weston/libweston/compositor.c

5467 lines
145 KiB
C
Raw Normal View History

2008-12-02 23:15:01 +03:00
/*
* Copyright © 2010-2011 Intel Corporation
* Copyright © 2008-2011 Kristian Høgsberg
* Copyright © 2012-2015 Collabora, Ltd.
2008-12-02 23:15:01 +03:00
*
* Permission is hereby granted, free of charge, to any person obtaining
* a copy of this software and associated documentation files (the
* "Software"), to deal in the Software without restriction, including
* without limitation the rights to use, copy, modify, merge, publish,
* distribute, sublicense, and/or sell copies of the Software, and to
* permit persons to whom the Software is furnished to do so, subject to
* the following conditions:
2008-12-02 23:15:01 +03:00
*
* The above copyright notice and this permission notice (including the
* next paragraph) shall be included in all copies or substantial
* portions of the Software.
*
* THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND,
* EXPRESS OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF
* MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND
* NONINFRINGEMENT. IN NO EVENT SHALL THE AUTHORS OR COPYRIGHT HOLDERS
* BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN
* ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN
* CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE
* SOFTWARE.
2008-12-02 23:15:01 +03:00
*/
#include "config.h"
#include <fcntl.h>
#include <stdio.h>
#include <string.h>
#include <stdlib.h>
#include <stdint.h>
#include <limits.h>
2008-11-09 08:22:51 +03:00
#include <stdarg.h>
#include <assert.h>
#include <sys/ioctl.h>
#include <sys/mman.h>
#include <sys/wait.h>
#include <sys/socket.h>
#include <sys/utsname.h>
#include <sys/stat.h>
#include <unistd.h>
#include <math.h>
#include <linux/input.h>
#include <dlfcn.h>
#include <signal.h>
#include <setjmp.h>
#include <sys/time.h>
#include <time.h>
#include <errno.h>
compositor: Implement JSON-timeline logging Logging is activated and deactivated with the debug key binding 't'. When activated, it creates a new log file, where it records the events. The log file contains events and detailed object information entries in JSON format, and is meant to be parsed in sequence from beginning to the end. The emitted events are mostly related to the output repaint cycle, like when repaint begins, is submitted to GPU, and when it completes on a vblank. This is recorded per-output. Also some per-surface events are recorded, including when surface damage is flushed. To reduce the log size, events refer to objects like outputs and surfaces by id numbers. Detailed object information is emitted only as needed: on the first object occurrence, and afterwards only if weston_timeline_object::force_refresh asks for it. The detailed information for surfaces includes the string returned by weston_surface::get_label. Therefore it is important to set weston_timeline_object::force_refresh = 1 whenever the string would change, so that the new details get recorded. A rudimentary parser and SVG generator can be found at: https://github.com/ppaalanen/wesgr The timeline logs can answer questions including: - How does the compositor repaint cycle work timing-wise? - When was the vblank deadline missed? - What is the latency from surface commit to showing the new content on screen? - How long does it take to process the scenegraph? v2: weston_surface::get_description renamed to get_label. v3: reafctor a bit into fprint_quoted_string(). Signed-off-by: Pekka Paalanen <pekka.paalanen@collabora.co.uk>
2014-11-12 16:09:24 +03:00
#include "timeline.h"
#include "compositor.h"
#include "viewporter-server-protocol.h"
#include "presentation-time-server-protocol.h"
#include "shared/helpers.h"
#include "shared/os-compatibility.h"
#include "shared/string-helpers.h"
#include "shared/timespec-util.h"
#include "git-version.h"
#include "version.h"
#include "plugin-registry.h"
compositor: add repaint delay timer This timer delays the output_repaint towards the end of the refresh period, reducing the time from repaint to present. The length of the repaint window can be set in weston.ini. The call to weston_output_schedule_repaint_reset() is delayed by one more period. If we exit the continuous repaint loop (set output->repaint_scheduled to false) in finish_frame, we may call start_repaint_loop() unnecessarily. The problem case was actually observed with two outputs on the DRM backend at 60 Hz, and 7 ms repaint-window. During a window move, one output was constantly falling off the continuous repaint loop and introducing additional one frame latency, leading to jerky window motion. This code now avoids the problem. Changes in v2: - Rename repaint_delay_timer to repaint_timer and output_repaint_delay_handler to output_repaint_timer_handler. - When computing the delay, take the current time into account. The timer uses a relative timeout, so we have to subtract any time already gone. Note, that 'gone' may also be negative. DRM has a habit of predicting the page flip timestamp so it may be still in the future when we get the completion event. - Do also a sanity check 'msec > 1000'. In the unlikely case that something fails to provide a good timestamp, never delay for more than one second. Signed-off-by: Pekka Paalanen <pekka.paalanen@collabora.co.uk> Reviewed-By: Derek Foreman <derekf@osg.samsung.com> Reviewed-by: Bryce Harrington <bryce@osg.samsung.com>
2014-05-21 17:17:27 +04:00
#define DEFAULT_REPAINT_WINDOW 7 /* milliseconds */
static void
weston_output_transform_scale_init(struct weston_output *output,
uint32_t transform, uint32_t scale);
static void
Split the geometry information from weston_surface out into weston_view The weston_surface structure is split into two structures: * The weston_surface structure storres everything required for a client-side or server-side surface. This includes buffers; callbacks; backend private data; input, damage, and opaque regions; and a few other bookkeeping bits. * The weston_view structure represents an entity in the scenegraph and storres all of the geometry information. This includes clip region, alpha, position, and the transformation list as well as all of the temporary information derived from the geometry state. Because a view, and not a surface, is a scenegraph element, the view is what is placed in layers and planes. There are a few things worth noting about the surface/view split: 1. This is *not* a modification to the protocol. It is, instead, a modification to Weston's internal scenegraph to allow a single surface to exist in multiple places at a time. Clients are completely unaware of how many views to a particular surface exist. 2. A view is considered a direct child of a surface and is destroyed when the surface is destroyed. Because of this, the view.surface pointer is always valid and non-null. 3. The compositor's surface_list is replaced with a view_list. Due to subsurfaces, building the view list is a little more complicated than it used to be and involves building a tree of views on the fly whenever subsurfaces are used. However, this means that backends can remain completely subsurface-agnostic. 4. Surfaces and views both keep track of which outputs they are on. 5. The weston_surface structure now has width and height fields. These are populated when a new buffer is attached before surface.configure is called. This is because there are many surface-based operations that really require the width and height and digging through the views didn't work well. Signed-off-by: Jason Ekstrand <jason@jlekstrand.net>
2013-10-13 07:38:11 +04:00
weston_compositor_build_view_list(struct weston_compositor *compositor);
static void weston_mode_switch_finish(struct weston_output *output,
int mode_changed,
int scale_changed)
{
struct weston_seat *seat;
struct wl_resource *resource;
pixman_region32_t old_output_region;
int version;
pixman_region32_init(&old_output_region);
pixman_region32_copy(&old_output_region, &output->region);
/* Update output region and transformation matrix */
weston_output_transform_scale_init(output, output->transform, output->current_scale);
pixman_region32_init(&output->previous_damage);
pixman_region32_init_rect(&output->region, output->x, output->y,
output->width, output->height);
weston_output_update_matrix(output);
/* If a pointer falls outside the outputs new geometry, move it to its
* lower-right corner */
wl_list_for_each(seat, &output->compositor->seat_list, link) {
struct weston_pointer *pointer = weston_seat_get_pointer(seat);
int32_t x, y;
if (!pointer)
continue;
x = wl_fixed_to_int(pointer->x);
y = wl_fixed_to_int(pointer->y);
if (!pixman_region32_contains_point(&old_output_region,
x, y, NULL) ||
pixman_region32_contains_point(&output->region,
x, y, NULL))
continue;
if (x >= output->x + output->width)
x = output->x + output->width - 1;
if (y >= output->y + output->height)
y = output->y + output->height - 1;
pointer->x = wl_fixed_from_int(x);
pointer->y = wl_fixed_from_int(y);
}
pixman_region32_fini(&old_output_region);
if (!mode_changed && !scale_changed)
return;
/* notify clients of the changes */
wl_resource_for_each(resource, &output->resource_list) {
if (mode_changed) {
wl_output_send_mode(resource,
output->current_mode->flags,
output->current_mode->width,
output->current_mode->height,
output->current_mode->refresh);
}
version = wl_resource_get_version(resource);
if (version >= WL_OUTPUT_SCALE_SINCE_VERSION && scale_changed)
wl_output_send_scale(resource, output->current_scale);
if (version >= WL_OUTPUT_DONE_SINCE_VERSION)
wl_output_send_done(resource);
}
}
static void
weston_compositor_reflow_outputs(struct weston_compositor *compositor,
struct weston_output *resized_output, int delta_width);
WL_EXPORT int
weston_output_mode_set_native(struct weston_output *output,
struct weston_mode *mode,
int32_t scale)
{
int ret;
int mode_changed = 0, scale_changed = 0;
int32_t old_width;
if (!output->switch_mode)
return -1;
if (!output->original_mode) {
mode_changed = 1;
ret = output->switch_mode(output, mode);
if (ret < 0)
return ret;
if (output->current_scale != scale) {
scale_changed = 1;
output->current_scale = scale;
}
}
old_width = output->width;
output->native_mode = mode;
output->native_scale = scale;
weston_mode_switch_finish(output, mode_changed, scale_changed);
if (mode_changed || scale_changed) {
weston_compositor_reflow_outputs(output->compositor, output, output->width - old_width);
wl_signal_emit(&output->compositor->output_resized_signal, output);
}
return 0;
}
WL_EXPORT int
weston_output_mode_switch_to_native(struct weston_output *output)
{
int ret;
int mode_changed = 0, scale_changed = 0;
if (!output->switch_mode)
return -1;
if (!output->original_mode) {
weston_log("already in the native mode\n");
return -1;
}
/* the non fullscreen clients haven't seen a mode set since we
* switched into a temporary, so we need to notify them if the
* mode at that time is different from the native mode now.
*/
mode_changed = (output->original_mode != output->native_mode);
scale_changed = (output->original_scale != output->native_scale);
ret = output->switch_mode(output, output->native_mode);
if (ret < 0)
return ret;
output->current_scale = output->native_scale;
output->original_mode = NULL;
output->original_scale = 0;
weston_mode_switch_finish(output, mode_changed, scale_changed);
return 0;
}
WL_EXPORT int
weston_output_mode_switch_to_temporary(struct weston_output *output,
struct weston_mode *mode,
int32_t scale)
{
int ret;
if (!output->switch_mode)
return -1;
/* original_mode is the last mode non full screen clients have seen,
* so we shouldn't change it if we already have one set.
*/
if (!output->original_mode) {
output->original_mode = output->native_mode;
output->original_scale = output->native_scale;
}
ret = output->switch_mode(output, mode);
if (ret < 0)
return ret;
output->current_scale = scale;
weston_mode_switch_finish(output, 0, 0);
return 0;
}
static void
region_init_infinite(pixman_region32_t *region)
{
pixman_region32_init_rect(region, INT32_MIN, INT32_MIN,
UINT32_MAX, UINT32_MAX);
}
compositor: introduce sub-surfaces Implement the basic protocol for sub-surfaces: - expose wl_subcompositor global interface - error checking on protocol calls - associate a parent wl_surface to a sub-surface - introduce the sub-surface role, which is exclusive - an implementation of the wl_subsurface interface - allow nesting of sub-surfaces - proper surface transformation inheritance from parent to sub-surfaces - two different modes of wl_surface.commit for sub-surfaces - hook sub-surfaces up to repaint by modifying the repaint list code Struct weston_subsurface is dynamically allocated. For sub-surfaces, it is completely populated. For parent surfaces, weston_subsurface acts only as a link for stacking order purposes. The wl_resource is unused, parent_destroy_listener is not registered, the transform is not linked, etc. Sub-surfaces are not added directly into layers for display or input. Instead, they are hooked up via the sub-surface list present in parent weston_surface. This way sub-surfaces are inherently linked to the parent surface, and cannot be displayed unless the parent is mapped, too. This also eases restacking, as only the parent will be in a layer list. Also, only the main surface should be subject to shell actions. The surface list rebuilding in weston_output_repaint() is modified to process sub-surface lists, if they are non-empty. The sub-surface list always contains the parent, too, unless empty. The collection of frame_callback_list is moved to a later loop, to streamline the surface list rebuild functions. Features still lacking are: - full-surface alpha support for compound windows Changes in v2: - fix a bug in surface mapping: commit a sub-surface would cause the main surface to never be mapped. - remove debug printfs - detect attempt of making a surface its own parent - always zero-alloc weston_subsurface - apply wl_subsurface.set_position in commit, not immediately - add weston_surface_to_subsurface() - implement sub-surface commit modes parent-cached and independent - implement wl_subcompositor.destroy and wl_subsurface.destroy Changes in v3: - rebased, and use the new transform inheritance code - squashed the commit "add sub-surfaces to repaint list" - fixed a buffer reference leak in commit_from_cache() - Rewrite the sub-surface destructor code, and make it leave the wl_subsurface protocol object inert, if one destroys the corresponding wl_surface. - replaced set_commit_mode with set_sync and set_desync - allowed sub-surface nesting, and fixed repaint accordingly - implemented nested sub-surface commit modes - Made the sub-surface order changes from wl_subsurface.place_above and .place_below to be applied when the parent surface state is applied, instead of immediately. This conforms with the protocol specification now. Signed-off-by: Pekka Paalanen <ppaalanen@gmail.com>
2013-04-25 14:57:42 +04:00
static struct weston_subsurface *
weston_surface_to_subsurface(struct weston_surface *surface);
Split the geometry information from weston_surface out into weston_view The weston_surface structure is split into two structures: * The weston_surface structure storres everything required for a client-side or server-side surface. This includes buffers; callbacks; backend private data; input, damage, and opaque regions; and a few other bookkeeping bits. * The weston_view structure represents an entity in the scenegraph and storres all of the geometry information. This includes clip region, alpha, position, and the transformation list as well as all of the temporary information derived from the geometry state. Because a view, and not a surface, is a scenegraph element, the view is what is placed in layers and planes. There are a few things worth noting about the surface/view split: 1. This is *not* a modification to the protocol. It is, instead, a modification to Weston's internal scenegraph to allow a single surface to exist in multiple places at a time. Clients are completely unaware of how many views to a particular surface exist. 2. A view is considered a direct child of a surface and is destroyed when the surface is destroyed. Because of this, the view.surface pointer is always valid and non-null. 3. The compositor's surface_list is replaced with a view_list. Due to subsurfaces, building the view list is a little more complicated than it used to be and involves building a tree of views on the fly whenever subsurfaces are used. However, this means that backends can remain completely subsurface-agnostic. 4. Surfaces and views both keep track of which outputs they are on. 5. The weston_surface structure now has width and height fields. These are populated when a new buffer is attached before surface.configure is called. This is because there are many surface-based operations that really require the width and height and digging through the views didn't work well. Signed-off-by: Jason Ekstrand <jason@jlekstrand.net>
2013-10-13 07:38:11 +04:00
WL_EXPORT struct weston_view *
weston_view_create(struct weston_surface *surface)
{
struct weston_view *view;
view = zalloc(sizeof *view);
Split the geometry information from weston_surface out into weston_view The weston_surface structure is split into two structures: * The weston_surface structure storres everything required for a client-side or server-side surface. This includes buffers; callbacks; backend private data; input, damage, and opaque regions; and a few other bookkeeping bits. * The weston_view structure represents an entity in the scenegraph and storres all of the geometry information. This includes clip region, alpha, position, and the transformation list as well as all of the temporary information derived from the geometry state. Because a view, and not a surface, is a scenegraph element, the view is what is placed in layers and planes. There are a few things worth noting about the surface/view split: 1. This is *not* a modification to the protocol. It is, instead, a modification to Weston's internal scenegraph to allow a single surface to exist in multiple places at a time. Clients are completely unaware of how many views to a particular surface exist. 2. A view is considered a direct child of a surface and is destroyed when the surface is destroyed. Because of this, the view.surface pointer is always valid and non-null. 3. The compositor's surface_list is replaced with a view_list. Due to subsurfaces, building the view list is a little more complicated than it used to be and involves building a tree of views on the fly whenever subsurfaces are used. However, this means that backends can remain completely subsurface-agnostic. 4. Surfaces and views both keep track of which outputs they are on. 5. The weston_surface structure now has width and height fields. These are populated when a new buffer is attached before surface.configure is called. This is because there are many surface-based operations that really require the width and height and digging through the views didn't work well. Signed-off-by: Jason Ekstrand <jason@jlekstrand.net>
2013-10-13 07:38:11 +04:00
if (view == NULL)
return NULL;
view->surface = surface;
compositor: Assign new views to the primary plane When we create a new view, assign it to the primary plane from the beginning. Currently, every view across the compositor will be assigned to a plane during every repaint cycle of every output: the DRM renderer's assign_planes hook will either move a view to a drm_plane, or to the primary plane if a suitable drm_plane could not be found for the output it is on. There are no other assign_planes implementation, and the fallback when none is provided, is to assign every view to the primary plane. DRM's behaviour is undesirable in multi-output situations, since it means that views which were on a plane on one output will be demoted to the primary plane; doing this causes damage, which will cause a spurious repaint for the output. This spurious repaint will have no effect on the other output, but it will do the same demotion of views to the primary plane, which will again provoke a repaint on the other output. With a simple fix for this behaviour (i.e. not moving views which are only visible on other outputs), the following behaviour is observed: - outputs A and B are present - views A and B are created for those outputs respectively, with SHM buffers attached; view->plane == NULL for both - current buffer content for views A and B are uploaded to the renderer - output A runs its repaint cycle, and sets keep_buffer to false on surface B's output, as it can never be promoted to a plane; it does not move view B to another plane - output B runs its repaint cycle, and moves view B to the primary plane - weston_view_assign_to_plane has work to do (as the plane is changing from NULL to the primary plane), calls weston_surface_damage and calls weston_surface_damage - weston_surface_damage re-uploads buffer content, possibly from nowhere at all; e508ce6a notes that this behaviour is broken Assigning views to the primary plane when created makes it possible to fix the DRM assign_planes implementation: assign_planes will always set keep_buffer to true if there is any chance the buffer can ever be promoted to a plane, regardless of view configruation. If the buffer cannot be promoted to a plane, it must by definition never migrate from the primary plane. This means that there is no opportunity to hit the same issue, where the buffer content has already been discarded, but weston_view_assign_to_plane is not a no-op. Signed-off-by: Daniel Stone <daniels@collabora.com> Reviewed: Pekka Paalanen <pekka.paalanen@collabora.co.uk>
2016-12-09 19:27:54 +03:00
view->plane = &surface->compositor->primary_plane;
Split the geometry information from weston_surface out into weston_view The weston_surface structure is split into two structures: * The weston_surface structure storres everything required for a client-side or server-side surface. This includes buffers; callbacks; backend private data; input, damage, and opaque regions; and a few other bookkeeping bits. * The weston_view structure represents an entity in the scenegraph and storres all of the geometry information. This includes clip region, alpha, position, and the transformation list as well as all of the temporary information derived from the geometry state. Because a view, and not a surface, is a scenegraph element, the view is what is placed in layers and planes. There are a few things worth noting about the surface/view split: 1. This is *not* a modification to the protocol. It is, instead, a modification to Weston's internal scenegraph to allow a single surface to exist in multiple places at a time. Clients are completely unaware of how many views to a particular surface exist. 2. A view is considered a direct child of a surface and is destroyed when the surface is destroyed. Because of this, the view.surface pointer is always valid and non-null. 3. The compositor's surface_list is replaced with a view_list. Due to subsurfaces, building the view list is a little more complicated than it used to be and involves building a tree of views on the fly whenever subsurfaces are used. However, this means that backends can remain completely subsurface-agnostic. 4. Surfaces and views both keep track of which outputs they are on. 5. The weston_surface structure now has width and height fields. These are populated when a new buffer is attached before surface.configure is called. This is because there are many surface-based operations that really require the width and height and digging through the views didn't work well. Signed-off-by: Jason Ekstrand <jason@jlekstrand.net>
2013-10-13 07:38:11 +04:00
/* Assign to surface */
wl_list_insert(&surface->views, &view->surface_link);
wl_signal_init(&view->destroy_signal);
wl_list_init(&view->link);
wl_list_init(&view->layer_link.link);
Split the geometry information from weston_surface out into weston_view The weston_surface structure is split into two structures: * The weston_surface structure storres everything required for a client-side or server-side surface. This includes buffers; callbacks; backend private data; input, damage, and opaque regions; and a few other bookkeeping bits. * The weston_view structure represents an entity in the scenegraph and storres all of the geometry information. This includes clip region, alpha, position, and the transformation list as well as all of the temporary information derived from the geometry state. Because a view, and not a surface, is a scenegraph element, the view is what is placed in layers and planes. There are a few things worth noting about the surface/view split: 1. This is *not* a modification to the protocol. It is, instead, a modification to Weston's internal scenegraph to allow a single surface to exist in multiple places at a time. Clients are completely unaware of how many views to a particular surface exist. 2. A view is considered a direct child of a surface and is destroyed when the surface is destroyed. Because of this, the view.surface pointer is always valid and non-null. 3. The compositor's surface_list is replaced with a view_list. Due to subsurfaces, building the view list is a little more complicated than it used to be and involves building a tree of views on the fly whenever subsurfaces are used. However, this means that backends can remain completely subsurface-agnostic. 4. Surfaces and views both keep track of which outputs they are on. 5. The weston_surface structure now has width and height fields. These are populated when a new buffer is attached before surface.configure is called. This is because there are many surface-based operations that really require the width and height and digging through the views didn't work well. Signed-off-by: Jason Ekstrand <jason@jlekstrand.net>
2013-10-13 07:38:11 +04:00
pixman_region32_init(&view->clip);
view->alpha = 1.0;
pixman_region32_init(&view->transform.opaque);
wl_list_init(&view->geometry.transformation_list);
wl_list_insert(&view->geometry.transformation_list,
&view->transform.position.link);
weston_matrix_init(&view->transform.position.matrix);
wl_list_init(&view->geometry.child_list);
compositor: add weston_view_set_mask() API and state Add API for setting a clip ('scissor' in the code) rectangle per view, in surface coordinates. Ivi-shell requires this feature to be able to implement the IVI Layer Manager API, which includes clipping of surfaces. The names weston_view_set_mask() and weston_view_set_mask_infinite() mirror the existing weston_layer_set_mask*() functions. This view clipping complements the weston_layer clipping, because view clipping is defined in surface local coordinates, while layer mask/clipping is defined in global coordinates. View clipping requires explicit support from the renderers. Therefore a new Weston capability bit is added: WESTON_CAP_VIEW_CLIP_MASK. Shells (and all users) of this new API are required to check the capability bit is set before using the API. Otherwise the rendering will not be what they expect. View clips are inherited through the transformation inheritance mechanism. However, there are restrictions. The clip rectangle can be set only on the root view of a transformation inheritance tree. The additional transformations in child views must not rotate the coordinate axes. These restrictions avoid corner cases in clip inheritance, and keep the renderer implementations as simple as they are right now. Renderers only need to do an additional intersection with the clip rectangle which is always aligned to the surface coordinate system. For more details, see the API documentation in the patch. Signed-off-by: Pekka Paalanen <pekka.paalanen@collabora.co.uk> Reviewed-by: Nobuhiko Tanibata <NOBUHIKO_TANIBATA@xddp.denso.co.jp> Tested-by: Nobuhiko Tanibata <NOBUHIKO_TANIBATA@xddp.denso.co.jp> Reviewed-By: Derek Foreman <derekf@osg.samsung.com>
2015-02-16 15:39:11 +03:00
pixman_region32_init(&view->geometry.scissor);
Split the geometry information from weston_surface out into weston_view The weston_surface structure is split into two structures: * The weston_surface structure storres everything required for a client-side or server-side surface. This includes buffers; callbacks; backend private data; input, damage, and opaque regions; and a few other bookkeeping bits. * The weston_view structure represents an entity in the scenegraph and storres all of the geometry information. This includes clip region, alpha, position, and the transformation list as well as all of the temporary information derived from the geometry state. Because a view, and not a surface, is a scenegraph element, the view is what is placed in layers and planes. There are a few things worth noting about the surface/view split: 1. This is *not* a modification to the protocol. It is, instead, a modification to Weston's internal scenegraph to allow a single surface to exist in multiple places at a time. Clients are completely unaware of how many views to a particular surface exist. 2. A view is considered a direct child of a surface and is destroyed when the surface is destroyed. Because of this, the view.surface pointer is always valid and non-null. 3. The compositor's surface_list is replaced with a view_list. Due to subsurfaces, building the view list is a little more complicated than it used to be and involves building a tree of views on the fly whenever subsurfaces are used. However, this means that backends can remain completely subsurface-agnostic. 4. Surfaces and views both keep track of which outputs they are on. 5. The weston_surface structure now has width and height fields. These are populated when a new buffer is attached before surface.configure is called. This is because there are many surface-based operations that really require the width and height and digging through the views didn't work well. Signed-off-by: Jason Ekstrand <jason@jlekstrand.net>
2013-10-13 07:38:11 +04:00
pixman_region32_init(&view->transform.boundingbox);
view->transform.dirty = 1;
return view;
}
struct weston_frame_callback {
struct wl_resource *resource;
struct wl_list link;
};
struct weston_presentation_feedback {
struct wl_resource *resource;
/* XXX: could use just wl_resource_get_link() instead */
struct wl_list link;
/* The per-surface feedback flags */
uint32_t psf_flags;
};
static void
weston_presentation_feedback_discard(
struct weston_presentation_feedback *feedback)
{
wp_presentation_feedback_send_discarded(feedback->resource);
wl_resource_destroy(feedback->resource);
}
static void
weston_presentation_feedback_discard_list(struct wl_list *list)
{
struct weston_presentation_feedback *feedback, *tmp;
wl_list_for_each_safe(feedback, tmp, list, link)
weston_presentation_feedback_discard(feedback);
}
static void
weston_presentation_feedback_present(
struct weston_presentation_feedback *feedback,
struct weston_output *output,
uint32_t refresh_nsec,
const struct timespec *ts,
compositor: set presentation.presented flags Change weston_output_finish_frame() signature so that backends are required to set the flags, that will be reported on the Presentation 'presented' event. This is meant for output-wide feedback flags. Flags that vary per wl_surface are subject for the following patch. All start_repaint_loop functions use the special private flag PRESENTATION_FEEDBACK_INVALID to mark, that this call of weston_output_finish_frame() cannot trigger the 'presented' event. If it does, we now hit an assert, and should then investigate why a fake update triggered Presentation feedback. DRM: Page flip is always vsync'd, and always gets the completion timestamp from the kernel which should correspond well to hardware. Completion is triggered by the kernel/hardware. Vblank handler is only used with the broken planes path, therefore do not report VSYNC, because we cannot guarantee all the planes updated at the same time. We cannot set the INVALID, because it would abort the compositor if the broken planes path was ever used. This is a hack that will get fixed with nuclear pageflip support in the future. fbdev: No vsync, update done by copy, no completion event from hardware, and completion time is totally fake. headless: No real output to update. RDP: Guessing that maybe no vsync, fake time, and copy make sense (pixels sent over network). Also no event that the pixels have been shown? RPI: Presumably Dispmanx updates are vsync'd. We get a completion event from the driver, but need to read the clock ourselves, so the completion time is somewhat unreliable. Zero-copy flag not implemented though it would be theoretically possible with EGL clients (zero-copy is a per-surface flag anyway, so in this patch). Wayland: No information how the host compositor is doing updates, so make a safe guess without assuming vsync or hardware completion event. While we do get some timestamp from the host compositor, it is not the completion time. Would need to hook to the Presentation extension of the host compositor to get more accurate flags. X11: No idea about vsync, completion event, or copying. Also the timestamp is a fake. Signed-off-by: Pekka Paalanen <pekka.paalanen@collabora.co.uk> Reviewed-by: Mario Kleiner <mario.kleiner.de@gmail.com> Tested-by: Mario Kleiner <mario.kleiner.de@gmail.com> Acked-by: Mario Kleiner <mario.kleiner.de@gmail.com>
2014-12-17 17:20:40 +03:00
uint64_t seq,
uint32_t flags)
{
struct wl_client *client = wl_resource_get_client(feedback->resource);
struct wl_resource *o;
uint64_t secs;
wl_resource_for_each(o, &output->resource_list) {
if (wl_resource_get_client(o) != client)
continue;
wp_presentation_feedback_send_sync_output(feedback->resource, o);
}
secs = ts->tv_sec;
wp_presentation_feedback_send_presented(feedback->resource,
secs >> 32, secs & 0xffffffff,
ts->tv_nsec,
refresh_nsec,
seq >> 32, seq & 0xffffffff,
flags | feedback->psf_flags);
wl_resource_destroy(feedback->resource);
}
static void
weston_presentation_feedback_present_list(struct wl_list *list,
struct weston_output *output,
uint32_t refresh_nsec,
const struct timespec *ts,
compositor: set presentation.presented flags Change weston_output_finish_frame() signature so that backends are required to set the flags, that will be reported on the Presentation 'presented' event. This is meant for output-wide feedback flags. Flags that vary per wl_surface are subject for the following patch. All start_repaint_loop functions use the special private flag PRESENTATION_FEEDBACK_INVALID to mark, that this call of weston_output_finish_frame() cannot trigger the 'presented' event. If it does, we now hit an assert, and should then investigate why a fake update triggered Presentation feedback. DRM: Page flip is always vsync'd, and always gets the completion timestamp from the kernel which should correspond well to hardware. Completion is triggered by the kernel/hardware. Vblank handler is only used with the broken planes path, therefore do not report VSYNC, because we cannot guarantee all the planes updated at the same time. We cannot set the INVALID, because it would abort the compositor if the broken planes path was ever used. This is a hack that will get fixed with nuclear pageflip support in the future. fbdev: No vsync, update done by copy, no completion event from hardware, and completion time is totally fake. headless: No real output to update. RDP: Guessing that maybe no vsync, fake time, and copy make sense (pixels sent over network). Also no event that the pixels have been shown? RPI: Presumably Dispmanx updates are vsync'd. We get a completion event from the driver, but need to read the clock ourselves, so the completion time is somewhat unreliable. Zero-copy flag not implemented though it would be theoretically possible with EGL clients (zero-copy is a per-surface flag anyway, so in this patch). Wayland: No information how the host compositor is doing updates, so make a safe guess without assuming vsync or hardware completion event. While we do get some timestamp from the host compositor, it is not the completion time. Would need to hook to the Presentation extension of the host compositor to get more accurate flags. X11: No idea about vsync, completion event, or copying. Also the timestamp is a fake. Signed-off-by: Pekka Paalanen <pekka.paalanen@collabora.co.uk> Reviewed-by: Mario Kleiner <mario.kleiner.de@gmail.com> Tested-by: Mario Kleiner <mario.kleiner.de@gmail.com> Acked-by: Mario Kleiner <mario.kleiner.de@gmail.com>
2014-12-17 17:20:40 +03:00
uint64_t seq,
uint32_t flags)
{
struct weston_presentation_feedback *feedback, *tmp;
assert(!(flags & WP_PRESENTATION_FEEDBACK_INVALID) ||
compositor: set presentation.presented flags Change weston_output_finish_frame() signature so that backends are required to set the flags, that will be reported on the Presentation 'presented' event. This is meant for output-wide feedback flags. Flags that vary per wl_surface are subject for the following patch. All start_repaint_loop functions use the special private flag PRESENTATION_FEEDBACK_INVALID to mark, that this call of weston_output_finish_frame() cannot trigger the 'presented' event. If it does, we now hit an assert, and should then investigate why a fake update triggered Presentation feedback. DRM: Page flip is always vsync'd, and always gets the completion timestamp from the kernel which should correspond well to hardware. Completion is triggered by the kernel/hardware. Vblank handler is only used with the broken planes path, therefore do not report VSYNC, because we cannot guarantee all the planes updated at the same time. We cannot set the INVALID, because it would abort the compositor if the broken planes path was ever used. This is a hack that will get fixed with nuclear pageflip support in the future. fbdev: No vsync, update done by copy, no completion event from hardware, and completion time is totally fake. headless: No real output to update. RDP: Guessing that maybe no vsync, fake time, and copy make sense (pixels sent over network). Also no event that the pixels have been shown? RPI: Presumably Dispmanx updates are vsync'd. We get a completion event from the driver, but need to read the clock ourselves, so the completion time is somewhat unreliable. Zero-copy flag not implemented though it would be theoretically possible with EGL clients (zero-copy is a per-surface flag anyway, so in this patch). Wayland: No information how the host compositor is doing updates, so make a safe guess without assuming vsync or hardware completion event. While we do get some timestamp from the host compositor, it is not the completion time. Would need to hook to the Presentation extension of the host compositor to get more accurate flags. X11: No idea about vsync, completion event, or copying. Also the timestamp is a fake. Signed-off-by: Pekka Paalanen <pekka.paalanen@collabora.co.uk> Reviewed-by: Mario Kleiner <mario.kleiner.de@gmail.com> Tested-by: Mario Kleiner <mario.kleiner.de@gmail.com> Acked-by: Mario Kleiner <mario.kleiner.de@gmail.com>
2014-12-17 17:20:40 +03:00
wl_list_empty(list));
wl_list_for_each_safe(feedback, tmp, list, link)
weston_presentation_feedback_present(feedback, output,
compositor: set presentation.presented flags Change weston_output_finish_frame() signature so that backends are required to set the flags, that will be reported on the Presentation 'presented' event. This is meant for output-wide feedback flags. Flags that vary per wl_surface are subject for the following patch. All start_repaint_loop functions use the special private flag PRESENTATION_FEEDBACK_INVALID to mark, that this call of weston_output_finish_frame() cannot trigger the 'presented' event. If it does, we now hit an assert, and should then investigate why a fake update triggered Presentation feedback. DRM: Page flip is always vsync'd, and always gets the completion timestamp from the kernel which should correspond well to hardware. Completion is triggered by the kernel/hardware. Vblank handler is only used with the broken planes path, therefore do not report VSYNC, because we cannot guarantee all the planes updated at the same time. We cannot set the INVALID, because it would abort the compositor if the broken planes path was ever used. This is a hack that will get fixed with nuclear pageflip support in the future. fbdev: No vsync, update done by copy, no completion event from hardware, and completion time is totally fake. headless: No real output to update. RDP: Guessing that maybe no vsync, fake time, and copy make sense (pixels sent over network). Also no event that the pixels have been shown? RPI: Presumably Dispmanx updates are vsync'd. We get a completion event from the driver, but need to read the clock ourselves, so the completion time is somewhat unreliable. Zero-copy flag not implemented though it would be theoretically possible with EGL clients (zero-copy is a per-surface flag anyway, so in this patch). Wayland: No information how the host compositor is doing updates, so make a safe guess without assuming vsync or hardware completion event. While we do get some timestamp from the host compositor, it is not the completion time. Would need to hook to the Presentation extension of the host compositor to get more accurate flags. X11: No idea about vsync, completion event, or copying. Also the timestamp is a fake. Signed-off-by: Pekka Paalanen <pekka.paalanen@collabora.co.uk> Reviewed-by: Mario Kleiner <mario.kleiner.de@gmail.com> Tested-by: Mario Kleiner <mario.kleiner.de@gmail.com> Acked-by: Mario Kleiner <mario.kleiner.de@gmail.com>
2014-12-17 17:20:40 +03:00
refresh_nsec, ts, seq,
flags);
}
static void
surface_state_handle_buffer_destroy(struct wl_listener *listener, void *data)
{
struct weston_surface_state *state =
container_of(listener, struct weston_surface_state,
buffer_destroy_listener);
state->buffer = NULL;
}
static void
weston_surface_state_init(struct weston_surface_state *state)
{
state->newly_attached = 0;
state->buffer = NULL;
state->buffer_destroy_listener.notify =
surface_state_handle_buffer_destroy;
state->sx = 0;
state->sy = 0;
pixman_region32_init(&state->damage_surface);
pixman_region32_init(&state->damage_buffer);
pixman_region32_init(&state->opaque);
region_init_infinite(&state->input);
wl_list_init(&state->frame_callback_list);
wl_list_init(&state->feedback_list);
state->buffer_viewport.buffer.transform = WL_OUTPUT_TRANSFORM_NORMAL;
state->buffer_viewport.buffer.scale = 1;
state->buffer_viewport.buffer.src_width = wl_fixed_from_int(-1);
state->buffer_viewport.surface.width = -1;
state->buffer_viewport.changed = 0;
}
static void
weston_surface_state_fini(struct weston_surface_state *state)
{
struct weston_frame_callback *cb, *next;
wl_list_for_each_safe(cb, next,
&state->frame_callback_list, link)
wl_resource_destroy(cb->resource);
weston_presentation_feedback_discard_list(&state->feedback_list);
pixman_region32_fini(&state->input);
pixman_region32_fini(&state->opaque);
pixman_region32_fini(&state->damage_surface);
pixman_region32_fini(&state->damage_buffer);
if (state->buffer)
wl_list_remove(&state->buffer_destroy_listener.link);
state->buffer = NULL;
}
static void
weston_surface_state_set_buffer(struct weston_surface_state *state,
struct weston_buffer *buffer)
{
if (state->buffer == buffer)
return;
if (state->buffer)
wl_list_remove(&state->buffer_destroy_listener.link);
state->buffer = buffer;
if (state->buffer)
wl_signal_add(&state->buffer->destroy_signal,
&state->buffer_destroy_listener);
}
WL_EXPORT struct weston_surface *
weston_surface_create(struct weston_compositor *compositor)
{
struct weston_surface *surface;
surface = zalloc(sizeof *surface);
if (surface == NULL)
return NULL;
wl_signal_init(&surface->destroy_signal);
wl_signal_init(&surface->commit_signal);
surface->compositor = compositor;
surface->ref_count = 1;
2010-06-07 06:23:21 +04:00
surface->buffer_viewport.buffer.transform = WL_OUTPUT_TRANSFORM_NORMAL;
surface->buffer_viewport.buffer.scale = 1;
surface->buffer_viewport.buffer.src_width = wl_fixed_from_int(-1);
surface->buffer_viewport.surface.width = -1;
weston_surface_state_init(&surface->pending);
pixman_region32_init(&surface->damage);
pixman_region32_init(&surface->opaque);
region_init_infinite(&surface->input);
Split the geometry information from weston_surface out into weston_view The weston_surface structure is split into two structures: * The weston_surface structure storres everything required for a client-side or server-side surface. This includes buffers; callbacks; backend private data; input, damage, and opaque regions; and a few other bookkeeping bits. * The weston_view structure represents an entity in the scenegraph and storres all of the geometry information. This includes clip region, alpha, position, and the transformation list as well as all of the temporary information derived from the geometry state. Because a view, and not a surface, is a scenegraph element, the view is what is placed in layers and planes. There are a few things worth noting about the surface/view split: 1. This is *not* a modification to the protocol. It is, instead, a modification to Weston's internal scenegraph to allow a single surface to exist in multiple places at a time. Clients are completely unaware of how many views to a particular surface exist. 2. A view is considered a direct child of a surface and is destroyed when the surface is destroyed. Because of this, the view.surface pointer is always valid and non-null. 3. The compositor's surface_list is replaced with a view_list. Due to subsurfaces, building the view list is a little more complicated than it used to be and involves building a tree of views on the fly whenever subsurfaces are used. However, this means that backends can remain completely subsurface-agnostic. 4. Surfaces and views both keep track of which outputs they are on. 5. The weston_surface structure now has width and height fields. These are populated when a new buffer is attached before surface.configure is called. This is because there are many surface-based operations that really require the width and height and digging through the views didn't work well. Signed-off-by: Jason Ekstrand <jason@jlekstrand.net>
2013-10-13 07:38:11 +04:00
wl_list_init(&surface->views);
wl_list_init(&surface->frame_callback_list);
wl_list_init(&surface->feedback_list);
compositor: introduce sub-surfaces Implement the basic protocol for sub-surfaces: - expose wl_subcompositor global interface - error checking on protocol calls - associate a parent wl_surface to a sub-surface - introduce the sub-surface role, which is exclusive - an implementation of the wl_subsurface interface - allow nesting of sub-surfaces - proper surface transformation inheritance from parent to sub-surfaces - two different modes of wl_surface.commit for sub-surfaces - hook sub-surfaces up to repaint by modifying the repaint list code Struct weston_subsurface is dynamically allocated. For sub-surfaces, it is completely populated. For parent surfaces, weston_subsurface acts only as a link for stacking order purposes. The wl_resource is unused, parent_destroy_listener is not registered, the transform is not linked, etc. Sub-surfaces are not added directly into layers for display or input. Instead, they are hooked up via the sub-surface list present in parent weston_surface. This way sub-surfaces are inherently linked to the parent surface, and cannot be displayed unless the parent is mapped, too. This also eases restacking, as only the parent will be in a layer list. Also, only the main surface should be subject to shell actions. The surface list rebuilding in weston_output_repaint() is modified to process sub-surface lists, if they are non-empty. The sub-surface list always contains the parent, too, unless empty. The collection of frame_callback_list is moved to a later loop, to streamline the surface list rebuild functions. Features still lacking are: - full-surface alpha support for compound windows Changes in v2: - fix a bug in surface mapping: commit a sub-surface would cause the main surface to never be mapped. - remove debug printfs - detect attempt of making a surface its own parent - always zero-alloc weston_subsurface - apply wl_subsurface.set_position in commit, not immediately - add weston_surface_to_subsurface() - implement sub-surface commit modes parent-cached and independent - implement wl_subcompositor.destroy and wl_subsurface.destroy Changes in v3: - rebased, and use the new transform inheritance code - squashed the commit "add sub-surfaces to repaint list" - fixed a buffer reference leak in commit_from_cache() - Rewrite the sub-surface destructor code, and make it leave the wl_subsurface protocol object inert, if one destroys the corresponding wl_surface. - replaced set_commit_mode with set_sync and set_desync - allowed sub-surface nesting, and fixed repaint accordingly - implemented nested sub-surface commit modes - Made the sub-surface order changes from wl_subsurface.place_above and .place_below to be applied when the parent surface state is applied, instead of immediately. This conforms with the protocol specification now. Signed-off-by: Pekka Paalanen <ppaalanen@gmail.com>
2013-04-25 14:57:42 +04:00
wl_list_init(&surface->subsurface_list);
wl_list_init(&surface->subsurface_list_pending);
weston_matrix_init(&surface->buffer_to_surface_matrix);
weston_matrix_init(&surface->surface_to_buffer_matrix);
wl_list_init(&surface->pointer_constraints);
return surface;
}
WL_EXPORT void
weston_surface_set_color(struct weston_surface *surface,
2012-09-30 04:57:21 +04:00
float red, float green, float blue, float alpha)
{
surface->compositor->renderer->surface_set_color(surface, red, green, blue, alpha);
}
WL_EXPORT void
Split the geometry information from weston_surface out into weston_view The weston_surface structure is split into two structures: * The weston_surface structure storres everything required for a client-side or server-side surface. This includes buffers; callbacks; backend private data; input, damage, and opaque regions; and a few other bookkeeping bits. * The weston_view structure represents an entity in the scenegraph and storres all of the geometry information. This includes clip region, alpha, position, and the transformation list as well as all of the temporary information derived from the geometry state. Because a view, and not a surface, is a scenegraph element, the view is what is placed in layers and planes. There are a few things worth noting about the surface/view split: 1. This is *not* a modification to the protocol. It is, instead, a modification to Weston's internal scenegraph to allow a single surface to exist in multiple places at a time. Clients are completely unaware of how many views to a particular surface exist. 2. A view is considered a direct child of a surface and is destroyed when the surface is destroyed. Because of this, the view.surface pointer is always valid and non-null. 3. The compositor's surface_list is replaced with a view_list. Due to subsurfaces, building the view list is a little more complicated than it used to be and involves building a tree of views on the fly whenever subsurfaces are used. However, this means that backends can remain completely subsurface-agnostic. 4. Surfaces and views both keep track of which outputs they are on. 5. The weston_surface structure now has width and height fields. These are populated when a new buffer is attached before surface.configure is called. This is because there are many surface-based operations that really require the width and height and digging through the views didn't work well. Signed-off-by: Jason Ekstrand <jason@jlekstrand.net>
2013-10-13 07:38:11 +04:00
weston_view_to_global_float(struct weston_view *view,
float sx, float sy, float *x, float *y)
{
Split the geometry information from weston_surface out into weston_view The weston_surface structure is split into two structures: * The weston_surface structure storres everything required for a client-side or server-side surface. This includes buffers; callbacks; backend private data; input, damage, and opaque regions; and a few other bookkeeping bits. * The weston_view structure represents an entity in the scenegraph and storres all of the geometry information. This includes clip region, alpha, position, and the transformation list as well as all of the temporary information derived from the geometry state. Because a view, and not a surface, is a scenegraph element, the view is what is placed in layers and planes. There are a few things worth noting about the surface/view split: 1. This is *not* a modification to the protocol. It is, instead, a modification to Weston's internal scenegraph to allow a single surface to exist in multiple places at a time. Clients are completely unaware of how many views to a particular surface exist. 2. A view is considered a direct child of a surface and is destroyed when the surface is destroyed. Because of this, the view.surface pointer is always valid and non-null. 3. The compositor's surface_list is replaced with a view_list. Due to subsurfaces, building the view list is a little more complicated than it used to be and involves building a tree of views on the fly whenever subsurfaces are used. However, this means that backends can remain completely subsurface-agnostic. 4. Surfaces and views both keep track of which outputs they are on. 5. The weston_surface structure now has width and height fields. These are populated when a new buffer is attached before surface.configure is called. This is because there are many surface-based operations that really require the width and height and digging through the views didn't work well. Signed-off-by: Jason Ekstrand <jason@jlekstrand.net>
2013-10-13 07:38:11 +04:00
if (view->transform.enabled) {
struct weston_vector v = { { sx, sy, 0.0f, 1.0f } };
Split the geometry information from weston_surface out into weston_view The weston_surface structure is split into two structures: * The weston_surface structure storres everything required for a client-side or server-side surface. This includes buffers; callbacks; backend private data; input, damage, and opaque regions; and a few other bookkeeping bits. * The weston_view structure represents an entity in the scenegraph and storres all of the geometry information. This includes clip region, alpha, position, and the transformation list as well as all of the temporary information derived from the geometry state. Because a view, and not a surface, is a scenegraph element, the view is what is placed in layers and planes. There are a few things worth noting about the surface/view split: 1. This is *not* a modification to the protocol. It is, instead, a modification to Weston's internal scenegraph to allow a single surface to exist in multiple places at a time. Clients are completely unaware of how many views to a particular surface exist. 2. A view is considered a direct child of a surface and is destroyed when the surface is destroyed. Because of this, the view.surface pointer is always valid and non-null. 3. The compositor's surface_list is replaced with a view_list. Due to subsurfaces, building the view list is a little more complicated than it used to be and involves building a tree of views on the fly whenever subsurfaces are used. However, this means that backends can remain completely subsurface-agnostic. 4. Surfaces and views both keep track of which outputs they are on. 5. The weston_surface structure now has width and height fields. These are populated when a new buffer is attached before surface.configure is called. This is because there are many surface-based operations that really require the width and height and digging through the views didn't work well. Signed-off-by: Jason Ekstrand <jason@jlekstrand.net>
2013-10-13 07:38:11 +04:00
weston_matrix_transform(&view->transform.matrix, &v);
if (fabsf(v.f[3]) < 1e-6) {
2012-06-07 20:01:59 +04:00
weston_log("warning: numerical instability in "
"%s(), divisor = %g\n", __func__,
v.f[3]);
*x = 0;
*y = 0;
return;
}
*x = v.f[0] / v.f[3];
*y = v.f[1] / v.f[3];
} else {
Split the geometry information from weston_surface out into weston_view The weston_surface structure is split into two structures: * The weston_surface structure storres everything required for a client-side or server-side surface. This includes buffers; callbacks; backend private data; input, damage, and opaque regions; and a few other bookkeeping bits. * The weston_view structure represents an entity in the scenegraph and storres all of the geometry information. This includes clip region, alpha, position, and the transformation list as well as all of the temporary information derived from the geometry state. Because a view, and not a surface, is a scenegraph element, the view is what is placed in layers and planes. There are a few things worth noting about the surface/view split: 1. This is *not* a modification to the protocol. It is, instead, a modification to Weston's internal scenegraph to allow a single surface to exist in multiple places at a time. Clients are completely unaware of how many views to a particular surface exist. 2. A view is considered a direct child of a surface and is destroyed when the surface is destroyed. Because of this, the view.surface pointer is always valid and non-null. 3. The compositor's surface_list is replaced with a view_list. Due to subsurfaces, building the view list is a little more complicated than it used to be and involves building a tree of views on the fly whenever subsurfaces are used. However, this means that backends can remain completely subsurface-agnostic. 4. Surfaces and views both keep track of which outputs they are on. 5. The weston_surface structure now has width and height fields. These are populated when a new buffer is attached before surface.configure is called. This is because there are many surface-based operations that really require the width and height and digging through the views didn't work well. Signed-off-by: Jason Ekstrand <jason@jlekstrand.net>
2013-10-13 07:38:11 +04:00
*x = sx + view->geometry.x;
*y = sy + view->geometry.y;
}
}
WL_EXPORT void
weston_transformed_coord(int width, int height,
enum wl_output_transform transform,
int32_t scale,
float sx, float sy, float *bx, float *by)
{
switch (transform) {
case WL_OUTPUT_TRANSFORM_NORMAL:
default:
*bx = sx;
*by = sy;
break;
case WL_OUTPUT_TRANSFORM_FLIPPED:
*bx = width - sx;
*by = sy;
break;
case WL_OUTPUT_TRANSFORM_90:
*bx = height - sy;
*by = sx;
break;
case WL_OUTPUT_TRANSFORM_FLIPPED_90:
*bx = height - sy;
*by = width - sx;
break;
case WL_OUTPUT_TRANSFORM_180:
*bx = width - sx;
*by = height - sy;
break;
case WL_OUTPUT_TRANSFORM_FLIPPED_180:
*bx = sx;
*by = height - sy;
break;
case WL_OUTPUT_TRANSFORM_270:
*bx = sy;
*by = width - sx;
break;
case WL_OUTPUT_TRANSFORM_FLIPPED_270:
*bx = sy;
*by = sx;
break;
}
*bx *= scale;
*by *= scale;
}
WL_EXPORT pixman_box32_t
weston_transformed_rect(int width, int height,
enum wl_output_transform transform,
int32_t scale,
pixman_box32_t rect)
{
float x1, x2, y1, y2;
pixman_box32_t ret;
weston_transformed_coord(width, height, transform, scale,
rect.x1, rect.y1, &x1, &y1);
weston_transformed_coord(width, height, transform, scale,
rect.x2, rect.y2, &x2, &y2);
if (x1 <= x2) {
ret.x1 = x1;
ret.x2 = x2;
} else {
ret.x1 = x2;
ret.x2 = x1;
}
if (y1 <= y2) {
ret.y1 = y1;
ret.y2 = y2;
} else {
ret.y1 = y2;
ret.y2 = y1;
}
return ret;
}
/** Transform a region by a matrix, restricted to axis-aligned transformations
*
* Warning: This function does not work for projective, affine, or matrices
* that encode arbitrary rotations. Only 90-degree step rotations are
* supported.
*/
WL_EXPORT void
weston_matrix_transform_region(pixman_region32_t *dest,
struct weston_matrix *matrix,
pixman_region32_t *src)
{
pixman_box32_t *src_rects, *dest_rects;
int nrects, i;
src_rects = pixman_region32_rectangles(src, &nrects);
dest_rects = malloc(nrects * sizeof(*dest_rects));
if (!dest_rects)
return;
for (i = 0; i < nrects; i++) {
struct weston_vector vec1 = {{
src_rects[i].x1, src_rects[i].y1, 0, 1
}};
weston_matrix_transform(matrix, &vec1);
vec1.f[0] /= vec1.f[3];
vec1.f[1] /= vec1.f[3];
struct weston_vector vec2 = {{
src_rects[i].x2, src_rects[i].y2, 0, 1
}};
weston_matrix_transform(matrix, &vec2);
vec2.f[0] /= vec2.f[3];
vec2.f[1] /= vec2.f[3];
if (vec1.f[0] < vec2.f[0]) {
dest_rects[i].x1 = floor(vec1.f[0]);
dest_rects[i].x2 = ceil(vec2.f[0]);
} else {
dest_rects[i].x1 = floor(vec2.f[0]);
dest_rects[i].x2 = ceil(vec1.f[0]);
}
if (vec1.f[1] < vec2.f[1]) {
dest_rects[i].y1 = floor(vec1.f[1]);
dest_rects[i].y2 = ceil(vec2.f[1]);
} else {
dest_rects[i].y1 = floor(vec2.f[1]);
dest_rects[i].y2 = ceil(vec1.f[1]);
}
}
pixman_region32_clear(dest);
pixman_region32_init_rects(dest, dest_rects, nrects);
free(dest_rects);
}
WL_EXPORT void
weston_transformed_region(int width, int height,
enum wl_output_transform transform,
int32_t scale,
pixman_region32_t *src, pixman_region32_t *dest)
{
pixman_box32_t *src_rects, *dest_rects;
int nrects, i;
if (transform == WL_OUTPUT_TRANSFORM_NORMAL && scale == 1) {
if (src != dest)
pixman_region32_copy(dest, src);
return;
}
src_rects = pixman_region32_rectangles(src, &nrects);
dest_rects = malloc(nrects * sizeof(*dest_rects));
if (!dest_rects)
return;
if (transform == WL_OUTPUT_TRANSFORM_NORMAL) {
memcpy(dest_rects, src_rects, nrects * sizeof(*dest_rects));
} else {
for (i = 0; i < nrects; i++) {
switch (transform) {
default:
case WL_OUTPUT_TRANSFORM_NORMAL:
dest_rects[i].x1 = src_rects[i].x1;
dest_rects[i].y1 = src_rects[i].y1;
dest_rects[i].x2 = src_rects[i].x2;
dest_rects[i].y2 = src_rects[i].y2;
break;
case WL_OUTPUT_TRANSFORM_90:
dest_rects[i].x1 = height - src_rects[i].y2;
dest_rects[i].y1 = src_rects[i].x1;
dest_rects[i].x2 = height - src_rects[i].y1;
dest_rects[i].y2 = src_rects[i].x2;
break;
case WL_OUTPUT_TRANSFORM_180:
dest_rects[i].x1 = width - src_rects[i].x2;
dest_rects[i].y1 = height - src_rects[i].y2;
dest_rects[i].x2 = width - src_rects[i].x1;
dest_rects[i].y2 = height - src_rects[i].y1;
break;
case WL_OUTPUT_TRANSFORM_270:
dest_rects[i].x1 = src_rects[i].y1;
dest_rects[i].y1 = width - src_rects[i].x2;
dest_rects[i].x2 = src_rects[i].y2;
dest_rects[i].y2 = width - src_rects[i].x1;
break;
case WL_OUTPUT_TRANSFORM_FLIPPED:
dest_rects[i].x1 = width - src_rects[i].x2;
dest_rects[i].y1 = src_rects[i].y1;
dest_rects[i].x2 = width - src_rects[i].x1;
dest_rects[i].y2 = src_rects[i].y2;
break;
case WL_OUTPUT_TRANSFORM_FLIPPED_90:
dest_rects[i].x1 = height - src_rects[i].y2;
dest_rects[i].y1 = width - src_rects[i].x2;
dest_rects[i].x2 = height - src_rects[i].y1;
dest_rects[i].y2 = width - src_rects[i].x1;
break;
case WL_OUTPUT_TRANSFORM_FLIPPED_180:
dest_rects[i].x1 = src_rects[i].x1;
dest_rects[i].y1 = height - src_rects[i].y2;
dest_rects[i].x2 = src_rects[i].x2;
dest_rects[i].y2 = height - src_rects[i].y1;
break;
case WL_OUTPUT_TRANSFORM_FLIPPED_270:
dest_rects[i].x1 = src_rects[i].y1;
dest_rects[i].y1 = src_rects[i].x1;
dest_rects[i].x2 = src_rects[i].y2;
dest_rects[i].y2 = src_rects[i].x2;
break;
}
}
}
if (scale != 1) {
for (i = 0; i < nrects; i++) {
dest_rects[i].x1 *= scale;
dest_rects[i].x2 *= scale;
dest_rects[i].y1 *= scale;
dest_rects[i].y2 *= scale;
}
}
pixman_region32_clear(dest);
pixman_region32_init_rects(dest, dest_rects, nrects);
free(dest_rects);
}
static void
viewport_surface_to_buffer(struct weston_surface *surface,
float sx, float sy, float *bx, float *by)
{
struct weston_buffer_viewport *vp = &surface->buffer_viewport;
double src_width, src_height;
double src_x, src_y;
if (vp->buffer.src_width == wl_fixed_from_int(-1)) {
if (vp->surface.width == -1) {
*bx = sx;
*by = sy;
return;
}
src_x = 0.0;
src_y = 0.0;
src_width = surface->width_from_buffer;
src_height = surface->height_from_buffer;
} else {
src_x = wl_fixed_to_double(vp->buffer.src_x);
src_y = wl_fixed_to_double(vp->buffer.src_y);
src_width = wl_fixed_to_double(vp->buffer.src_width);
src_height = wl_fixed_to_double(vp->buffer.src_height);
}
*bx = sx * src_width / surface->width + src_x;
*by = sy * src_height / surface->height + src_y;
}
WL_EXPORT void
weston_surface_to_buffer_float(struct weston_surface *surface,
float sx, float sy, float *bx, float *by)
{
struct weston_buffer_viewport *vp = &surface->buffer_viewport;
/* first transform coordinates if the viewport is set */
viewport_surface_to_buffer(surface, sx, sy, bx, by);
weston_transformed_coord(surface->width_from_buffer,
surface->height_from_buffer,
vp->buffer.transform, vp->buffer.scale,
*bx, *by, bx, by);
}
/** Transform a rectangle from surface coordinates to buffer coordinates
*
* \param surface The surface to fetch wp_viewport and buffer transformation
* from.
* \param rect The rectangle to transform.
* \return The transformed rectangle.
*
* Viewport and buffer transformations can only do translation, scaling,
* and rotations in 90-degree steps. Therefore the only loss in the
* conversion is coordinate rounding.
*
* However, some coordinate rounding takes place as an intermediate
* step before the buffer scale factor is applied, so the rectangle
* boundary may not be exactly as expected.
*
* This is OK for damage tracking since a little extra coverage is
* not a problem.
*/
WL_EXPORT pixman_box32_t
weston_surface_to_buffer_rect(struct weston_surface *surface,
pixman_box32_t rect)
{
struct weston_buffer_viewport *vp = &surface->buffer_viewport;
float xf, yf;
/* first transform box coordinates if the viewport is set */
viewport_surface_to_buffer(surface, rect.x1, rect.y1, &xf, &yf);
rect.x1 = floorf(xf);
rect.y1 = floorf(yf);
viewport_surface_to_buffer(surface, rect.x2, rect.y2, &xf, &yf);
rect.x2 = ceilf(xf);
rect.y2 = ceilf(yf);
return weston_transformed_rect(surface->width_from_buffer,
surface->height_from_buffer,
vp->buffer.transform, vp->buffer.scale,
rect);
}
/** Transform a region from surface coordinates to buffer coordinates
*
* \param surface The surface to fetch wp_viewport and buffer transformation
* from.
* \param surface_region[in] The region in surface coordinates.
* \param buffer_region[out] The region converted to buffer coordinates.
*
* Buffer_region must be init'd, but will be completely overwritten.
*
* Viewport and buffer transformations can only do translation, scaling,
* and rotations in 90-degree steps. Therefore the only loss in the
* conversion is from the coordinate rounding that takes place in
* \ref weston_surface_to_buffer_rect.
*/
WL_EXPORT void
weston_surface_to_buffer_region(struct weston_surface *surface,
pixman_region32_t *surface_region,
pixman_region32_t *buffer_region)
{
pixman_box32_t *src_rects, *dest_rects;
int nrects, i;
src_rects = pixman_region32_rectangles(surface_region, &nrects);
dest_rects = malloc(nrects * sizeof(*dest_rects));
if (!dest_rects)
return;
for (i = 0; i < nrects; i++) {
dest_rects[i] = weston_surface_to_buffer_rect(surface,
src_rects[i]);
}
pixman_region32_fini(buffer_region);
pixman_region32_init_rects(buffer_region, dest_rects, nrects);
free(dest_rects);
}
WL_EXPORT void
Split the geometry information from weston_surface out into weston_view The weston_surface structure is split into two structures: * The weston_surface structure storres everything required for a client-side or server-side surface. This includes buffers; callbacks; backend private data; input, damage, and opaque regions; and a few other bookkeeping bits. * The weston_view structure represents an entity in the scenegraph and storres all of the geometry information. This includes clip region, alpha, position, and the transformation list as well as all of the temporary information derived from the geometry state. Because a view, and not a surface, is a scenegraph element, the view is what is placed in layers and planes. There are a few things worth noting about the surface/view split: 1. This is *not* a modification to the protocol. It is, instead, a modification to Weston's internal scenegraph to allow a single surface to exist in multiple places at a time. Clients are completely unaware of how many views to a particular surface exist. 2. A view is considered a direct child of a surface and is destroyed when the surface is destroyed. Because of this, the view.surface pointer is always valid and non-null. 3. The compositor's surface_list is replaced with a view_list. Due to subsurfaces, building the view list is a little more complicated than it used to be and involves building a tree of views on the fly whenever subsurfaces are used. However, this means that backends can remain completely subsurface-agnostic. 4. Surfaces and views both keep track of which outputs they are on. 5. The weston_surface structure now has width and height fields. These are populated when a new buffer is attached before surface.configure is called. This is because there are many surface-based operations that really require the width and height and digging through the views didn't work well. Signed-off-by: Jason Ekstrand <jason@jlekstrand.net>
2013-10-13 07:38:11 +04:00
weston_view_move_to_plane(struct weston_view *view,
struct weston_plane *plane)
{
Split the geometry information from weston_surface out into weston_view The weston_surface structure is split into two structures: * The weston_surface structure storres everything required for a client-side or server-side surface. This includes buffers; callbacks; backend private data; input, damage, and opaque regions; and a few other bookkeeping bits. * The weston_view structure represents an entity in the scenegraph and storres all of the geometry information. This includes clip region, alpha, position, and the transformation list as well as all of the temporary information derived from the geometry state. Because a view, and not a surface, is a scenegraph element, the view is what is placed in layers and planes. There are a few things worth noting about the surface/view split: 1. This is *not* a modification to the protocol. It is, instead, a modification to Weston's internal scenegraph to allow a single surface to exist in multiple places at a time. Clients are completely unaware of how many views to a particular surface exist. 2. A view is considered a direct child of a surface and is destroyed when the surface is destroyed. Because of this, the view.surface pointer is always valid and non-null. 3. The compositor's surface_list is replaced with a view_list. Due to subsurfaces, building the view list is a little more complicated than it used to be and involves building a tree of views on the fly whenever subsurfaces are used. However, this means that backends can remain completely subsurface-agnostic. 4. Surfaces and views both keep track of which outputs they are on. 5. The weston_surface structure now has width and height fields. These are populated when a new buffer is attached before surface.configure is called. This is because there are many surface-based operations that really require the width and height and digging through the views didn't work well. Signed-off-by: Jason Ekstrand <jason@jlekstrand.net>
2013-10-13 07:38:11 +04:00
if (view->plane == plane)
return;
Split the geometry information from weston_surface out into weston_view The weston_surface structure is split into two structures: * The weston_surface structure storres everything required for a client-side or server-side surface. This includes buffers; callbacks; backend private data; input, damage, and opaque regions; and a few other bookkeeping bits. * The weston_view structure represents an entity in the scenegraph and storres all of the geometry information. This includes clip region, alpha, position, and the transformation list as well as all of the temporary information derived from the geometry state. Because a view, and not a surface, is a scenegraph element, the view is what is placed in layers and planes. There are a few things worth noting about the surface/view split: 1. This is *not* a modification to the protocol. It is, instead, a modification to Weston's internal scenegraph to allow a single surface to exist in multiple places at a time. Clients are completely unaware of how many views to a particular surface exist. 2. A view is considered a direct child of a surface and is destroyed when the surface is destroyed. Because of this, the view.surface pointer is always valid and non-null. 3. The compositor's surface_list is replaced with a view_list. Due to subsurfaces, building the view list is a little more complicated than it used to be and involves building a tree of views on the fly whenever subsurfaces are used. However, this means that backends can remain completely subsurface-agnostic. 4. Surfaces and views both keep track of which outputs they are on. 5. The weston_surface structure now has width and height fields. These are populated when a new buffer is attached before surface.configure is called. This is because there are many surface-based operations that really require the width and height and digging through the views didn't work well. Signed-off-by: Jason Ekstrand <jason@jlekstrand.net>
2013-10-13 07:38:11 +04:00
weston_view_damage_below(view);
view->plane = plane;
weston_surface_damage(view->surface);
}
/** Inflict damage on the plane where the view is visible.
*
* \param view The view that causes the damage.
*
* If the view is currently on a plane (including the primary plane),
* take the view's boundingbox, subtract all the opaque views that cover it,
* and add the remaining region as damage to the plane. This corresponds
* to the damage inflicted to the plane if this view disappeared.
*
* A repaint is scheduled for this view.
*
* The region of all opaque views covering this view is stored in
* weston_view::clip and updated by view_accumulate_damage() during
* weston_output_repaint(). Specifically, that region matches the
* scenegraph as it was last painted.
*/
WL_EXPORT void
Split the geometry information from weston_surface out into weston_view The weston_surface structure is split into two structures: * The weston_surface structure storres everything required for a client-side or server-side surface. This includes buffers; callbacks; backend private data; input, damage, and opaque regions; and a few other bookkeeping bits. * The weston_view structure represents an entity in the scenegraph and storres all of the geometry information. This includes clip region, alpha, position, and the transformation list as well as all of the temporary information derived from the geometry state. Because a view, and not a surface, is a scenegraph element, the view is what is placed in layers and planes. There are a few things worth noting about the surface/view split: 1. This is *not* a modification to the protocol. It is, instead, a modification to Weston's internal scenegraph to allow a single surface to exist in multiple places at a time. Clients are completely unaware of how many views to a particular surface exist. 2. A view is considered a direct child of a surface and is destroyed when the surface is destroyed. Because of this, the view.surface pointer is always valid and non-null. 3. The compositor's surface_list is replaced with a view_list. Due to subsurfaces, building the view list is a little more complicated than it used to be and involves building a tree of views on the fly whenever subsurfaces are used. However, this means that backends can remain completely subsurface-agnostic. 4. Surfaces and views both keep track of which outputs they are on. 5. The weston_surface structure now has width and height fields. These are populated when a new buffer is attached before surface.configure is called. This is because there are many surface-based operations that really require the width and height and digging through the views didn't work well. Signed-off-by: Jason Ekstrand <jason@jlekstrand.net>
2013-10-13 07:38:11 +04:00
weston_view_damage_below(struct weston_view *view)
{
pixman_region32_t damage;
pixman_region32_init(&damage);
pixman_region32_subtract(&damage, &view->transform.boundingbox,
Split the geometry information from weston_surface out into weston_view The weston_surface structure is split into two structures: * The weston_surface structure storres everything required for a client-side or server-side surface. This includes buffers; callbacks; backend private data; input, damage, and opaque regions; and a few other bookkeeping bits. * The weston_view structure represents an entity in the scenegraph and storres all of the geometry information. This includes clip region, alpha, position, and the transformation list as well as all of the temporary information derived from the geometry state. Because a view, and not a surface, is a scenegraph element, the view is what is placed in layers and planes. There are a few things worth noting about the surface/view split: 1. This is *not* a modification to the protocol. It is, instead, a modification to Weston's internal scenegraph to allow a single surface to exist in multiple places at a time. Clients are completely unaware of how many views to a particular surface exist. 2. A view is considered a direct child of a surface and is destroyed when the surface is destroyed. Because of this, the view.surface pointer is always valid and non-null. 3. The compositor's surface_list is replaced with a view_list. Due to subsurfaces, building the view list is a little more complicated than it used to be and involves building a tree of views on the fly whenever subsurfaces are used. However, this means that backends can remain completely subsurface-agnostic. 4. Surfaces and views both keep track of which outputs they are on. 5. The weston_surface structure now has width and height fields. These are populated when a new buffer is attached before surface.configure is called. This is because there are many surface-based operations that really require the width and height and digging through the views didn't work well. Signed-off-by: Jason Ekstrand <jason@jlekstrand.net>
2013-10-13 07:38:11 +04:00
&view->clip);
if (view->plane)
pixman_region32_union(&view->plane->damage,
&view->plane->damage, &damage);
pixman_region32_fini(&damage);
weston_view_schedule_repaint(view);
}
/**
* \param es The surface
* \param mask The new set of outputs for the surface
*
* Sets the surface's set of outputs to the ones specified by
* the new output mask provided. Identifies the outputs that
* have changed, the posts enter and leave events for these
* outputs as appropriate.
*/
static void
weston_surface_update_output_mask(struct weston_surface *es, uint32_t mask)
{
uint32_t different = es->output_mask ^ mask;
uint32_t entered = mask & different;
uint32_t left = es->output_mask & different;
struct weston_output *output;
struct wl_resource *resource = NULL;
struct wl_client *client;
es->output_mask = mask;
if (es->resource == NULL)
return;
if (different == 0)
return;
client = wl_resource_get_client(es->resource);
wl_list_for_each(output, &es->compositor->output_list, link) {
if (1u << output->id & different)
resource =
wl_resource_find_for_client(&output->resource_list,
client);
if (resource == NULL)
continue;
if (1u << output->id & entered)
wl_surface_send_enter(es->resource, resource);
if (1u << output->id & left)
wl_surface_send_leave(es->resource, resource);
}
}
/** Recalculate which output(s) the surface has views displayed on
*
* \param es The surface to remap to outputs
*
* Finds the output that is showing the largest amount of one
* of the surface's various views. This output becomes the
* surface's primary output for vsync and frame callback purposes.
*
* Also notes all outputs of all of the surface's views
* in the output_mask for the surface.
*/
static void
weston_surface_assign_output(struct weston_surface *es)
{
Split the geometry information from weston_surface out into weston_view The weston_surface structure is split into two structures: * The weston_surface structure storres everything required for a client-side or server-side surface. This includes buffers; callbacks; backend private data; input, damage, and opaque regions; and a few other bookkeeping bits. * The weston_view structure represents an entity in the scenegraph and storres all of the geometry information. This includes clip region, alpha, position, and the transformation list as well as all of the temporary information derived from the geometry state. Because a view, and not a surface, is a scenegraph element, the view is what is placed in layers and planes. There are a few things worth noting about the surface/view split: 1. This is *not* a modification to the protocol. It is, instead, a modification to Weston's internal scenegraph to allow a single surface to exist in multiple places at a time. Clients are completely unaware of how many views to a particular surface exist. 2. A view is considered a direct child of a surface and is destroyed when the surface is destroyed. Because of this, the view.surface pointer is always valid and non-null. 3. The compositor's surface_list is replaced with a view_list. Due to subsurfaces, building the view list is a little more complicated than it used to be and involves building a tree of views on the fly whenever subsurfaces are used. However, this means that backends can remain completely subsurface-agnostic. 4. Surfaces and views both keep track of which outputs they are on. 5. The weston_surface structure now has width and height fields. These are populated when a new buffer is attached before surface.configure is called. This is because there are many surface-based operations that really require the width and height and digging through the views didn't work well. Signed-off-by: Jason Ekstrand <jason@jlekstrand.net>
2013-10-13 07:38:11 +04:00
struct weston_output *new_output;
struct weston_view *view;
pixman_region32_t region;
uint32_t max, area, mask;
pixman_box32_t *e;
new_output = NULL;
max = 0;
mask = 0;
pixman_region32_init(&region);
wl_list_for_each(view, &es->views, surface_link) {
if (!view->output)
continue;
pixman_region32_intersect(&region, &view->transform.boundingbox,
&view->output->region);
e = pixman_region32_extents(&region);
area = (e->x2 - e->x1) * (e->y2 - e->y1);
mask |= view->output_mask;
if (area >= max) {
new_output = view->output;
max = area;
}
}
pixman_region32_fini(&region);
es->output = new_output;
weston_surface_update_output_mask(es, mask);
}
/** Recalculate which output(s) the view is displayed on
*
* \param ev The view to remap to outputs
*
* Identifies the set of outputs that the view is visible on,
* noting them into the output_mask. The output that the view
* is most visible on is set as the view's primary output.
*
* Also does the same for the view's surface. See
* weston_surface_assign_output().
*/
Split the geometry information from weston_surface out into weston_view The weston_surface structure is split into two structures: * The weston_surface structure storres everything required for a client-side or server-side surface. This includes buffers; callbacks; backend private data; input, damage, and opaque regions; and a few other bookkeeping bits. * The weston_view structure represents an entity in the scenegraph and storres all of the geometry information. This includes clip region, alpha, position, and the transformation list as well as all of the temporary information derived from the geometry state. Because a view, and not a surface, is a scenegraph element, the view is what is placed in layers and planes. There are a few things worth noting about the surface/view split: 1. This is *not* a modification to the protocol. It is, instead, a modification to Weston's internal scenegraph to allow a single surface to exist in multiple places at a time. Clients are completely unaware of how many views to a particular surface exist. 2. A view is considered a direct child of a surface and is destroyed when the surface is destroyed. Because of this, the view.surface pointer is always valid and non-null. 3. The compositor's surface_list is replaced with a view_list. Due to subsurfaces, building the view list is a little more complicated than it used to be and involves building a tree of views on the fly whenever subsurfaces are used. However, this means that backends can remain completely subsurface-agnostic. 4. Surfaces and views both keep track of which outputs they are on. 5. The weston_surface structure now has width and height fields. These are populated when a new buffer is attached before surface.configure is called. This is because there are many surface-based operations that really require the width and height and digging through the views didn't work well. Signed-off-by: Jason Ekstrand <jason@jlekstrand.net>
2013-10-13 07:38:11 +04:00
static void
weston_view_assign_output(struct weston_view *ev)
{
struct weston_compositor *ec = ev->surface->compositor;
struct weston_output *output, *new_output;
pixman_region32_t region;
uint32_t max, area, mask;
pixman_box32_t *e;
new_output = NULL;
max = 0;
mask = 0;
pixman_region32_init(&region);
wl_list_for_each(output, &ec->output_list, link) {
if (output->destroying)
continue;
Split the geometry information from weston_surface out into weston_view The weston_surface structure is split into two structures: * The weston_surface structure storres everything required for a client-side or server-side surface. This includes buffers; callbacks; backend private data; input, damage, and opaque regions; and a few other bookkeeping bits. * The weston_view structure represents an entity in the scenegraph and storres all of the geometry information. This includes clip region, alpha, position, and the transformation list as well as all of the temporary information derived from the geometry state. Because a view, and not a surface, is a scenegraph element, the view is what is placed in layers and planes. There are a few things worth noting about the surface/view split: 1. This is *not* a modification to the protocol. It is, instead, a modification to Weston's internal scenegraph to allow a single surface to exist in multiple places at a time. Clients are completely unaware of how many views to a particular surface exist. 2. A view is considered a direct child of a surface and is destroyed when the surface is destroyed. Because of this, the view.surface pointer is always valid and non-null. 3. The compositor's surface_list is replaced with a view_list. Due to subsurfaces, building the view list is a little more complicated than it used to be and involves building a tree of views on the fly whenever subsurfaces are used. However, this means that backends can remain completely subsurface-agnostic. 4. Surfaces and views both keep track of which outputs they are on. 5. The weston_surface structure now has width and height fields. These are populated when a new buffer is attached before surface.configure is called. This is because there are many surface-based operations that really require the width and height and digging through the views didn't work well. Signed-off-by: Jason Ekstrand <jason@jlekstrand.net>
2013-10-13 07:38:11 +04:00
pixman_region32_intersect(&region, &ev->transform.boundingbox,
&output->region);
e = pixman_region32_extents(&region);
area = (e->x2 - e->x1) * (e->y2 - e->y1);
if (area > 0)
mask |= 1u << output->id;
if (area >= max) {
new_output = output;
max = area;
}
}
pixman_region32_fini(&region);
Split the geometry information from weston_surface out into weston_view The weston_surface structure is split into two structures: * The weston_surface structure storres everything required for a client-side or server-side surface. This includes buffers; callbacks; backend private data; input, damage, and opaque regions; and a few other bookkeeping bits. * The weston_view structure represents an entity in the scenegraph and storres all of the geometry information. This includes clip region, alpha, position, and the transformation list as well as all of the temporary information derived from the geometry state. Because a view, and not a surface, is a scenegraph element, the view is what is placed in layers and planes. There are a few things worth noting about the surface/view split: 1. This is *not* a modification to the protocol. It is, instead, a modification to Weston's internal scenegraph to allow a single surface to exist in multiple places at a time. Clients are completely unaware of how many views to a particular surface exist. 2. A view is considered a direct child of a surface and is destroyed when the surface is destroyed. Because of this, the view.surface pointer is always valid and non-null. 3. The compositor's surface_list is replaced with a view_list. Due to subsurfaces, building the view list is a little more complicated than it used to be and involves building a tree of views on the fly whenever subsurfaces are used. However, this means that backends can remain completely subsurface-agnostic. 4. Surfaces and views both keep track of which outputs they are on. 5. The weston_surface structure now has width and height fields. These are populated when a new buffer is attached before surface.configure is called. This is because there are many surface-based operations that really require the width and height and digging through the views didn't work well. Signed-off-by: Jason Ekstrand <jason@jlekstrand.net>
2013-10-13 07:38:11 +04:00
ev->output = new_output;
ev->output_mask = mask;
weston_surface_assign_output(ev->surface);
}
compositor: add weston_view_set_mask() API and state Add API for setting a clip ('scissor' in the code) rectangle per view, in surface coordinates. Ivi-shell requires this feature to be able to implement the IVI Layer Manager API, which includes clipping of surfaces. The names weston_view_set_mask() and weston_view_set_mask_infinite() mirror the existing weston_layer_set_mask*() functions. This view clipping complements the weston_layer clipping, because view clipping is defined in surface local coordinates, while layer mask/clipping is defined in global coordinates. View clipping requires explicit support from the renderers. Therefore a new Weston capability bit is added: WESTON_CAP_VIEW_CLIP_MASK. Shells (and all users) of this new API are required to check the capability bit is set before using the API. Otherwise the rendering will not be what they expect. View clips are inherited through the transformation inheritance mechanism. However, there are restrictions. The clip rectangle can be set only on the root view of a transformation inheritance tree. The additional transformations in child views must not rotate the coordinate axes. These restrictions avoid corner cases in clip inheritance, and keep the renderer implementations as simple as they are right now. Renderers only need to do an additional intersection with the clip rectangle which is always aligned to the surface coordinate system. For more details, see the API documentation in the patch. Signed-off-by: Pekka Paalanen <pekka.paalanen@collabora.co.uk> Reviewed-by: Nobuhiko Tanibata <NOBUHIKO_TANIBATA@xddp.denso.co.jp> Tested-by: Nobuhiko Tanibata <NOBUHIKO_TANIBATA@xddp.denso.co.jp> Reviewed-By: Derek Foreman <derekf@osg.samsung.com>
2015-02-16 15:39:11 +03:00
static void
weston_view_to_view_map(struct weston_view *from, struct weston_view *to,
int from_x, int from_y, int *to_x, int *to_y)
{
float x, y;
weston_view_to_global_float(from, from_x, from_y, &x, &y);
weston_view_from_global_float(to, x, y, &x, &y);
*to_x = round(x);
*to_y = round(y);
}
static void
weston_view_transfer_scissor(struct weston_view *from, struct weston_view *to)
{
pixman_box32_t *a;
pixman_box32_t b;
a = pixman_region32_extents(&from->geometry.scissor);
weston_view_to_view_map(from, to, a->x1, a->y1, &b.x1, &b.y1);
weston_view_to_view_map(from, to, a->x2, a->y2, &b.x2, &b.y2);
pixman_region32_fini(&to->geometry.scissor);
pixman_region32_init_with_extents(&to->geometry.scissor, &b);
}
static void
view_compute_bbox(struct weston_view *view, const pixman_box32_t *inbox,
Split the geometry information from weston_surface out into weston_view The weston_surface structure is split into two structures: * The weston_surface structure storres everything required for a client-side or server-side surface. This includes buffers; callbacks; backend private data; input, damage, and opaque regions; and a few other bookkeeping bits. * The weston_view structure represents an entity in the scenegraph and storres all of the geometry information. This includes clip region, alpha, position, and the transformation list as well as all of the temporary information derived from the geometry state. Because a view, and not a surface, is a scenegraph element, the view is what is placed in layers and planes. There are a few things worth noting about the surface/view split: 1. This is *not* a modification to the protocol. It is, instead, a modification to Weston's internal scenegraph to allow a single surface to exist in multiple places at a time. Clients are completely unaware of how many views to a particular surface exist. 2. A view is considered a direct child of a surface and is destroyed when the surface is destroyed. Because of this, the view.surface pointer is always valid and non-null. 3. The compositor's surface_list is replaced with a view_list. Due to subsurfaces, building the view list is a little more complicated than it used to be and involves building a tree of views on the fly whenever subsurfaces are used. However, this means that backends can remain completely subsurface-agnostic. 4. Surfaces and views both keep track of which outputs they are on. 5. The weston_surface structure now has width and height fields. These are populated when a new buffer is attached before surface.configure is called. This is because there are many surface-based operations that really require the width and height and digging through the views didn't work well. Signed-off-by: Jason Ekstrand <jason@jlekstrand.net>
2013-10-13 07:38:11 +04:00
pixman_region32_t *bbox)
{
2012-09-30 04:57:21 +04:00
float min_x = HUGE_VALF, min_y = HUGE_VALF;
float max_x = -HUGE_VALF, max_y = -HUGE_VALF;
int32_t s[4][2] = {
{ inbox->x1, inbox->y1 },
{ inbox->x1, inbox->y2 },
{ inbox->x2, inbox->y1 },
{ inbox->x2, inbox->y2 },
};
2012-09-30 04:57:21 +04:00
float int_x, int_y;
int i;
if (inbox->x1 == inbox->x2 || inbox->y1 == inbox->y2) {
/* avoid rounding empty bbox to 1x1 */
pixman_region32_init(bbox);
return;
}
for (i = 0; i < 4; ++i) {
2012-09-30 04:57:21 +04:00
float x, y;
Split the geometry information from weston_surface out into weston_view The weston_surface structure is split into two structures: * The weston_surface structure storres everything required for a client-side or server-side surface. This includes buffers; callbacks; backend private data; input, damage, and opaque regions; and a few other bookkeeping bits. * The weston_view structure represents an entity in the scenegraph and storres all of the geometry information. This includes clip region, alpha, position, and the transformation list as well as all of the temporary information derived from the geometry state. Because a view, and not a surface, is a scenegraph element, the view is what is placed in layers and planes. There are a few things worth noting about the surface/view split: 1. This is *not* a modification to the protocol. It is, instead, a modification to Weston's internal scenegraph to allow a single surface to exist in multiple places at a time. Clients are completely unaware of how many views to a particular surface exist. 2. A view is considered a direct child of a surface and is destroyed when the surface is destroyed. Because of this, the view.surface pointer is always valid and non-null. 3. The compositor's surface_list is replaced with a view_list. Due to subsurfaces, building the view list is a little more complicated than it used to be and involves building a tree of views on the fly whenever subsurfaces are used. However, this means that backends can remain completely subsurface-agnostic. 4. Surfaces and views both keep track of which outputs they are on. 5. The weston_surface structure now has width and height fields. These are populated when a new buffer is attached before surface.configure is called. This is because there are many surface-based operations that really require the width and height and digging through the views didn't work well. Signed-off-by: Jason Ekstrand <jason@jlekstrand.net>
2013-10-13 07:38:11 +04:00
weston_view_to_global_float(view, s[i][0], s[i][1], &x, &y);
if (x < min_x)
min_x = x;
if (x > max_x)
max_x = x;
if (y < min_y)
min_y = y;
if (y > max_y)
max_y = y;
}
int_x = floorf(min_x);
int_y = floorf(min_y);
pixman_region32_init_rect(bbox, int_x, int_y,
ceilf(max_x) - int_x, ceilf(max_y) - int_y);
}
static void
Split the geometry information from weston_surface out into weston_view The weston_surface structure is split into two structures: * The weston_surface structure storres everything required for a client-side or server-side surface. This includes buffers; callbacks; backend private data; input, damage, and opaque regions; and a few other bookkeeping bits. * The weston_view structure represents an entity in the scenegraph and storres all of the geometry information. This includes clip region, alpha, position, and the transformation list as well as all of the temporary information derived from the geometry state. Because a view, and not a surface, is a scenegraph element, the view is what is placed in layers and planes. There are a few things worth noting about the surface/view split: 1. This is *not* a modification to the protocol. It is, instead, a modification to Weston's internal scenegraph to allow a single surface to exist in multiple places at a time. Clients are completely unaware of how many views to a particular surface exist. 2. A view is considered a direct child of a surface and is destroyed when the surface is destroyed. Because of this, the view.surface pointer is always valid and non-null. 3. The compositor's surface_list is replaced with a view_list. Due to subsurfaces, building the view list is a little more complicated than it used to be and involves building a tree of views on the fly whenever subsurfaces are used. However, this means that backends can remain completely subsurface-agnostic. 4. Surfaces and views both keep track of which outputs they are on. 5. The weston_surface structure now has width and height fields. These are populated when a new buffer is attached before surface.configure is called. This is because there are many surface-based operations that really require the width and height and digging through the views didn't work well. Signed-off-by: Jason Ekstrand <jason@jlekstrand.net>
2013-10-13 07:38:11 +04:00
weston_view_update_transform_disable(struct weston_view *view)
{
Split the geometry information from weston_surface out into weston_view The weston_surface structure is split into two structures: * The weston_surface structure storres everything required for a client-side or server-side surface. This includes buffers; callbacks; backend private data; input, damage, and opaque regions; and a few other bookkeeping bits. * The weston_view structure represents an entity in the scenegraph and storres all of the geometry information. This includes clip region, alpha, position, and the transformation list as well as all of the temporary information derived from the geometry state. Because a view, and not a surface, is a scenegraph element, the view is what is placed in layers and planes. There are a few things worth noting about the surface/view split: 1. This is *not* a modification to the protocol. It is, instead, a modification to Weston's internal scenegraph to allow a single surface to exist in multiple places at a time. Clients are completely unaware of how many views to a particular surface exist. 2. A view is considered a direct child of a surface and is destroyed when the surface is destroyed. Because of this, the view.surface pointer is always valid and non-null. 3. The compositor's surface_list is replaced with a view_list. Due to subsurfaces, building the view list is a little more complicated than it used to be and involves building a tree of views on the fly whenever subsurfaces are used. However, this means that backends can remain completely subsurface-agnostic. 4. Surfaces and views both keep track of which outputs they are on. 5. The weston_surface structure now has width and height fields. These are populated when a new buffer is attached before surface.configure is called. This is because there are many surface-based operations that really require the width and height and digging through the views didn't work well. Signed-off-by: Jason Ekstrand <jason@jlekstrand.net>
2013-10-13 07:38:11 +04:00
view->transform.enabled = 0;
/* round off fractions when not transformed */
Split the geometry information from weston_surface out into weston_view The weston_surface structure is split into two structures: * The weston_surface structure storres everything required for a client-side or server-side surface. This includes buffers; callbacks; backend private data; input, damage, and opaque regions; and a few other bookkeeping bits. * The weston_view structure represents an entity in the scenegraph and storres all of the geometry information. This includes clip region, alpha, position, and the transformation list as well as all of the temporary information derived from the geometry state. Because a view, and not a surface, is a scenegraph element, the view is what is placed in layers and planes. There are a few things worth noting about the surface/view split: 1. This is *not* a modification to the protocol. It is, instead, a modification to Weston's internal scenegraph to allow a single surface to exist in multiple places at a time. Clients are completely unaware of how many views to a particular surface exist. 2. A view is considered a direct child of a surface and is destroyed when the surface is destroyed. Because of this, the view.surface pointer is always valid and non-null. 3. The compositor's surface_list is replaced with a view_list. Due to subsurfaces, building the view list is a little more complicated than it used to be and involves building a tree of views on the fly whenever subsurfaces are used. However, this means that backends can remain completely subsurface-agnostic. 4. Surfaces and views both keep track of which outputs they are on. 5. The weston_surface structure now has width and height fields. These are populated when a new buffer is attached before surface.configure is called. This is because there are many surface-based operations that really require the width and height and digging through the views didn't work well. Signed-off-by: Jason Ekstrand <jason@jlekstrand.net>
2013-10-13 07:38:11 +04:00
view->geometry.x = roundf(view->geometry.x);
view->geometry.y = roundf(view->geometry.y);
/* Otherwise identity matrix, but with x and y translation. */
Split the geometry information from weston_surface out into weston_view The weston_surface structure is split into two structures: * The weston_surface structure storres everything required for a client-side or server-side surface. This includes buffers; callbacks; backend private data; input, damage, and opaque regions; and a few other bookkeeping bits. * The weston_view structure represents an entity in the scenegraph and storres all of the geometry information. This includes clip region, alpha, position, and the transformation list as well as all of the temporary information derived from the geometry state. Because a view, and not a surface, is a scenegraph element, the view is what is placed in layers and planes. There are a few things worth noting about the surface/view split: 1. This is *not* a modification to the protocol. It is, instead, a modification to Weston's internal scenegraph to allow a single surface to exist in multiple places at a time. Clients are completely unaware of how many views to a particular surface exist. 2. A view is considered a direct child of a surface and is destroyed when the surface is destroyed. Because of this, the view.surface pointer is always valid and non-null. 3. The compositor's surface_list is replaced with a view_list. Due to subsurfaces, building the view list is a little more complicated than it used to be and involves building a tree of views on the fly whenever subsurfaces are used. However, this means that backends can remain completely subsurface-agnostic. 4. Surfaces and views both keep track of which outputs they are on. 5. The weston_surface structure now has width and height fields. These are populated when a new buffer is attached before surface.configure is called. This is because there are many surface-based operations that really require the width and height and digging through the views didn't work well. Signed-off-by: Jason Ekstrand <jason@jlekstrand.net>
2013-10-13 07:38:11 +04:00
view->transform.position.matrix.type = WESTON_MATRIX_TRANSFORM_TRANSLATE;
view->transform.position.matrix.d[12] = view->geometry.x;
view->transform.position.matrix.d[13] = view->geometry.y;
Split the geometry information from weston_surface out into weston_view The weston_surface structure is split into two structures: * The weston_surface structure storres everything required for a client-side or server-side surface. This includes buffers; callbacks; backend private data; input, damage, and opaque regions; and a few other bookkeeping bits. * The weston_view structure represents an entity in the scenegraph and storres all of the geometry information. This includes clip region, alpha, position, and the transformation list as well as all of the temporary information derived from the geometry state. Because a view, and not a surface, is a scenegraph element, the view is what is placed in layers and planes. There are a few things worth noting about the surface/view split: 1. This is *not* a modification to the protocol. It is, instead, a modification to Weston's internal scenegraph to allow a single surface to exist in multiple places at a time. Clients are completely unaware of how many views to a particular surface exist. 2. A view is considered a direct child of a surface and is destroyed when the surface is destroyed. Because of this, the view.surface pointer is always valid and non-null. 3. The compositor's surface_list is replaced with a view_list. Due to subsurfaces, building the view list is a little more complicated than it used to be and involves building a tree of views on the fly whenever subsurfaces are used. However, this means that backends can remain completely subsurface-agnostic. 4. Surfaces and views both keep track of which outputs they are on. 5. The weston_surface structure now has width and height fields. These are populated when a new buffer is attached before surface.configure is called. This is because there are many surface-based operations that really require the width and height and digging through the views didn't work well. Signed-off-by: Jason Ekstrand <jason@jlekstrand.net>
2013-10-13 07:38:11 +04:00
view->transform.matrix = view->transform.position.matrix;
Split the geometry information from weston_surface out into weston_view The weston_surface structure is split into two structures: * The weston_surface structure storres everything required for a client-side or server-side surface. This includes buffers; callbacks; backend private data; input, damage, and opaque regions; and a few other bookkeeping bits. * The weston_view structure represents an entity in the scenegraph and storres all of the geometry information. This includes clip region, alpha, position, and the transformation list as well as all of the temporary information derived from the geometry state. Because a view, and not a surface, is a scenegraph element, the view is what is placed in layers and planes. There are a few things worth noting about the surface/view split: 1. This is *not* a modification to the protocol. It is, instead, a modification to Weston's internal scenegraph to allow a single surface to exist in multiple places at a time. Clients are completely unaware of how many views to a particular surface exist. 2. A view is considered a direct child of a surface and is destroyed when the surface is destroyed. Because of this, the view.surface pointer is always valid and non-null. 3. The compositor's surface_list is replaced with a view_list. Due to subsurfaces, building the view list is a little more complicated than it used to be and involves building a tree of views on the fly whenever subsurfaces are used. However, this means that backends can remain completely subsurface-agnostic. 4. Surfaces and views both keep track of which outputs they are on. 5. The weston_surface structure now has width and height fields. These are populated when a new buffer is attached before surface.configure is called. This is because there are many surface-based operations that really require the width and height and digging through the views didn't work well. Signed-off-by: Jason Ekstrand <jason@jlekstrand.net>
2013-10-13 07:38:11 +04:00
view->transform.inverse = view->transform.position.matrix;
view->transform.inverse.d[12] = -view->geometry.x;
view->transform.inverse.d[13] = -view->geometry.y;
Split the geometry information from weston_surface out into weston_view The weston_surface structure is split into two structures: * The weston_surface structure storres everything required for a client-side or server-side surface. This includes buffers; callbacks; backend private data; input, damage, and opaque regions; and a few other bookkeeping bits. * The weston_view structure represents an entity in the scenegraph and storres all of the geometry information. This includes clip region, alpha, position, and the transformation list as well as all of the temporary information derived from the geometry state. Because a view, and not a surface, is a scenegraph element, the view is what is placed in layers and planes. There are a few things worth noting about the surface/view split: 1. This is *not* a modification to the protocol. It is, instead, a modification to Weston's internal scenegraph to allow a single surface to exist in multiple places at a time. Clients are completely unaware of how many views to a particular surface exist. 2. A view is considered a direct child of a surface and is destroyed when the surface is destroyed. Because of this, the view.surface pointer is always valid and non-null. 3. The compositor's surface_list is replaced with a view_list. Due to subsurfaces, building the view list is a little more complicated than it used to be and involves building a tree of views on the fly whenever subsurfaces are used. However, this means that backends can remain completely subsurface-agnostic. 4. Surfaces and views both keep track of which outputs they are on. 5. The weston_surface structure now has width and height fields. These are populated when a new buffer is attached before surface.configure is called. This is because there are many surface-based operations that really require the width and height and digging through the views didn't work well. Signed-off-by: Jason Ekstrand <jason@jlekstrand.net>
2013-10-13 07:38:11 +04:00
pixman_region32_init_rect(&view->transform.boundingbox,
compositor: add weston_view_set_mask() API and state Add API for setting a clip ('scissor' in the code) rectangle per view, in surface coordinates. Ivi-shell requires this feature to be able to implement the IVI Layer Manager API, which includes clipping of surfaces. The names weston_view_set_mask() and weston_view_set_mask_infinite() mirror the existing weston_layer_set_mask*() functions. This view clipping complements the weston_layer clipping, because view clipping is defined in surface local coordinates, while layer mask/clipping is defined in global coordinates. View clipping requires explicit support from the renderers. Therefore a new Weston capability bit is added: WESTON_CAP_VIEW_CLIP_MASK. Shells (and all users) of this new API are required to check the capability bit is set before using the API. Otherwise the rendering will not be what they expect. View clips are inherited through the transformation inheritance mechanism. However, there are restrictions. The clip rectangle can be set only on the root view of a transformation inheritance tree. The additional transformations in child views must not rotate the coordinate axes. These restrictions avoid corner cases in clip inheritance, and keep the renderer implementations as simple as they are right now. Renderers only need to do an additional intersection with the clip rectangle which is always aligned to the surface coordinate system. For more details, see the API documentation in the patch. Signed-off-by: Pekka Paalanen <pekka.paalanen@collabora.co.uk> Reviewed-by: Nobuhiko Tanibata <NOBUHIKO_TANIBATA@xddp.denso.co.jp> Tested-by: Nobuhiko Tanibata <NOBUHIKO_TANIBATA@xddp.denso.co.jp> Reviewed-By: Derek Foreman <derekf@osg.samsung.com>
2015-02-16 15:39:11 +03:00
0, 0,
view->surface->width,
view->surface->height);
compositor: add weston_view_set_mask() API and state Add API for setting a clip ('scissor' in the code) rectangle per view, in surface coordinates. Ivi-shell requires this feature to be able to implement the IVI Layer Manager API, which includes clipping of surfaces. The names weston_view_set_mask() and weston_view_set_mask_infinite() mirror the existing weston_layer_set_mask*() functions. This view clipping complements the weston_layer clipping, because view clipping is defined in surface local coordinates, while layer mask/clipping is defined in global coordinates. View clipping requires explicit support from the renderers. Therefore a new Weston capability bit is added: WESTON_CAP_VIEW_CLIP_MASK. Shells (and all users) of this new API are required to check the capability bit is set before using the API. Otherwise the rendering will not be what they expect. View clips are inherited through the transformation inheritance mechanism. However, there are restrictions. The clip rectangle can be set only on the root view of a transformation inheritance tree. The additional transformations in child views must not rotate the coordinate axes. These restrictions avoid corner cases in clip inheritance, and keep the renderer implementations as simple as they are right now. Renderers only need to do an additional intersection with the clip rectangle which is always aligned to the surface coordinate system. For more details, see the API documentation in the patch. Signed-off-by: Pekka Paalanen <pekka.paalanen@collabora.co.uk> Reviewed-by: Nobuhiko Tanibata <NOBUHIKO_TANIBATA@xddp.denso.co.jp> Tested-by: Nobuhiko Tanibata <NOBUHIKO_TANIBATA@xddp.denso.co.jp> Reviewed-By: Derek Foreman <derekf@osg.samsung.com>
2015-02-16 15:39:11 +03:00
if (view->geometry.scissor_enabled)
pixman_region32_intersect(&view->transform.boundingbox,
&view->transform.boundingbox,
&view->geometry.scissor);
pixman_region32_translate(&view->transform.boundingbox,
view->geometry.x, view->geometry.y);
Split the geometry information from weston_surface out into weston_view The weston_surface structure is split into two structures: * The weston_surface structure storres everything required for a client-side or server-side surface. This includes buffers; callbacks; backend private data; input, damage, and opaque regions; and a few other bookkeeping bits. * The weston_view structure represents an entity in the scenegraph and storres all of the geometry information. This includes clip region, alpha, position, and the transformation list as well as all of the temporary information derived from the geometry state. Because a view, and not a surface, is a scenegraph element, the view is what is placed in layers and planes. There are a few things worth noting about the surface/view split: 1. This is *not* a modification to the protocol. It is, instead, a modification to Weston's internal scenegraph to allow a single surface to exist in multiple places at a time. Clients are completely unaware of how many views to a particular surface exist. 2. A view is considered a direct child of a surface and is destroyed when the surface is destroyed. Because of this, the view.surface pointer is always valid and non-null. 3. The compositor's surface_list is replaced with a view_list. Due to subsurfaces, building the view list is a little more complicated than it used to be and involves building a tree of views on the fly whenever subsurfaces are used. However, this means that backends can remain completely subsurface-agnostic. 4. Surfaces and views both keep track of which outputs they are on. 5. The weston_surface structure now has width and height fields. These are populated when a new buffer is attached before surface.configure is called. This is because there are many surface-based operations that really require the width and height and digging through the views didn't work well. Signed-off-by: Jason Ekstrand <jason@jlekstrand.net>
2013-10-13 07:38:11 +04:00
if (view->alpha == 1.0) {
pixman_region32_copy(&view->transform.opaque,
&view->surface->opaque);
pixman_region32_translate(&view->transform.opaque,
view->geometry.x,
view->geometry.y);
}
}
static int
Split the geometry information from weston_surface out into weston_view The weston_surface structure is split into two structures: * The weston_surface structure storres everything required for a client-side or server-side surface. This includes buffers; callbacks; backend private data; input, damage, and opaque regions; and a few other bookkeeping bits. * The weston_view structure represents an entity in the scenegraph and storres all of the geometry information. This includes clip region, alpha, position, and the transformation list as well as all of the temporary information derived from the geometry state. Because a view, and not a surface, is a scenegraph element, the view is what is placed in layers and planes. There are a few things worth noting about the surface/view split: 1. This is *not* a modification to the protocol. It is, instead, a modification to Weston's internal scenegraph to allow a single surface to exist in multiple places at a time. Clients are completely unaware of how many views to a particular surface exist. 2. A view is considered a direct child of a surface and is destroyed when the surface is destroyed. Because of this, the view.surface pointer is always valid and non-null. 3. The compositor's surface_list is replaced with a view_list. Due to subsurfaces, building the view list is a little more complicated than it used to be and involves building a tree of views on the fly whenever subsurfaces are used. However, this means that backends can remain completely subsurface-agnostic. 4. Surfaces and views both keep track of which outputs they are on. 5. The weston_surface structure now has width and height fields. These are populated when a new buffer is attached before surface.configure is called. This is because there are many surface-based operations that really require the width and height and digging through the views didn't work well. Signed-off-by: Jason Ekstrand <jason@jlekstrand.net>
2013-10-13 07:38:11 +04:00
weston_view_update_transform_enable(struct weston_view *view)
{
Split the geometry information from weston_surface out into weston_view The weston_surface structure is split into two structures: * The weston_surface structure storres everything required for a client-side or server-side surface. This includes buffers; callbacks; backend private data; input, damage, and opaque regions; and a few other bookkeeping bits. * The weston_view structure represents an entity in the scenegraph and storres all of the geometry information. This includes clip region, alpha, position, and the transformation list as well as all of the temporary information derived from the geometry state. Because a view, and not a surface, is a scenegraph element, the view is what is placed in layers and planes. There are a few things worth noting about the surface/view split: 1. This is *not* a modification to the protocol. It is, instead, a modification to Weston's internal scenegraph to allow a single surface to exist in multiple places at a time. Clients are completely unaware of how many views to a particular surface exist. 2. A view is considered a direct child of a surface and is destroyed when the surface is destroyed. Because of this, the view.surface pointer is always valid and non-null. 3. The compositor's surface_list is replaced with a view_list. Due to subsurfaces, building the view list is a little more complicated than it used to be and involves building a tree of views on the fly whenever subsurfaces are used. However, this means that backends can remain completely subsurface-agnostic. 4. Surfaces and views both keep track of which outputs they are on. 5. The weston_surface structure now has width and height fields. These are populated when a new buffer is attached before surface.configure is called. This is because there are many surface-based operations that really require the width and height and digging through the views didn't work well. Signed-off-by: Jason Ekstrand <jason@jlekstrand.net>
2013-10-13 07:38:11 +04:00
struct weston_view *parent = view->geometry.parent;
struct weston_matrix *matrix = &view->transform.matrix;
struct weston_matrix *inverse = &view->transform.inverse;
struct weston_transform *tform;
compositor: add weston_view_set_mask() API and state Add API for setting a clip ('scissor' in the code) rectangle per view, in surface coordinates. Ivi-shell requires this feature to be able to implement the IVI Layer Manager API, which includes clipping of surfaces. The names weston_view_set_mask() and weston_view_set_mask_infinite() mirror the existing weston_layer_set_mask*() functions. This view clipping complements the weston_layer clipping, because view clipping is defined in surface local coordinates, while layer mask/clipping is defined in global coordinates. View clipping requires explicit support from the renderers. Therefore a new Weston capability bit is added: WESTON_CAP_VIEW_CLIP_MASK. Shells (and all users) of this new API are required to check the capability bit is set before using the API. Otherwise the rendering will not be what they expect. View clips are inherited through the transformation inheritance mechanism. However, there are restrictions. The clip rectangle can be set only on the root view of a transformation inheritance tree. The additional transformations in child views must not rotate the coordinate axes. These restrictions avoid corner cases in clip inheritance, and keep the renderer implementations as simple as they are right now. Renderers only need to do an additional intersection with the clip rectangle which is always aligned to the surface coordinate system. For more details, see the API documentation in the patch. Signed-off-by: Pekka Paalanen <pekka.paalanen@collabora.co.uk> Reviewed-by: Nobuhiko Tanibata <NOBUHIKO_TANIBATA@xddp.denso.co.jp> Tested-by: Nobuhiko Tanibata <NOBUHIKO_TANIBATA@xddp.denso.co.jp> Reviewed-By: Derek Foreman <derekf@osg.samsung.com>
2015-02-16 15:39:11 +03:00
pixman_region32_t surfregion;
const pixman_box32_t *surfbox;
Split the geometry information from weston_surface out into weston_view The weston_surface structure is split into two structures: * The weston_surface structure storres everything required for a client-side or server-side surface. This includes buffers; callbacks; backend private data; input, damage, and opaque regions; and a few other bookkeeping bits. * The weston_view structure represents an entity in the scenegraph and storres all of the geometry information. This includes clip region, alpha, position, and the transformation list as well as all of the temporary information derived from the geometry state. Because a view, and not a surface, is a scenegraph element, the view is what is placed in layers and planes. There are a few things worth noting about the surface/view split: 1. This is *not* a modification to the protocol. It is, instead, a modification to Weston's internal scenegraph to allow a single surface to exist in multiple places at a time. Clients are completely unaware of how many views to a particular surface exist. 2. A view is considered a direct child of a surface and is destroyed when the surface is destroyed. Because of this, the view.surface pointer is always valid and non-null. 3. The compositor's surface_list is replaced with a view_list. Due to subsurfaces, building the view list is a little more complicated than it used to be and involves building a tree of views on the fly whenever subsurfaces are used. However, this means that backends can remain completely subsurface-agnostic. 4. Surfaces and views both keep track of which outputs they are on. 5. The weston_surface structure now has width and height fields. These are populated when a new buffer is attached before surface.configure is called. This is because there are many surface-based operations that really require the width and height and digging through the views didn't work well. Signed-off-by: Jason Ekstrand <jason@jlekstrand.net>
2013-10-13 07:38:11 +04:00
view->transform.enabled = 1;
/* Otherwise identity matrix, but with x and y translation. */
Split the geometry information from weston_surface out into weston_view The weston_surface structure is split into two structures: * The weston_surface structure storres everything required for a client-side or server-side surface. This includes buffers; callbacks; backend private data; input, damage, and opaque regions; and a few other bookkeeping bits. * The weston_view structure represents an entity in the scenegraph and storres all of the geometry information. This includes clip region, alpha, position, and the transformation list as well as all of the temporary information derived from the geometry state. Because a view, and not a surface, is a scenegraph element, the view is what is placed in layers and planes. There are a few things worth noting about the surface/view split: 1. This is *not* a modification to the protocol. It is, instead, a modification to Weston's internal scenegraph to allow a single surface to exist in multiple places at a time. Clients are completely unaware of how many views to a particular surface exist. 2. A view is considered a direct child of a surface and is destroyed when the surface is destroyed. Because of this, the view.surface pointer is always valid and non-null. 3. The compositor's surface_list is replaced with a view_list. Due to subsurfaces, building the view list is a little more complicated than it used to be and involves building a tree of views on the fly whenever subsurfaces are used. However, this means that backends can remain completely subsurface-agnostic. 4. Surfaces and views both keep track of which outputs they are on. 5. The weston_surface structure now has width and height fields. These are populated when a new buffer is attached before surface.configure is called. This is because there are many surface-based operations that really require the width and height and digging through the views didn't work well. Signed-off-by: Jason Ekstrand <jason@jlekstrand.net>
2013-10-13 07:38:11 +04:00
view->transform.position.matrix.type = WESTON_MATRIX_TRANSFORM_TRANSLATE;
view->transform.position.matrix.d[12] = view->geometry.x;
view->transform.position.matrix.d[13] = view->geometry.y;
weston_matrix_init(matrix);
Split the geometry information from weston_surface out into weston_view The weston_surface structure is split into two structures: * The weston_surface structure storres everything required for a client-side or server-side surface. This includes buffers; callbacks; backend private data; input, damage, and opaque regions; and a few other bookkeeping bits. * The weston_view structure represents an entity in the scenegraph and storres all of the geometry information. This includes clip region, alpha, position, and the transformation list as well as all of the temporary information derived from the geometry state. Because a view, and not a surface, is a scenegraph element, the view is what is placed in layers and planes. There are a few things worth noting about the surface/view split: 1. This is *not* a modification to the protocol. It is, instead, a modification to Weston's internal scenegraph to allow a single surface to exist in multiple places at a time. Clients are completely unaware of how many views to a particular surface exist. 2. A view is considered a direct child of a surface and is destroyed when the surface is destroyed. Because of this, the view.surface pointer is always valid and non-null. 3. The compositor's surface_list is replaced with a view_list. Due to subsurfaces, building the view list is a little more complicated than it used to be and involves building a tree of views on the fly whenever subsurfaces are used. However, this means that backends can remain completely subsurface-agnostic. 4. Surfaces and views both keep track of which outputs they are on. 5. The weston_surface structure now has width and height fields. These are populated when a new buffer is attached before surface.configure is called. This is because there are many surface-based operations that really require the width and height and digging through the views didn't work well. Signed-off-by: Jason Ekstrand <jason@jlekstrand.net>
2013-10-13 07:38:11 +04:00
wl_list_for_each(tform, &view->geometry.transformation_list, link)
weston_matrix_multiply(matrix, &tform->matrix);
if (parent)
weston_matrix_multiply(matrix, &parent->transform.matrix);
if (weston_matrix_invert(inverse, matrix) < 0) {
/* Oops, bad total transformation, not invertible */
Split the geometry information from weston_surface out into weston_view The weston_surface structure is split into two structures: * The weston_surface structure storres everything required for a client-side or server-side surface. This includes buffers; callbacks; backend private data; input, damage, and opaque regions; and a few other bookkeeping bits. * The weston_view structure represents an entity in the scenegraph and storres all of the geometry information. This includes clip region, alpha, position, and the transformation list as well as all of the temporary information derived from the geometry state. Because a view, and not a surface, is a scenegraph element, the view is what is placed in layers and planes. There are a few things worth noting about the surface/view split: 1. This is *not* a modification to the protocol. It is, instead, a modification to Weston's internal scenegraph to allow a single surface to exist in multiple places at a time. Clients are completely unaware of how many views to a particular surface exist. 2. A view is considered a direct child of a surface and is destroyed when the surface is destroyed. Because of this, the view.surface pointer is always valid and non-null. 3. The compositor's surface_list is replaced with a view_list. Due to subsurfaces, building the view list is a little more complicated than it used to be and involves building a tree of views on the fly whenever subsurfaces are used. However, this means that backends can remain completely subsurface-agnostic. 4. Surfaces and views both keep track of which outputs they are on. 5. The weston_surface structure now has width and height fields. These are populated when a new buffer is attached before surface.configure is called. This is because there are many surface-based operations that really require the width and height and digging through the views didn't work well. Signed-off-by: Jason Ekstrand <jason@jlekstrand.net>
2013-10-13 07:38:11 +04:00
weston_log("error: weston_view %p"
" transformation not invertible.\n", view);
return -1;
}
if (view->alpha == 1.0 &&
matrix->type == WESTON_MATRIX_TRANSFORM_TRANSLATE) {
pixman_region32_copy(&view->transform.opaque,
&view->surface->opaque);
pixman_region32_translate(&view->transform.opaque,
matrix->d[12],
matrix->d[13]);
}
compositor: add weston_view_set_mask() API and state Add API for setting a clip ('scissor' in the code) rectangle per view, in surface coordinates. Ivi-shell requires this feature to be able to implement the IVI Layer Manager API, which includes clipping of surfaces. The names weston_view_set_mask() and weston_view_set_mask_infinite() mirror the existing weston_layer_set_mask*() functions. This view clipping complements the weston_layer clipping, because view clipping is defined in surface local coordinates, while layer mask/clipping is defined in global coordinates. View clipping requires explicit support from the renderers. Therefore a new Weston capability bit is added: WESTON_CAP_VIEW_CLIP_MASK. Shells (and all users) of this new API are required to check the capability bit is set before using the API. Otherwise the rendering will not be what they expect. View clips are inherited through the transformation inheritance mechanism. However, there are restrictions. The clip rectangle can be set only on the root view of a transformation inheritance tree. The additional transformations in child views must not rotate the coordinate axes. These restrictions avoid corner cases in clip inheritance, and keep the renderer implementations as simple as they are right now. Renderers only need to do an additional intersection with the clip rectangle which is always aligned to the surface coordinate system. For more details, see the API documentation in the patch. Signed-off-by: Pekka Paalanen <pekka.paalanen@collabora.co.uk> Reviewed-by: Nobuhiko Tanibata <NOBUHIKO_TANIBATA@xddp.denso.co.jp> Tested-by: Nobuhiko Tanibata <NOBUHIKO_TANIBATA@xddp.denso.co.jp> Reviewed-By: Derek Foreman <derekf@osg.samsung.com>
2015-02-16 15:39:11 +03:00
pixman_region32_init_rect(&surfregion, 0, 0,
view->surface->width, view->surface->height);
if (view->geometry.scissor_enabled)
pixman_region32_intersect(&surfregion, &surfregion,
&view->geometry.scissor);
surfbox = pixman_region32_extents(&surfregion);
view_compute_bbox(view, surfbox, &view->transform.boundingbox);
pixman_region32_fini(&surfregion);
return 0;
}
static struct weston_layer *
get_view_layer(struct weston_view *view)
{
if (view->parent_view)
return get_view_layer(view->parent_view);
return view->layer_link.layer;
}
WL_EXPORT void
Split the geometry information from weston_surface out into weston_view The weston_surface structure is split into two structures: * The weston_surface structure storres everything required for a client-side or server-side surface. This includes buffers; callbacks; backend private data; input, damage, and opaque regions; and a few other bookkeeping bits. * The weston_view structure represents an entity in the scenegraph and storres all of the geometry information. This includes clip region, alpha, position, and the transformation list as well as all of the temporary information derived from the geometry state. Because a view, and not a surface, is a scenegraph element, the view is what is placed in layers and planes. There are a few things worth noting about the surface/view split: 1. This is *not* a modification to the protocol. It is, instead, a modification to Weston's internal scenegraph to allow a single surface to exist in multiple places at a time. Clients are completely unaware of how many views to a particular surface exist. 2. A view is considered a direct child of a surface and is destroyed when the surface is destroyed. Because of this, the view.surface pointer is always valid and non-null. 3. The compositor's surface_list is replaced with a view_list. Due to subsurfaces, building the view list is a little more complicated than it used to be and involves building a tree of views on the fly whenever subsurfaces are used. However, this means that backends can remain completely subsurface-agnostic. 4. Surfaces and views both keep track of which outputs they are on. 5. The weston_surface structure now has width and height fields. These are populated when a new buffer is attached before surface.configure is called. This is because there are many surface-based operations that really require the width and height and digging through the views didn't work well. Signed-off-by: Jason Ekstrand <jason@jlekstrand.net>
2013-10-13 07:38:11 +04:00
weston_view_update_transform(struct weston_view *view)
{
Split the geometry information from weston_surface out into weston_view The weston_surface structure is split into two structures: * The weston_surface structure storres everything required for a client-side or server-side surface. This includes buffers; callbacks; backend private data; input, damage, and opaque regions; and a few other bookkeeping bits. * The weston_view structure represents an entity in the scenegraph and storres all of the geometry information. This includes clip region, alpha, position, and the transformation list as well as all of the temporary information derived from the geometry state. Because a view, and not a surface, is a scenegraph element, the view is what is placed in layers and planes. There are a few things worth noting about the surface/view split: 1. This is *not* a modification to the protocol. It is, instead, a modification to Weston's internal scenegraph to allow a single surface to exist in multiple places at a time. Clients are completely unaware of how many views to a particular surface exist. 2. A view is considered a direct child of a surface and is destroyed when the surface is destroyed. Because of this, the view.surface pointer is always valid and non-null. 3. The compositor's surface_list is replaced with a view_list. Due to subsurfaces, building the view list is a little more complicated than it used to be and involves building a tree of views on the fly whenever subsurfaces are used. However, this means that backends can remain completely subsurface-agnostic. 4. Surfaces and views both keep track of which outputs they are on. 5. The weston_surface structure now has width and height fields. These are populated when a new buffer is attached before surface.configure is called. This is because there are many surface-based operations that really require the width and height and digging through the views didn't work well. Signed-off-by: Jason Ekstrand <jason@jlekstrand.net>
2013-10-13 07:38:11 +04:00
struct weston_view *parent = view->geometry.parent;
struct weston_layer *layer;
pixman_region32_t mask;
Split the geometry information from weston_surface out into weston_view The weston_surface structure is split into two structures: * The weston_surface structure storres everything required for a client-side or server-side surface. This includes buffers; callbacks; backend private data; input, damage, and opaque regions; and a few other bookkeeping bits. * The weston_view structure represents an entity in the scenegraph and storres all of the geometry information. This includes clip region, alpha, position, and the transformation list as well as all of the temporary information derived from the geometry state. Because a view, and not a surface, is a scenegraph element, the view is what is placed in layers and planes. There are a few things worth noting about the surface/view split: 1. This is *not* a modification to the protocol. It is, instead, a modification to Weston's internal scenegraph to allow a single surface to exist in multiple places at a time. Clients are completely unaware of how many views to a particular surface exist. 2. A view is considered a direct child of a surface and is destroyed when the surface is destroyed. Because of this, the view.surface pointer is always valid and non-null. 3. The compositor's surface_list is replaced with a view_list. Due to subsurfaces, building the view list is a little more complicated than it used to be and involves building a tree of views on the fly whenever subsurfaces are used. However, this means that backends can remain completely subsurface-agnostic. 4. Surfaces and views both keep track of which outputs they are on. 5. The weston_surface structure now has width and height fields. These are populated when a new buffer is attached before surface.configure is called. This is because there are many surface-based operations that really require the width and height and digging through the views didn't work well. Signed-off-by: Jason Ekstrand <jason@jlekstrand.net>
2013-10-13 07:38:11 +04:00
if (!view->transform.dirty)
return;
if (parent)
Split the geometry information from weston_surface out into weston_view The weston_surface structure is split into two structures: * The weston_surface structure storres everything required for a client-side or server-side surface. This includes buffers; callbacks; backend private data; input, damage, and opaque regions; and a few other bookkeeping bits. * The weston_view structure represents an entity in the scenegraph and storres all of the geometry information. This includes clip region, alpha, position, and the transformation list as well as all of the temporary information derived from the geometry state. Because a view, and not a surface, is a scenegraph element, the view is what is placed in layers and planes. There are a few things worth noting about the surface/view split: 1. This is *not* a modification to the protocol. It is, instead, a modification to Weston's internal scenegraph to allow a single surface to exist in multiple places at a time. Clients are completely unaware of how many views to a particular surface exist. 2. A view is considered a direct child of a surface and is destroyed when the surface is destroyed. Because of this, the view.surface pointer is always valid and non-null. 3. The compositor's surface_list is replaced with a view_list. Due to subsurfaces, building the view list is a little more complicated than it used to be and involves building a tree of views on the fly whenever subsurfaces are used. However, this means that backends can remain completely subsurface-agnostic. 4. Surfaces and views both keep track of which outputs they are on. 5. The weston_surface structure now has width and height fields. These are populated when a new buffer is attached before surface.configure is called. This is because there are many surface-based operations that really require the width and height and digging through the views didn't work well. Signed-off-by: Jason Ekstrand <jason@jlekstrand.net>
2013-10-13 07:38:11 +04:00
weston_view_update_transform(parent);
Split the geometry information from weston_surface out into weston_view The weston_surface structure is split into two structures: * The weston_surface structure storres everything required for a client-side or server-side surface. This includes buffers; callbacks; backend private data; input, damage, and opaque regions; and a few other bookkeeping bits. * The weston_view structure represents an entity in the scenegraph and storres all of the geometry information. This includes clip region, alpha, position, and the transformation list as well as all of the temporary information derived from the geometry state. Because a view, and not a surface, is a scenegraph element, the view is what is placed in layers and planes. There are a few things worth noting about the surface/view split: 1. This is *not* a modification to the protocol. It is, instead, a modification to Weston's internal scenegraph to allow a single surface to exist in multiple places at a time. Clients are completely unaware of how many views to a particular surface exist. 2. A view is considered a direct child of a surface and is destroyed when the surface is destroyed. Because of this, the view.surface pointer is always valid and non-null. 3. The compositor's surface_list is replaced with a view_list. Due to subsurfaces, building the view list is a little more complicated than it used to be and involves building a tree of views on the fly whenever subsurfaces are used. However, this means that backends can remain completely subsurface-agnostic. 4. Surfaces and views both keep track of which outputs they are on. 5. The weston_surface structure now has width and height fields. These are populated when a new buffer is attached before surface.configure is called. This is because there are many surface-based operations that really require the width and height and digging through the views didn't work well. Signed-off-by: Jason Ekstrand <jason@jlekstrand.net>
2013-10-13 07:38:11 +04:00
view->transform.dirty = 0;
Split the geometry information from weston_surface out into weston_view The weston_surface structure is split into two structures: * The weston_surface structure storres everything required for a client-side or server-side surface. This includes buffers; callbacks; backend private data; input, damage, and opaque regions; and a few other bookkeeping bits. * The weston_view structure represents an entity in the scenegraph and storres all of the geometry information. This includes clip region, alpha, position, and the transformation list as well as all of the temporary information derived from the geometry state. Because a view, and not a surface, is a scenegraph element, the view is what is placed in layers and planes. There are a few things worth noting about the surface/view split: 1. This is *not* a modification to the protocol. It is, instead, a modification to Weston's internal scenegraph to allow a single surface to exist in multiple places at a time. Clients are completely unaware of how many views to a particular surface exist. 2. A view is considered a direct child of a surface and is destroyed when the surface is destroyed. Because of this, the view.surface pointer is always valid and non-null. 3. The compositor's surface_list is replaced with a view_list. Due to subsurfaces, building the view list is a little more complicated than it used to be and involves building a tree of views on the fly whenever subsurfaces are used. However, this means that backends can remain completely subsurface-agnostic. 4. Surfaces and views both keep track of which outputs they are on. 5. The weston_surface structure now has width and height fields. These are populated when a new buffer is attached before surface.configure is called. This is because there are many surface-based operations that really require the width and height and digging through the views didn't work well. Signed-off-by: Jason Ekstrand <jason@jlekstrand.net>
2013-10-13 07:38:11 +04:00
weston_view_damage_below(view);
Split the geometry information from weston_surface out into weston_view The weston_surface structure is split into two structures: * The weston_surface structure storres everything required for a client-side or server-side surface. This includes buffers; callbacks; backend private data; input, damage, and opaque regions; and a few other bookkeeping bits. * The weston_view structure represents an entity in the scenegraph and storres all of the geometry information. This includes clip region, alpha, position, and the transformation list as well as all of the temporary information derived from the geometry state. Because a view, and not a surface, is a scenegraph element, the view is what is placed in layers and planes. There are a few things worth noting about the surface/view split: 1. This is *not* a modification to the protocol. It is, instead, a modification to Weston's internal scenegraph to allow a single surface to exist in multiple places at a time. Clients are completely unaware of how many views to a particular surface exist. 2. A view is considered a direct child of a surface and is destroyed when the surface is destroyed. Because of this, the view.surface pointer is always valid and non-null. 3. The compositor's surface_list is replaced with a view_list. Due to subsurfaces, building the view list is a little more complicated than it used to be and involves building a tree of views on the fly whenever subsurfaces are used. However, this means that backends can remain completely subsurface-agnostic. 4. Surfaces and views both keep track of which outputs they are on. 5. The weston_surface structure now has width and height fields. These are populated when a new buffer is attached before surface.configure is called. This is because there are many surface-based operations that really require the width and height and digging through the views didn't work well. Signed-off-by: Jason Ekstrand <jason@jlekstrand.net>
2013-10-13 07:38:11 +04:00
pixman_region32_fini(&view->transform.boundingbox);
pixman_region32_fini(&view->transform.opaque);
pixman_region32_init(&view->transform.opaque);
/* transform.position is always in transformation_list */
Split the geometry information from weston_surface out into weston_view The weston_surface structure is split into two structures: * The weston_surface structure storres everything required for a client-side or server-side surface. This includes buffers; callbacks; backend private data; input, damage, and opaque regions; and a few other bookkeeping bits. * The weston_view structure represents an entity in the scenegraph and storres all of the geometry information. This includes clip region, alpha, position, and the transformation list as well as all of the temporary information derived from the geometry state. Because a view, and not a surface, is a scenegraph element, the view is what is placed in layers and planes. There are a few things worth noting about the surface/view split: 1. This is *not* a modification to the protocol. It is, instead, a modification to Weston's internal scenegraph to allow a single surface to exist in multiple places at a time. Clients are completely unaware of how many views to a particular surface exist. 2. A view is considered a direct child of a surface and is destroyed when the surface is destroyed. Because of this, the view.surface pointer is always valid and non-null. 3. The compositor's surface_list is replaced with a view_list. Due to subsurfaces, building the view list is a little more complicated than it used to be and involves building a tree of views on the fly whenever subsurfaces are used. However, this means that backends can remain completely subsurface-agnostic. 4. Surfaces and views both keep track of which outputs they are on. 5. The weston_surface structure now has width and height fields. These are populated when a new buffer is attached before surface.configure is called. This is because there are many surface-based operations that really require the width and height and digging through the views didn't work well. Signed-off-by: Jason Ekstrand <jason@jlekstrand.net>
2013-10-13 07:38:11 +04:00
if (view->geometry.transformation_list.next ==
&view->transform.position.link &&
view->geometry.transformation_list.prev ==
&view->transform.position.link &&
!parent) {
Split the geometry information from weston_surface out into weston_view The weston_surface structure is split into two structures: * The weston_surface structure storres everything required for a client-side or server-side surface. This includes buffers; callbacks; backend private data; input, damage, and opaque regions; and a few other bookkeeping bits. * The weston_view structure represents an entity in the scenegraph and storres all of the geometry information. This includes clip region, alpha, position, and the transformation list as well as all of the temporary information derived from the geometry state. Because a view, and not a surface, is a scenegraph element, the view is what is placed in layers and planes. There are a few things worth noting about the surface/view split: 1. This is *not* a modification to the protocol. It is, instead, a modification to Weston's internal scenegraph to allow a single surface to exist in multiple places at a time. Clients are completely unaware of how many views to a particular surface exist. 2. A view is considered a direct child of a surface and is destroyed when the surface is destroyed. Because of this, the view.surface pointer is always valid and non-null. 3. The compositor's surface_list is replaced with a view_list. Due to subsurfaces, building the view list is a little more complicated than it used to be and involves building a tree of views on the fly whenever subsurfaces are used. However, this means that backends can remain completely subsurface-agnostic. 4. Surfaces and views both keep track of which outputs they are on. 5. The weston_surface structure now has width and height fields. These are populated when a new buffer is attached before surface.configure is called. This is because there are many surface-based operations that really require the width and height and digging through the views didn't work well. Signed-off-by: Jason Ekstrand <jason@jlekstrand.net>
2013-10-13 07:38:11 +04:00
weston_view_update_transform_disable(view);
} else {
Split the geometry information from weston_surface out into weston_view The weston_surface structure is split into two structures: * The weston_surface structure storres everything required for a client-side or server-side surface. This includes buffers; callbacks; backend private data; input, damage, and opaque regions; and a few other bookkeeping bits. * The weston_view structure represents an entity in the scenegraph and storres all of the geometry information. This includes clip region, alpha, position, and the transformation list as well as all of the temporary information derived from the geometry state. Because a view, and not a surface, is a scenegraph element, the view is what is placed in layers and planes. There are a few things worth noting about the surface/view split: 1. This is *not* a modification to the protocol. It is, instead, a modification to Weston's internal scenegraph to allow a single surface to exist in multiple places at a time. Clients are completely unaware of how many views to a particular surface exist. 2. A view is considered a direct child of a surface and is destroyed when the surface is destroyed. Because of this, the view.surface pointer is always valid and non-null. 3. The compositor's surface_list is replaced with a view_list. Due to subsurfaces, building the view list is a little more complicated than it used to be and involves building a tree of views on the fly whenever subsurfaces are used. However, this means that backends can remain completely subsurface-agnostic. 4. Surfaces and views both keep track of which outputs they are on. 5. The weston_surface structure now has width and height fields. These are populated when a new buffer is attached before surface.configure is called. This is because there are many surface-based operations that really require the width and height and digging through the views didn't work well. Signed-off-by: Jason Ekstrand <jason@jlekstrand.net>
2013-10-13 07:38:11 +04:00
if (weston_view_update_transform_enable(view) < 0)
weston_view_update_transform_disable(view);
}
layer = get_view_layer(view);
if (layer) {
pixman_region32_init_with_extents(&mask, &layer->mask);
pixman_region32_intersect(&view->transform.boundingbox,
&view->transform.boundingbox, &mask);
pixman_region32_intersect(&view->transform.opaque,
&view->transform.opaque, &mask);
pixman_region32_fini(&mask);
}
compositor: add weston_view_set_mask() API and state Add API for setting a clip ('scissor' in the code) rectangle per view, in surface coordinates. Ivi-shell requires this feature to be able to implement the IVI Layer Manager API, which includes clipping of surfaces. The names weston_view_set_mask() and weston_view_set_mask_infinite() mirror the existing weston_layer_set_mask*() functions. This view clipping complements the weston_layer clipping, because view clipping is defined in surface local coordinates, while layer mask/clipping is defined in global coordinates. View clipping requires explicit support from the renderers. Therefore a new Weston capability bit is added: WESTON_CAP_VIEW_CLIP_MASK. Shells (and all users) of this new API are required to check the capability bit is set before using the API. Otherwise the rendering will not be what they expect. View clips are inherited through the transformation inheritance mechanism. However, there are restrictions. The clip rectangle can be set only on the root view of a transformation inheritance tree. The additional transformations in child views must not rotate the coordinate axes. These restrictions avoid corner cases in clip inheritance, and keep the renderer implementations as simple as they are right now. Renderers only need to do an additional intersection with the clip rectangle which is always aligned to the surface coordinate system. For more details, see the API documentation in the patch. Signed-off-by: Pekka Paalanen <pekka.paalanen@collabora.co.uk> Reviewed-by: Nobuhiko Tanibata <NOBUHIKO_TANIBATA@xddp.denso.co.jp> Tested-by: Nobuhiko Tanibata <NOBUHIKO_TANIBATA@xddp.denso.co.jp> Reviewed-By: Derek Foreman <derekf@osg.samsung.com>
2015-02-16 15:39:11 +03:00
if (parent) {
if (parent->geometry.scissor_enabled) {
view->geometry.scissor_enabled = true;
weston_view_transfer_scissor(parent, view);
} else {
view->geometry.scissor_enabled = false;
}
}
Split the geometry information from weston_surface out into weston_view The weston_surface structure is split into two structures: * The weston_surface structure storres everything required for a client-side or server-side surface. This includes buffers; callbacks; backend private data; input, damage, and opaque regions; and a few other bookkeeping bits. * The weston_view structure represents an entity in the scenegraph and storres all of the geometry information. This includes clip region, alpha, position, and the transformation list as well as all of the temporary information derived from the geometry state. Because a view, and not a surface, is a scenegraph element, the view is what is placed in layers and planes. There are a few things worth noting about the surface/view split: 1. This is *not* a modification to the protocol. It is, instead, a modification to Weston's internal scenegraph to allow a single surface to exist in multiple places at a time. Clients are completely unaware of how many views to a particular surface exist. 2. A view is considered a direct child of a surface and is destroyed when the surface is destroyed. Because of this, the view.surface pointer is always valid and non-null. 3. The compositor's surface_list is replaced with a view_list. Due to subsurfaces, building the view list is a little more complicated than it used to be and involves building a tree of views on the fly whenever subsurfaces are used. However, this means that backends can remain completely subsurface-agnostic. 4. Surfaces and views both keep track of which outputs they are on. 5. The weston_surface structure now has width and height fields. These are populated when a new buffer is attached before surface.configure is called. This is because there are many surface-based operations that really require the width and height and digging through the views didn't work well. Signed-off-by: Jason Ekstrand <jason@jlekstrand.net>
2013-10-13 07:38:11 +04:00
weston_view_damage_below(view);
Split the geometry information from weston_surface out into weston_view The weston_surface structure is split into two structures: * The weston_surface structure storres everything required for a client-side or server-side surface. This includes buffers; callbacks; backend private data; input, damage, and opaque regions; and a few other bookkeeping bits. * The weston_view structure represents an entity in the scenegraph and storres all of the geometry information. This includes clip region, alpha, position, and the transformation list as well as all of the temporary information derived from the geometry state. Because a view, and not a surface, is a scenegraph element, the view is what is placed in layers and planes. There are a few things worth noting about the surface/view split: 1. This is *not* a modification to the protocol. It is, instead, a modification to Weston's internal scenegraph to allow a single surface to exist in multiple places at a time. Clients are completely unaware of how many views to a particular surface exist. 2. A view is considered a direct child of a surface and is destroyed when the surface is destroyed. Because of this, the view.surface pointer is always valid and non-null. 3. The compositor's surface_list is replaced with a view_list. Due to subsurfaces, building the view list is a little more complicated than it used to be and involves building a tree of views on the fly whenever subsurfaces are used. However, this means that backends can remain completely subsurface-agnostic. 4. Surfaces and views both keep track of which outputs they are on. 5. The weston_surface structure now has width and height fields. These are populated when a new buffer is attached before surface.configure is called. This is because there are many surface-based operations that really require the width and height and digging through the views didn't work well. Signed-off-by: Jason Ekstrand <jason@jlekstrand.net>
2013-10-13 07:38:11 +04:00
weston_view_assign_output(view);
Split the geometry information from weston_surface out into weston_view The weston_surface structure is split into two structures: * The weston_surface structure storres everything required for a client-side or server-side surface. This includes buffers; callbacks; backend private data; input, damage, and opaque regions; and a few other bookkeeping bits. * The weston_view structure represents an entity in the scenegraph and storres all of the geometry information. This includes clip region, alpha, position, and the transformation list as well as all of the temporary information derived from the geometry state. Because a view, and not a surface, is a scenegraph element, the view is what is placed in layers and planes. There are a few things worth noting about the surface/view split: 1. This is *not* a modification to the protocol. It is, instead, a modification to Weston's internal scenegraph to allow a single surface to exist in multiple places at a time. Clients are completely unaware of how many views to a particular surface exist. 2. A view is considered a direct child of a surface and is destroyed when the surface is destroyed. Because of this, the view.surface pointer is always valid and non-null. 3. The compositor's surface_list is replaced with a view_list. Due to subsurfaces, building the view list is a little more complicated than it used to be and involves building a tree of views on the fly whenever subsurfaces are used. However, this means that backends can remain completely subsurface-agnostic. 4. Surfaces and views both keep track of which outputs they are on. 5. The weston_surface structure now has width and height fields. These are populated when a new buffer is attached before surface.configure is called. This is because there are many surface-based operations that really require the width and height and digging through the views didn't work well. Signed-off-by: Jason Ekstrand <jason@jlekstrand.net>
2013-10-13 07:38:11 +04:00
wl_signal_emit(&view->surface->compositor->transform_signal,
view->surface);
}
WL_EXPORT void
Split the geometry information from weston_surface out into weston_view The weston_surface structure is split into two structures: * The weston_surface structure storres everything required for a client-side or server-side surface. This includes buffers; callbacks; backend private data; input, damage, and opaque regions; and a few other bookkeeping bits. * The weston_view structure represents an entity in the scenegraph and storres all of the geometry information. This includes clip region, alpha, position, and the transformation list as well as all of the temporary information derived from the geometry state. Because a view, and not a surface, is a scenegraph element, the view is what is placed in layers and planes. There are a few things worth noting about the surface/view split: 1. This is *not* a modification to the protocol. It is, instead, a modification to Weston's internal scenegraph to allow a single surface to exist in multiple places at a time. Clients are completely unaware of how many views to a particular surface exist. 2. A view is considered a direct child of a surface and is destroyed when the surface is destroyed. Because of this, the view.surface pointer is always valid and non-null. 3. The compositor's surface_list is replaced with a view_list. Due to subsurfaces, building the view list is a little more complicated than it used to be and involves building a tree of views on the fly whenever subsurfaces are used. However, this means that backends can remain completely subsurface-agnostic. 4. Surfaces and views both keep track of which outputs they are on. 5. The weston_surface structure now has width and height fields. These are populated when a new buffer is attached before surface.configure is called. This is because there are many surface-based operations that really require the width and height and digging through the views didn't work well. Signed-off-by: Jason Ekstrand <jason@jlekstrand.net>
2013-10-13 07:38:11 +04:00
weston_view_geometry_dirty(struct weston_view *view)
{
Split the geometry information from weston_surface out into weston_view The weston_surface structure is split into two structures: * The weston_surface structure storres everything required for a client-side or server-side surface. This includes buffers; callbacks; backend private data; input, damage, and opaque regions; and a few other bookkeeping bits. * The weston_view structure represents an entity in the scenegraph and storres all of the geometry information. This includes clip region, alpha, position, and the transformation list as well as all of the temporary information derived from the geometry state. Because a view, and not a surface, is a scenegraph element, the view is what is placed in layers and planes. There are a few things worth noting about the surface/view split: 1. This is *not* a modification to the protocol. It is, instead, a modification to Weston's internal scenegraph to allow a single surface to exist in multiple places at a time. Clients are completely unaware of how many views to a particular surface exist. 2. A view is considered a direct child of a surface and is destroyed when the surface is destroyed. Because of this, the view.surface pointer is always valid and non-null. 3. The compositor's surface_list is replaced with a view_list. Due to subsurfaces, building the view list is a little more complicated than it used to be and involves building a tree of views on the fly whenever subsurfaces are used. However, this means that backends can remain completely subsurface-agnostic. 4. Surfaces and views both keep track of which outputs they are on. 5. The weston_surface structure now has width and height fields. These are populated when a new buffer is attached before surface.configure is called. This is because there are many surface-based operations that really require the width and height and digging through the views didn't work well. Signed-off-by: Jason Ekstrand <jason@jlekstrand.net>
2013-10-13 07:38:11 +04:00
struct weston_view *child;
/*
Split the geometry information from weston_surface out into weston_view The weston_surface structure is split into two structures: * The weston_surface structure storres everything required for a client-side or server-side surface. This includes buffers; callbacks; backend private data; input, damage, and opaque regions; and a few other bookkeeping bits. * The weston_view structure represents an entity in the scenegraph and storres all of the geometry information. This includes clip region, alpha, position, and the transformation list as well as all of the temporary information derived from the geometry state. Because a view, and not a surface, is a scenegraph element, the view is what is placed in layers and planes. There are a few things worth noting about the surface/view split: 1. This is *not* a modification to the protocol. It is, instead, a modification to Weston's internal scenegraph to allow a single surface to exist in multiple places at a time. Clients are completely unaware of how many views to a particular surface exist. 2. A view is considered a direct child of a surface and is destroyed when the surface is destroyed. Because of this, the view.surface pointer is always valid and non-null. 3. The compositor's surface_list is replaced with a view_list. Due to subsurfaces, building the view list is a little more complicated than it used to be and involves building a tree of views on the fly whenever subsurfaces are used. However, this means that backends can remain completely subsurface-agnostic. 4. Surfaces and views both keep track of which outputs they are on. 5. The weston_surface structure now has width and height fields. These are populated when a new buffer is attached before surface.configure is called. This is because there are many surface-based operations that really require the width and height and digging through the views didn't work well. Signed-off-by: Jason Ekstrand <jason@jlekstrand.net>
2013-10-13 07:38:11 +04:00
* The invariant: if view->geometry.dirty, then all views
* in view->geometry.child_list have geometry.dirty too.
* Corollary: if not parent->geometry.dirty, then all ancestors
* are not dirty.
*/
Split the geometry information from weston_surface out into weston_view The weston_surface structure is split into two structures: * The weston_surface structure storres everything required for a client-side or server-side surface. This includes buffers; callbacks; backend private data; input, damage, and opaque regions; and a few other bookkeeping bits. * The weston_view structure represents an entity in the scenegraph and storres all of the geometry information. This includes clip region, alpha, position, and the transformation list as well as all of the temporary information derived from the geometry state. Because a view, and not a surface, is a scenegraph element, the view is what is placed in layers and planes. There are a few things worth noting about the surface/view split: 1. This is *not* a modification to the protocol. It is, instead, a modification to Weston's internal scenegraph to allow a single surface to exist in multiple places at a time. Clients are completely unaware of how many views to a particular surface exist. 2. A view is considered a direct child of a surface and is destroyed when the surface is destroyed. Because of this, the view.surface pointer is always valid and non-null. 3. The compositor's surface_list is replaced with a view_list. Due to subsurfaces, building the view list is a little more complicated than it used to be and involves building a tree of views on the fly whenever subsurfaces are used. However, this means that backends can remain completely subsurface-agnostic. 4. Surfaces and views both keep track of which outputs they are on. 5. The weston_surface structure now has width and height fields. These are populated when a new buffer is attached before surface.configure is called. This is because there are many surface-based operations that really require the width and height and digging through the views didn't work well. Signed-off-by: Jason Ekstrand <jason@jlekstrand.net>
2013-10-13 07:38:11 +04:00
if (view->transform.dirty)
return;
Split the geometry information from weston_surface out into weston_view The weston_surface structure is split into two structures: * The weston_surface structure storres everything required for a client-side or server-side surface. This includes buffers; callbacks; backend private data; input, damage, and opaque regions; and a few other bookkeeping bits. * The weston_view structure represents an entity in the scenegraph and storres all of the geometry information. This includes clip region, alpha, position, and the transformation list as well as all of the temporary information derived from the geometry state. Because a view, and not a surface, is a scenegraph element, the view is what is placed in layers and planes. There are a few things worth noting about the surface/view split: 1. This is *not* a modification to the protocol. It is, instead, a modification to Weston's internal scenegraph to allow a single surface to exist in multiple places at a time. Clients are completely unaware of how many views to a particular surface exist. 2. A view is considered a direct child of a surface and is destroyed when the surface is destroyed. Because of this, the view.surface pointer is always valid and non-null. 3. The compositor's surface_list is replaced with a view_list. Due to subsurfaces, building the view list is a little more complicated than it used to be and involves building a tree of views on the fly whenever subsurfaces are used. However, this means that backends can remain completely subsurface-agnostic. 4. Surfaces and views both keep track of which outputs they are on. 5. The weston_surface structure now has width and height fields. These are populated when a new buffer is attached before surface.configure is called. This is because there are many surface-based operations that really require the width and height and digging through the views didn't work well. Signed-off-by: Jason Ekstrand <jason@jlekstrand.net>
2013-10-13 07:38:11 +04:00
view->transform.dirty = 1;
Split the geometry information from weston_surface out into weston_view The weston_surface structure is split into two structures: * The weston_surface structure storres everything required for a client-side or server-side surface. This includes buffers; callbacks; backend private data; input, damage, and opaque regions; and a few other bookkeeping bits. * The weston_view structure represents an entity in the scenegraph and storres all of the geometry information. This includes clip region, alpha, position, and the transformation list as well as all of the temporary information derived from the geometry state. Because a view, and not a surface, is a scenegraph element, the view is what is placed in layers and planes. There are a few things worth noting about the surface/view split: 1. This is *not* a modification to the protocol. It is, instead, a modification to Weston's internal scenegraph to allow a single surface to exist in multiple places at a time. Clients are completely unaware of how many views to a particular surface exist. 2. A view is considered a direct child of a surface and is destroyed when the surface is destroyed. Because of this, the view.surface pointer is always valid and non-null. 3. The compositor's surface_list is replaced with a view_list. Due to subsurfaces, building the view list is a little more complicated than it used to be and involves building a tree of views on the fly whenever subsurfaces are used. However, this means that backends can remain completely subsurface-agnostic. 4. Surfaces and views both keep track of which outputs they are on. 5. The weston_surface structure now has width and height fields. These are populated when a new buffer is attached before surface.configure is called. This is because there are many surface-based operations that really require the width and height and digging through the views didn't work well. Signed-off-by: Jason Ekstrand <jason@jlekstrand.net>
2013-10-13 07:38:11 +04:00
wl_list_for_each(child, &view->geometry.child_list,
geometry.parent_link)
Split the geometry information from weston_surface out into weston_view The weston_surface structure is split into two structures: * The weston_surface structure storres everything required for a client-side or server-side surface. This includes buffers; callbacks; backend private data; input, damage, and opaque regions; and a few other bookkeeping bits. * The weston_view structure represents an entity in the scenegraph and storres all of the geometry information. This includes clip region, alpha, position, and the transformation list as well as all of the temporary information derived from the geometry state. Because a view, and not a surface, is a scenegraph element, the view is what is placed in layers and planes. There are a few things worth noting about the surface/view split: 1. This is *not* a modification to the protocol. It is, instead, a modification to Weston's internal scenegraph to allow a single surface to exist in multiple places at a time. Clients are completely unaware of how many views to a particular surface exist. 2. A view is considered a direct child of a surface and is destroyed when the surface is destroyed. Because of this, the view.surface pointer is always valid and non-null. 3. The compositor's surface_list is replaced with a view_list. Due to subsurfaces, building the view list is a little more complicated than it used to be and involves building a tree of views on the fly whenever subsurfaces are used. However, this means that backends can remain completely subsurface-agnostic. 4. Surfaces and views both keep track of which outputs they are on. 5. The weston_surface structure now has width and height fields. These are populated when a new buffer is attached before surface.configure is called. This is because there are many surface-based operations that really require the width and height and digging through the views didn't work well. Signed-off-by: Jason Ekstrand <jason@jlekstrand.net>
2013-10-13 07:38:11 +04:00
weston_view_geometry_dirty(child);
}
WL_EXPORT void
Split the geometry information from weston_surface out into weston_view The weston_surface structure is split into two structures: * The weston_surface structure storres everything required for a client-side or server-side surface. This includes buffers; callbacks; backend private data; input, damage, and opaque regions; and a few other bookkeeping bits. * The weston_view structure represents an entity in the scenegraph and storres all of the geometry information. This includes clip region, alpha, position, and the transformation list as well as all of the temporary information derived from the geometry state. Because a view, and not a surface, is a scenegraph element, the view is what is placed in layers and planes. There are a few things worth noting about the surface/view split: 1. This is *not* a modification to the protocol. It is, instead, a modification to Weston's internal scenegraph to allow a single surface to exist in multiple places at a time. Clients are completely unaware of how many views to a particular surface exist. 2. A view is considered a direct child of a surface and is destroyed when the surface is destroyed. Because of this, the view.surface pointer is always valid and non-null. 3. The compositor's surface_list is replaced with a view_list. Due to subsurfaces, building the view list is a little more complicated than it used to be and involves building a tree of views on the fly whenever subsurfaces are used. However, this means that backends can remain completely subsurface-agnostic. 4. Surfaces and views both keep track of which outputs they are on. 5. The weston_surface structure now has width and height fields. These are populated when a new buffer is attached before surface.configure is called. This is because there are many surface-based operations that really require the width and height and digging through the views didn't work well. Signed-off-by: Jason Ekstrand <jason@jlekstrand.net>
2013-10-13 07:38:11 +04:00
weston_view_to_global_fixed(struct weston_view *view,
wl_fixed_t vx, wl_fixed_t vy,
wl_fixed_t *x, wl_fixed_t *y)
{
2012-09-30 04:57:21 +04:00
float xf, yf;
Split the geometry information from weston_surface out into weston_view The weston_surface structure is split into two structures: * The weston_surface structure storres everything required for a client-side or server-side surface. This includes buffers; callbacks; backend private data; input, damage, and opaque regions; and a few other bookkeeping bits. * The weston_view structure represents an entity in the scenegraph and storres all of the geometry information. This includes clip region, alpha, position, and the transformation list as well as all of the temporary information derived from the geometry state. Because a view, and not a surface, is a scenegraph element, the view is what is placed in layers and planes. There are a few things worth noting about the surface/view split: 1. This is *not* a modification to the protocol. It is, instead, a modification to Weston's internal scenegraph to allow a single surface to exist in multiple places at a time. Clients are completely unaware of how many views to a particular surface exist. 2. A view is considered a direct child of a surface and is destroyed when the surface is destroyed. Because of this, the view.surface pointer is always valid and non-null. 3. The compositor's surface_list is replaced with a view_list. Due to subsurfaces, building the view list is a little more complicated than it used to be and involves building a tree of views on the fly whenever subsurfaces are used. However, this means that backends can remain completely subsurface-agnostic. 4. Surfaces and views both keep track of which outputs they are on. 5. The weston_surface structure now has width and height fields. These are populated when a new buffer is attached before surface.configure is called. This is because there are many surface-based operations that really require the width and height and digging through the views didn't work well. Signed-off-by: Jason Ekstrand <jason@jlekstrand.net>
2013-10-13 07:38:11 +04:00
weston_view_to_global_float(view,
wl_fixed_to_double(vx),
wl_fixed_to_double(vy),
&xf, &yf);
*x = wl_fixed_from_double(xf);
*y = wl_fixed_from_double(yf);
}
WL_EXPORT void
Split the geometry information from weston_surface out into weston_view The weston_surface structure is split into two structures: * The weston_surface structure storres everything required for a client-side or server-side surface. This includes buffers; callbacks; backend private data; input, damage, and opaque regions; and a few other bookkeeping bits. * The weston_view structure represents an entity in the scenegraph and storres all of the geometry information. This includes clip region, alpha, position, and the transformation list as well as all of the temporary information derived from the geometry state. Because a view, and not a surface, is a scenegraph element, the view is what is placed in layers and planes. There are a few things worth noting about the surface/view split: 1. This is *not* a modification to the protocol. It is, instead, a modification to Weston's internal scenegraph to allow a single surface to exist in multiple places at a time. Clients are completely unaware of how many views to a particular surface exist. 2. A view is considered a direct child of a surface and is destroyed when the surface is destroyed. Because of this, the view.surface pointer is always valid and non-null. 3. The compositor's surface_list is replaced with a view_list. Due to subsurfaces, building the view list is a little more complicated than it used to be and involves building a tree of views on the fly whenever subsurfaces are used. However, this means that backends can remain completely subsurface-agnostic. 4. Surfaces and views both keep track of which outputs they are on. 5. The weston_surface structure now has width and height fields. These are populated when a new buffer is attached before surface.configure is called. This is because there are many surface-based operations that really require the width and height and digging through the views didn't work well. Signed-off-by: Jason Ekstrand <jason@jlekstrand.net>
2013-10-13 07:38:11 +04:00
weston_view_from_global_float(struct weston_view *view,
float x, float y, float *vx, float *vy)
{
Split the geometry information from weston_surface out into weston_view The weston_surface structure is split into two structures: * The weston_surface structure storres everything required for a client-side or server-side surface. This includes buffers; callbacks; backend private data; input, damage, and opaque regions; and a few other bookkeeping bits. * The weston_view structure represents an entity in the scenegraph and storres all of the geometry information. This includes clip region, alpha, position, and the transformation list as well as all of the temporary information derived from the geometry state. Because a view, and not a surface, is a scenegraph element, the view is what is placed in layers and planes. There are a few things worth noting about the surface/view split: 1. This is *not* a modification to the protocol. It is, instead, a modification to Weston's internal scenegraph to allow a single surface to exist in multiple places at a time. Clients are completely unaware of how many views to a particular surface exist. 2. A view is considered a direct child of a surface and is destroyed when the surface is destroyed. Because of this, the view.surface pointer is always valid and non-null. 3. The compositor's surface_list is replaced with a view_list. Due to subsurfaces, building the view list is a little more complicated than it used to be and involves building a tree of views on the fly whenever subsurfaces are used. However, this means that backends can remain completely subsurface-agnostic. 4. Surfaces and views both keep track of which outputs they are on. 5. The weston_surface structure now has width and height fields. These are populated when a new buffer is attached before surface.configure is called. This is because there are many surface-based operations that really require the width and height and digging through the views didn't work well. Signed-off-by: Jason Ekstrand <jason@jlekstrand.net>
2013-10-13 07:38:11 +04:00
if (view->transform.enabled) {
struct weston_vector v = { { x, y, 0.0f, 1.0f } };
Split the geometry information from weston_surface out into weston_view The weston_surface structure is split into two structures: * The weston_surface structure storres everything required for a client-side or server-side surface. This includes buffers; callbacks; backend private data; input, damage, and opaque regions; and a few other bookkeeping bits. * The weston_view structure represents an entity in the scenegraph and storres all of the geometry information. This includes clip region, alpha, position, and the transformation list as well as all of the temporary information derived from the geometry state. Because a view, and not a surface, is a scenegraph element, the view is what is placed in layers and planes. There are a few things worth noting about the surface/view split: 1. This is *not* a modification to the protocol. It is, instead, a modification to Weston's internal scenegraph to allow a single surface to exist in multiple places at a time. Clients are completely unaware of how many views to a particular surface exist. 2. A view is considered a direct child of a surface and is destroyed when the surface is destroyed. Because of this, the view.surface pointer is always valid and non-null. 3. The compositor's surface_list is replaced with a view_list. Due to subsurfaces, building the view list is a little more complicated than it used to be and involves building a tree of views on the fly whenever subsurfaces are used. However, this means that backends can remain completely subsurface-agnostic. 4. Surfaces and views both keep track of which outputs they are on. 5. The weston_surface structure now has width and height fields. These are populated when a new buffer is attached before surface.configure is called. This is because there are many surface-based operations that really require the width and height and digging through the views didn't work well. Signed-off-by: Jason Ekstrand <jason@jlekstrand.net>
2013-10-13 07:38:11 +04:00
weston_matrix_transform(&view->transform.inverse, &v);
if (fabsf(v.f[3]) < 1e-6) {
2012-06-07 20:01:59 +04:00
weston_log("warning: numerical instability in "
Split the geometry information from weston_surface out into weston_view The weston_surface structure is split into two structures: * The weston_surface structure storres everything required for a client-side or server-side surface. This includes buffers; callbacks; backend private data; input, damage, and opaque regions; and a few other bookkeeping bits. * The weston_view structure represents an entity in the scenegraph and storres all of the geometry information. This includes clip region, alpha, position, and the transformation list as well as all of the temporary information derived from the geometry state. Because a view, and not a surface, is a scenegraph element, the view is what is placed in layers and planes. There are a few things worth noting about the surface/view split: 1. This is *not* a modification to the protocol. It is, instead, a modification to Weston's internal scenegraph to allow a single surface to exist in multiple places at a time. Clients are completely unaware of how many views to a particular surface exist. 2. A view is considered a direct child of a surface and is destroyed when the surface is destroyed. Because of this, the view.surface pointer is always valid and non-null. 3. The compositor's surface_list is replaced with a view_list. Due to subsurfaces, building the view list is a little more complicated than it used to be and involves building a tree of views on the fly whenever subsurfaces are used. However, this means that backends can remain completely subsurface-agnostic. 4. Surfaces and views both keep track of which outputs they are on. 5. The weston_surface structure now has width and height fields. These are populated when a new buffer is attached before surface.configure is called. This is because there are many surface-based operations that really require the width and height and digging through the views didn't work well. Signed-off-by: Jason Ekstrand <jason@jlekstrand.net>
2013-10-13 07:38:11 +04:00
"weston_view_from_global(), divisor = %g\n",
v.f[3]);
Split the geometry information from weston_surface out into weston_view The weston_surface structure is split into two structures: * The weston_surface structure storres everything required for a client-side or server-side surface. This includes buffers; callbacks; backend private data; input, damage, and opaque regions; and a few other bookkeeping bits. * The weston_view structure represents an entity in the scenegraph and storres all of the geometry information. This includes clip region, alpha, position, and the transformation list as well as all of the temporary information derived from the geometry state. Because a view, and not a surface, is a scenegraph element, the view is what is placed in layers and planes. There are a few things worth noting about the surface/view split: 1. This is *not* a modification to the protocol. It is, instead, a modification to Weston's internal scenegraph to allow a single surface to exist in multiple places at a time. Clients are completely unaware of how many views to a particular surface exist. 2. A view is considered a direct child of a surface and is destroyed when the surface is destroyed. Because of this, the view.surface pointer is always valid and non-null. 3. The compositor's surface_list is replaced with a view_list. Due to subsurfaces, building the view list is a little more complicated than it used to be and involves building a tree of views on the fly whenever subsurfaces are used. However, this means that backends can remain completely subsurface-agnostic. 4. Surfaces and views both keep track of which outputs they are on. 5. The weston_surface structure now has width and height fields. These are populated when a new buffer is attached before surface.configure is called. This is because there are many surface-based operations that really require the width and height and digging through the views didn't work well. Signed-off-by: Jason Ekstrand <jason@jlekstrand.net>
2013-10-13 07:38:11 +04:00
*vx = 0;
*vy = 0;
return;
}
Split the geometry information from weston_surface out into weston_view The weston_surface structure is split into two structures: * The weston_surface structure storres everything required for a client-side or server-side surface. This includes buffers; callbacks; backend private data; input, damage, and opaque regions; and a few other bookkeeping bits. * The weston_view structure represents an entity in the scenegraph and storres all of the geometry information. This includes clip region, alpha, position, and the transformation list as well as all of the temporary information derived from the geometry state. Because a view, and not a surface, is a scenegraph element, the view is what is placed in layers and planes. There are a few things worth noting about the surface/view split: 1. This is *not* a modification to the protocol. It is, instead, a modification to Weston's internal scenegraph to allow a single surface to exist in multiple places at a time. Clients are completely unaware of how many views to a particular surface exist. 2. A view is considered a direct child of a surface and is destroyed when the surface is destroyed. Because of this, the view.surface pointer is always valid and non-null. 3. The compositor's surface_list is replaced with a view_list. Due to subsurfaces, building the view list is a little more complicated than it used to be and involves building a tree of views on the fly whenever subsurfaces are used. However, this means that backends can remain completely subsurface-agnostic. 4. Surfaces and views both keep track of which outputs they are on. 5. The weston_surface structure now has width and height fields. These are populated when a new buffer is attached before surface.configure is called. This is because there are many surface-based operations that really require the width and height and digging through the views didn't work well. Signed-off-by: Jason Ekstrand <jason@jlekstrand.net>
2013-10-13 07:38:11 +04:00
*vx = v.f[0] / v.f[3];
*vy = v.f[1] / v.f[3];
} else {
Split the geometry information from weston_surface out into weston_view The weston_surface structure is split into two structures: * The weston_surface structure storres everything required for a client-side or server-side surface. This includes buffers; callbacks; backend private data; input, damage, and opaque regions; and a few other bookkeeping bits. * The weston_view structure represents an entity in the scenegraph and storres all of the geometry information. This includes clip region, alpha, position, and the transformation list as well as all of the temporary information derived from the geometry state. Because a view, and not a surface, is a scenegraph element, the view is what is placed in layers and planes. There are a few things worth noting about the surface/view split: 1. This is *not* a modification to the protocol. It is, instead, a modification to Weston's internal scenegraph to allow a single surface to exist in multiple places at a time. Clients are completely unaware of how many views to a particular surface exist. 2. A view is considered a direct child of a surface and is destroyed when the surface is destroyed. Because of this, the view.surface pointer is always valid and non-null. 3. The compositor's surface_list is replaced with a view_list. Due to subsurfaces, building the view list is a little more complicated than it used to be and involves building a tree of views on the fly whenever subsurfaces are used. However, this means that backends can remain completely subsurface-agnostic. 4. Surfaces and views both keep track of which outputs they are on. 5. The weston_surface structure now has width and height fields. These are populated when a new buffer is attached before surface.configure is called. This is because there are many surface-based operations that really require the width and height and digging through the views didn't work well. Signed-off-by: Jason Ekstrand <jason@jlekstrand.net>
2013-10-13 07:38:11 +04:00
*vx = x - view->geometry.x;
*vy = y - view->geometry.y;
}
}
WL_EXPORT void
Split the geometry information from weston_surface out into weston_view The weston_surface structure is split into two structures: * The weston_surface structure storres everything required for a client-side or server-side surface. This includes buffers; callbacks; backend private data; input, damage, and opaque regions; and a few other bookkeeping bits. * The weston_view structure represents an entity in the scenegraph and storres all of the geometry information. This includes clip region, alpha, position, and the transformation list as well as all of the temporary information derived from the geometry state. Because a view, and not a surface, is a scenegraph element, the view is what is placed in layers and planes. There are a few things worth noting about the surface/view split: 1. This is *not* a modification to the protocol. It is, instead, a modification to Weston's internal scenegraph to allow a single surface to exist in multiple places at a time. Clients are completely unaware of how many views to a particular surface exist. 2. A view is considered a direct child of a surface and is destroyed when the surface is destroyed. Because of this, the view.surface pointer is always valid and non-null. 3. The compositor's surface_list is replaced with a view_list. Due to subsurfaces, building the view list is a little more complicated than it used to be and involves building a tree of views on the fly whenever subsurfaces are used. However, this means that backends can remain completely subsurface-agnostic. 4. Surfaces and views both keep track of which outputs they are on. 5. The weston_surface structure now has width and height fields. These are populated when a new buffer is attached before surface.configure is called. This is because there are many surface-based operations that really require the width and height and digging through the views didn't work well. Signed-off-by: Jason Ekstrand <jason@jlekstrand.net>
2013-10-13 07:38:11 +04:00
weston_view_from_global_fixed(struct weston_view *view,
wl_fixed_t x, wl_fixed_t y,
wl_fixed_t *vx, wl_fixed_t *vy)
{
Split the geometry information from weston_surface out into weston_view The weston_surface structure is split into two structures: * The weston_surface structure storres everything required for a client-side or server-side surface. This includes buffers; callbacks; backend private data; input, damage, and opaque regions; and a few other bookkeeping bits. * The weston_view structure represents an entity in the scenegraph and storres all of the geometry information. This includes clip region, alpha, position, and the transformation list as well as all of the temporary information derived from the geometry state. Because a view, and not a surface, is a scenegraph element, the view is what is placed in layers and planes. There are a few things worth noting about the surface/view split: 1. This is *not* a modification to the protocol. It is, instead, a modification to Weston's internal scenegraph to allow a single surface to exist in multiple places at a time. Clients are completely unaware of how many views to a particular surface exist. 2. A view is considered a direct child of a surface and is destroyed when the surface is destroyed. Because of this, the view.surface pointer is always valid and non-null. 3. The compositor's surface_list is replaced with a view_list. Due to subsurfaces, building the view list is a little more complicated than it used to be and involves building a tree of views on the fly whenever subsurfaces are used. However, this means that backends can remain completely subsurface-agnostic. 4. Surfaces and views both keep track of which outputs they are on. 5. The weston_surface structure now has width and height fields. These are populated when a new buffer is attached before surface.configure is called. This is because there are many surface-based operations that really require the width and height and digging through the views didn't work well. Signed-off-by: Jason Ekstrand <jason@jlekstrand.net>
2013-10-13 07:38:11 +04:00
float vxf, vyf;
Split the geometry information from weston_surface out into weston_view The weston_surface structure is split into two structures: * The weston_surface structure storres everything required for a client-side or server-side surface. This includes buffers; callbacks; backend private data; input, damage, and opaque regions; and a few other bookkeeping bits. * The weston_view structure represents an entity in the scenegraph and storres all of the geometry information. This includes clip region, alpha, position, and the transformation list as well as all of the temporary information derived from the geometry state. Because a view, and not a surface, is a scenegraph element, the view is what is placed in layers and planes. There are a few things worth noting about the surface/view split: 1. This is *not* a modification to the protocol. It is, instead, a modification to Weston's internal scenegraph to allow a single surface to exist in multiple places at a time. Clients are completely unaware of how many views to a particular surface exist. 2. A view is considered a direct child of a surface and is destroyed when the surface is destroyed. Because of this, the view.surface pointer is always valid and non-null. 3. The compositor's surface_list is replaced with a view_list. Due to subsurfaces, building the view list is a little more complicated than it used to be and involves building a tree of views on the fly whenever subsurfaces are used. However, this means that backends can remain completely subsurface-agnostic. 4. Surfaces and views both keep track of which outputs they are on. 5. The weston_surface structure now has width and height fields. These are populated when a new buffer is attached before surface.configure is called. This is because there are many surface-based operations that really require the width and height and digging through the views didn't work well. Signed-off-by: Jason Ekstrand <jason@jlekstrand.net>
2013-10-13 07:38:11 +04:00
weston_view_from_global_float(view,
wl_fixed_to_double(x),
wl_fixed_to_double(y),
&vxf, &vyf);
*vx = wl_fixed_from_double(vxf);
*vy = wl_fixed_from_double(vyf);
}
WL_EXPORT void
Split the geometry information from weston_surface out into weston_view The weston_surface structure is split into two structures: * The weston_surface structure storres everything required for a client-side or server-side surface. This includes buffers; callbacks; backend private data; input, damage, and opaque regions; and a few other bookkeeping bits. * The weston_view structure represents an entity in the scenegraph and storres all of the geometry information. This includes clip region, alpha, position, and the transformation list as well as all of the temporary information derived from the geometry state. Because a view, and not a surface, is a scenegraph element, the view is what is placed in layers and planes. There are a few things worth noting about the surface/view split: 1. This is *not* a modification to the protocol. It is, instead, a modification to Weston's internal scenegraph to allow a single surface to exist in multiple places at a time. Clients are completely unaware of how many views to a particular surface exist. 2. A view is considered a direct child of a surface and is destroyed when the surface is destroyed. Because of this, the view.surface pointer is always valid and non-null. 3. The compositor's surface_list is replaced with a view_list. Due to subsurfaces, building the view list is a little more complicated than it used to be and involves building a tree of views on the fly whenever subsurfaces are used. However, this means that backends can remain completely subsurface-agnostic. 4. Surfaces and views both keep track of which outputs they are on. 5. The weston_surface structure now has width and height fields. These are populated when a new buffer is attached before surface.configure is called. This is because there are many surface-based operations that really require the width and height and digging through the views didn't work well. Signed-off-by: Jason Ekstrand <jason@jlekstrand.net>
2013-10-13 07:38:11 +04:00
weston_view_from_global(struct weston_view *view,
int32_t x, int32_t y, int32_t *vx, int32_t *vy)
{
Split the geometry information from weston_surface out into weston_view The weston_surface structure is split into two structures: * The weston_surface structure storres everything required for a client-side or server-side surface. This includes buffers; callbacks; backend private data; input, damage, and opaque regions; and a few other bookkeeping bits. * The weston_view structure represents an entity in the scenegraph and storres all of the geometry information. This includes clip region, alpha, position, and the transformation list as well as all of the temporary information derived from the geometry state. Because a view, and not a surface, is a scenegraph element, the view is what is placed in layers and planes. There are a few things worth noting about the surface/view split: 1. This is *not* a modification to the protocol. It is, instead, a modification to Weston's internal scenegraph to allow a single surface to exist in multiple places at a time. Clients are completely unaware of how many views to a particular surface exist. 2. A view is considered a direct child of a surface and is destroyed when the surface is destroyed. Because of this, the view.surface pointer is always valid and non-null. 3. The compositor's surface_list is replaced with a view_list. Due to subsurfaces, building the view list is a little more complicated than it used to be and involves building a tree of views on the fly whenever subsurfaces are used. However, this means that backends can remain completely subsurface-agnostic. 4. Surfaces and views both keep track of which outputs they are on. 5. The weston_surface structure now has width and height fields. These are populated when a new buffer is attached before surface.configure is called. This is because there are many surface-based operations that really require the width and height and digging through the views didn't work well. Signed-off-by: Jason Ekstrand <jason@jlekstrand.net>
2013-10-13 07:38:11 +04:00
float vxf, vyf;
Split the geometry information from weston_surface out into weston_view The weston_surface structure is split into two structures: * The weston_surface structure storres everything required for a client-side or server-side surface. This includes buffers; callbacks; backend private data; input, damage, and opaque regions; and a few other bookkeeping bits. * The weston_view structure represents an entity in the scenegraph and storres all of the geometry information. This includes clip region, alpha, position, and the transformation list as well as all of the temporary information derived from the geometry state. Because a view, and not a surface, is a scenegraph element, the view is what is placed in layers and planes. There are a few things worth noting about the surface/view split: 1. This is *not* a modification to the protocol. It is, instead, a modification to Weston's internal scenegraph to allow a single surface to exist in multiple places at a time. Clients are completely unaware of how many views to a particular surface exist. 2. A view is considered a direct child of a surface and is destroyed when the surface is destroyed. Because of this, the view.surface pointer is always valid and non-null. 3. The compositor's surface_list is replaced with a view_list. Due to subsurfaces, building the view list is a little more complicated than it used to be and involves building a tree of views on the fly whenever subsurfaces are used. However, this means that backends can remain completely subsurface-agnostic. 4. Surfaces and views both keep track of which outputs they are on. 5. The weston_surface structure now has width and height fields. These are populated when a new buffer is attached before surface.configure is called. This is because there are many surface-based operations that really require the width and height and digging through the views didn't work well. Signed-off-by: Jason Ekstrand <jason@jlekstrand.net>
2013-10-13 07:38:11 +04:00
weston_view_from_global_float(view, x, y, &vxf, &vyf);
*vx = floorf(vxf);
*vy = floorf(vyf);
}
/**
* \param surface The surface to be repainted
*
* Marks the output(s) that the surface is shown on as needing to be
* repainted. See weston_output_schedule_repaint().
*/
WL_EXPORT void
weston_surface_schedule_repaint(struct weston_surface *surface)
{
struct weston_output *output;
wl_list_for_each(output, &surface->compositor->output_list, link)
if (surface->output_mask & (1u << output->id))
weston_output_schedule_repaint(output);
}
/**
* \param view The view to be repainted
*
* Marks the output(s) that the view is shown on as needing to be
* repainted. See weston_output_schedule_repaint().
*/
Split the geometry information from weston_surface out into weston_view The weston_surface structure is split into two structures: * The weston_surface structure storres everything required for a client-side or server-side surface. This includes buffers; callbacks; backend private data; input, damage, and opaque regions; and a few other bookkeeping bits. * The weston_view structure represents an entity in the scenegraph and storres all of the geometry information. This includes clip region, alpha, position, and the transformation list as well as all of the temporary information derived from the geometry state. Because a view, and not a surface, is a scenegraph element, the view is what is placed in layers and planes. There are a few things worth noting about the surface/view split: 1. This is *not* a modification to the protocol. It is, instead, a modification to Weston's internal scenegraph to allow a single surface to exist in multiple places at a time. Clients are completely unaware of how many views to a particular surface exist. 2. A view is considered a direct child of a surface and is destroyed when the surface is destroyed. Because of this, the view.surface pointer is always valid and non-null. 3. The compositor's surface_list is replaced with a view_list. Due to subsurfaces, building the view list is a little more complicated than it used to be and involves building a tree of views on the fly whenever subsurfaces are used. However, this means that backends can remain completely subsurface-agnostic. 4. Surfaces and views both keep track of which outputs they are on. 5. The weston_surface structure now has width and height fields. These are populated when a new buffer is attached before surface.configure is called. This is because there are many surface-based operations that really require the width and height and digging through the views didn't work well. Signed-off-by: Jason Ekstrand <jason@jlekstrand.net>
2013-10-13 07:38:11 +04:00
WL_EXPORT void
weston_view_schedule_repaint(struct weston_view *view)
{
struct weston_output *output;
wl_list_for_each(output, &view->surface->compositor->output_list, link)
if (view->output_mask & (1u << output->id))
Split the geometry information from weston_surface out into weston_view The weston_surface structure is split into two structures: * The weston_surface structure storres everything required for a client-side or server-side surface. This includes buffers; callbacks; backend private data; input, damage, and opaque regions; and a few other bookkeeping bits. * The weston_view structure represents an entity in the scenegraph and storres all of the geometry information. This includes clip region, alpha, position, and the transformation list as well as all of the temporary information derived from the geometry state. Because a view, and not a surface, is a scenegraph element, the view is what is placed in layers and planes. There are a few things worth noting about the surface/view split: 1. This is *not* a modification to the protocol. It is, instead, a modification to Weston's internal scenegraph to allow a single surface to exist in multiple places at a time. Clients are completely unaware of how many views to a particular surface exist. 2. A view is considered a direct child of a surface and is destroyed when the surface is destroyed. Because of this, the view.surface pointer is always valid and non-null. 3. The compositor's surface_list is replaced with a view_list. Due to subsurfaces, building the view list is a little more complicated than it used to be and involves building a tree of views on the fly whenever subsurfaces are used. However, this means that backends can remain completely subsurface-agnostic. 4. Surfaces and views both keep track of which outputs they are on. 5. The weston_surface structure now has width and height fields. These are populated when a new buffer is attached before surface.configure is called. This is because there are many surface-based operations that really require the width and height and digging through the views didn't work well. Signed-off-by: Jason Ekstrand <jason@jlekstrand.net>
2013-10-13 07:38:11 +04:00
weston_output_schedule_repaint(output);
}
/**
* XXX: This function does it the wrong way.
* surface->damage is the damage from the client, and causes
* surface_flush_damage() to copy pixels. No window management action can
* cause damage to the client-provided content, warranting re-upload!
*
* Instead of surface->damage, this function should record the damage
* with all the views for this surface to avoid extraneous texture
* uploads.
*/
WL_EXPORT void
weston_surface_damage(struct weston_surface *surface)
{
pixman_region32_union_rect(&surface->damage, &surface->damage,
Split the geometry information from weston_surface out into weston_view The weston_surface structure is split into two structures: * The weston_surface structure storres everything required for a client-side or server-side surface. This includes buffers; callbacks; backend private data; input, damage, and opaque regions; and a few other bookkeeping bits. * The weston_view structure represents an entity in the scenegraph and storres all of the geometry information. This includes clip region, alpha, position, and the transformation list as well as all of the temporary information derived from the geometry state. Because a view, and not a surface, is a scenegraph element, the view is what is placed in layers and planes. There are a few things worth noting about the surface/view split: 1. This is *not* a modification to the protocol. It is, instead, a modification to Weston's internal scenegraph to allow a single surface to exist in multiple places at a time. Clients are completely unaware of how many views to a particular surface exist. 2. A view is considered a direct child of a surface and is destroyed when the surface is destroyed. Because of this, the view.surface pointer is always valid and non-null. 3. The compositor's surface_list is replaced with a view_list. Due to subsurfaces, building the view list is a little more complicated than it used to be and involves building a tree of views on the fly whenever subsurfaces are used. However, this means that backends can remain completely subsurface-agnostic. 4. Surfaces and views both keep track of which outputs they are on. 5. The weston_surface structure now has width and height fields. These are populated when a new buffer is attached before surface.configure is called. This is because there are many surface-based operations that really require the width and height and digging through the views didn't work well. Signed-off-by: Jason Ekstrand <jason@jlekstrand.net>
2013-10-13 07:38:11 +04:00
0, 0, surface->width,
surface->height);
weston_surface_schedule_repaint(surface);
}
WL_EXPORT void
Split the geometry information from weston_surface out into weston_view The weston_surface structure is split into two structures: * The weston_surface structure storres everything required for a client-side or server-side surface. This includes buffers; callbacks; backend private data; input, damage, and opaque regions; and a few other bookkeeping bits. * The weston_view structure represents an entity in the scenegraph and storres all of the geometry information. This includes clip region, alpha, position, and the transformation list as well as all of the temporary information derived from the geometry state. Because a view, and not a surface, is a scenegraph element, the view is what is placed in layers and planes. There are a few things worth noting about the surface/view split: 1. This is *not* a modification to the protocol. It is, instead, a modification to Weston's internal scenegraph to allow a single surface to exist in multiple places at a time. Clients are completely unaware of how many views to a particular surface exist. 2. A view is considered a direct child of a surface and is destroyed when the surface is destroyed. Because of this, the view.surface pointer is always valid and non-null. 3. The compositor's surface_list is replaced with a view_list. Due to subsurfaces, building the view list is a little more complicated than it used to be and involves building a tree of views on the fly whenever subsurfaces are used. However, this means that backends can remain completely subsurface-agnostic. 4. Surfaces and views both keep track of which outputs they are on. 5. The weston_surface structure now has width and height fields. These are populated when a new buffer is attached before surface.configure is called. This is because there are many surface-based operations that really require the width and height and digging through the views didn't work well. Signed-off-by: Jason Ekstrand <jason@jlekstrand.net>
2013-10-13 07:38:11 +04:00
weston_view_set_position(struct weston_view *view, float x, float y)
{
if (view->geometry.x == x && view->geometry.y == y)
return;
Split the geometry information from weston_surface out into weston_view The weston_surface structure is split into two structures: * The weston_surface structure storres everything required for a client-side or server-side surface. This includes buffers; callbacks; backend private data; input, damage, and opaque regions; and a few other bookkeeping bits. * The weston_view structure represents an entity in the scenegraph and storres all of the geometry information. This includes clip region, alpha, position, and the transformation list as well as all of the temporary information derived from the geometry state. Because a view, and not a surface, is a scenegraph element, the view is what is placed in layers and planes. There are a few things worth noting about the surface/view split: 1. This is *not* a modification to the protocol. It is, instead, a modification to Weston's internal scenegraph to allow a single surface to exist in multiple places at a time. Clients are completely unaware of how many views to a particular surface exist. 2. A view is considered a direct child of a surface and is destroyed when the surface is destroyed. Because of this, the view.surface pointer is always valid and non-null. 3. The compositor's surface_list is replaced with a view_list. Due to subsurfaces, building the view list is a little more complicated than it used to be and involves building a tree of views on the fly whenever subsurfaces are used. However, this means that backends can remain completely subsurface-agnostic. 4. Surfaces and views both keep track of which outputs they are on. 5. The weston_surface structure now has width and height fields. These are populated when a new buffer is attached before surface.configure is called. This is because there are many surface-based operations that really require the width and height and digging through the views didn't work well. Signed-off-by: Jason Ekstrand <jason@jlekstrand.net>
2013-10-13 07:38:11 +04:00
view->geometry.x = x;
view->geometry.y = y;
weston_view_geometry_dirty(view);
}
static void
transform_parent_handle_parent_destroy(struct wl_listener *listener,
void *data)
{
Split the geometry information from weston_surface out into weston_view The weston_surface structure is split into two structures: * The weston_surface structure storres everything required for a client-side or server-side surface. This includes buffers; callbacks; backend private data; input, damage, and opaque regions; and a few other bookkeeping bits. * The weston_view structure represents an entity in the scenegraph and storres all of the geometry information. This includes clip region, alpha, position, and the transformation list as well as all of the temporary information derived from the geometry state. Because a view, and not a surface, is a scenegraph element, the view is what is placed in layers and planes. There are a few things worth noting about the surface/view split: 1. This is *not* a modification to the protocol. It is, instead, a modification to Weston's internal scenegraph to allow a single surface to exist in multiple places at a time. Clients are completely unaware of how many views to a particular surface exist. 2. A view is considered a direct child of a surface and is destroyed when the surface is destroyed. Because of this, the view.surface pointer is always valid and non-null. 3. The compositor's surface_list is replaced with a view_list. Due to subsurfaces, building the view list is a little more complicated than it used to be and involves building a tree of views on the fly whenever subsurfaces are used. However, this means that backends can remain completely subsurface-agnostic. 4. Surfaces and views both keep track of which outputs they are on. 5. The weston_surface structure now has width and height fields. These are populated when a new buffer is attached before surface.configure is called. This is because there are many surface-based operations that really require the width and height and digging through the views didn't work well. Signed-off-by: Jason Ekstrand <jason@jlekstrand.net>
2013-10-13 07:38:11 +04:00
struct weston_view *view =
container_of(listener, struct weston_view,
geometry.parent_destroy_listener);
Split the geometry information from weston_surface out into weston_view The weston_surface structure is split into two structures: * The weston_surface structure storres everything required for a client-side or server-side surface. This includes buffers; callbacks; backend private data; input, damage, and opaque regions; and a few other bookkeeping bits. * The weston_view structure represents an entity in the scenegraph and storres all of the geometry information. This includes clip region, alpha, position, and the transformation list as well as all of the temporary information derived from the geometry state. Because a view, and not a surface, is a scenegraph element, the view is what is placed in layers and planes. There are a few things worth noting about the surface/view split: 1. This is *not* a modification to the protocol. It is, instead, a modification to Weston's internal scenegraph to allow a single surface to exist in multiple places at a time. Clients are completely unaware of how many views to a particular surface exist. 2. A view is considered a direct child of a surface and is destroyed when the surface is destroyed. Because of this, the view.surface pointer is always valid and non-null. 3. The compositor's surface_list is replaced with a view_list. Due to subsurfaces, building the view list is a little more complicated than it used to be and involves building a tree of views on the fly whenever subsurfaces are used. However, this means that backends can remain completely subsurface-agnostic. 4. Surfaces and views both keep track of which outputs they are on. 5. The weston_surface structure now has width and height fields. These are populated when a new buffer is attached before surface.configure is called. This is because there are many surface-based operations that really require the width and height and digging through the views didn't work well. Signed-off-by: Jason Ekstrand <jason@jlekstrand.net>
2013-10-13 07:38:11 +04:00
weston_view_set_transform_parent(view, NULL);
}
WL_EXPORT void
Split the geometry information from weston_surface out into weston_view The weston_surface structure is split into two structures: * The weston_surface structure storres everything required for a client-side or server-side surface. This includes buffers; callbacks; backend private data; input, damage, and opaque regions; and a few other bookkeeping bits. * The weston_view structure represents an entity in the scenegraph and storres all of the geometry information. This includes clip region, alpha, position, and the transformation list as well as all of the temporary information derived from the geometry state. Because a view, and not a surface, is a scenegraph element, the view is what is placed in layers and planes. There are a few things worth noting about the surface/view split: 1. This is *not* a modification to the protocol. It is, instead, a modification to Weston's internal scenegraph to allow a single surface to exist in multiple places at a time. Clients are completely unaware of how many views to a particular surface exist. 2. A view is considered a direct child of a surface and is destroyed when the surface is destroyed. Because of this, the view.surface pointer is always valid and non-null. 3. The compositor's surface_list is replaced with a view_list. Due to subsurfaces, building the view list is a little more complicated than it used to be and involves building a tree of views on the fly whenever subsurfaces are used. However, this means that backends can remain completely subsurface-agnostic. 4. Surfaces and views both keep track of which outputs they are on. 5. The weston_surface structure now has width and height fields. These are populated when a new buffer is attached before surface.configure is called. This is because there are many surface-based operations that really require the width and height and digging through the views didn't work well. Signed-off-by: Jason Ekstrand <jason@jlekstrand.net>
2013-10-13 07:38:11 +04:00
weston_view_set_transform_parent(struct weston_view *view,
compositor: add weston_view_set_mask() API and state Add API for setting a clip ('scissor' in the code) rectangle per view, in surface coordinates. Ivi-shell requires this feature to be able to implement the IVI Layer Manager API, which includes clipping of surfaces. The names weston_view_set_mask() and weston_view_set_mask_infinite() mirror the existing weston_layer_set_mask*() functions. This view clipping complements the weston_layer clipping, because view clipping is defined in surface local coordinates, while layer mask/clipping is defined in global coordinates. View clipping requires explicit support from the renderers. Therefore a new Weston capability bit is added: WESTON_CAP_VIEW_CLIP_MASK. Shells (and all users) of this new API are required to check the capability bit is set before using the API. Otherwise the rendering will not be what they expect. View clips are inherited through the transformation inheritance mechanism. However, there are restrictions. The clip rectangle can be set only on the root view of a transformation inheritance tree. The additional transformations in child views must not rotate the coordinate axes. These restrictions avoid corner cases in clip inheritance, and keep the renderer implementations as simple as they are right now. Renderers only need to do an additional intersection with the clip rectangle which is always aligned to the surface coordinate system. For more details, see the API documentation in the patch. Signed-off-by: Pekka Paalanen <pekka.paalanen@collabora.co.uk> Reviewed-by: Nobuhiko Tanibata <NOBUHIKO_TANIBATA@xddp.denso.co.jp> Tested-by: Nobuhiko Tanibata <NOBUHIKO_TANIBATA@xddp.denso.co.jp> Reviewed-By: Derek Foreman <derekf@osg.samsung.com>
2015-02-16 15:39:11 +03:00
struct weston_view *parent)
{
Split the geometry information from weston_surface out into weston_view The weston_surface structure is split into two structures: * The weston_surface structure storres everything required for a client-side or server-side surface. This includes buffers; callbacks; backend private data; input, damage, and opaque regions; and a few other bookkeeping bits. * The weston_view structure represents an entity in the scenegraph and storres all of the geometry information. This includes clip region, alpha, position, and the transformation list as well as all of the temporary information derived from the geometry state. Because a view, and not a surface, is a scenegraph element, the view is what is placed in layers and planes. There are a few things worth noting about the surface/view split: 1. This is *not* a modification to the protocol. It is, instead, a modification to Weston's internal scenegraph to allow a single surface to exist in multiple places at a time. Clients are completely unaware of how many views to a particular surface exist. 2. A view is considered a direct child of a surface and is destroyed when the surface is destroyed. Because of this, the view.surface pointer is always valid and non-null. 3. The compositor's surface_list is replaced with a view_list. Due to subsurfaces, building the view list is a little more complicated than it used to be and involves building a tree of views on the fly whenever subsurfaces are used. However, this means that backends can remain completely subsurface-agnostic. 4. Surfaces and views both keep track of which outputs they are on. 5. The weston_surface structure now has width and height fields. These are populated when a new buffer is attached before surface.configure is called. This is because there are many surface-based operations that really require the width and height and digging through the views didn't work well. Signed-off-by: Jason Ekstrand <jason@jlekstrand.net>
2013-10-13 07:38:11 +04:00
if (view->geometry.parent) {
wl_list_remove(&view->geometry.parent_destroy_listener.link);
wl_list_remove(&view->geometry.parent_link);
compositor: add weston_view_set_mask() API and state Add API for setting a clip ('scissor' in the code) rectangle per view, in surface coordinates. Ivi-shell requires this feature to be able to implement the IVI Layer Manager API, which includes clipping of surfaces. The names weston_view_set_mask() and weston_view_set_mask_infinite() mirror the existing weston_layer_set_mask*() functions. This view clipping complements the weston_layer clipping, because view clipping is defined in surface local coordinates, while layer mask/clipping is defined in global coordinates. View clipping requires explicit support from the renderers. Therefore a new Weston capability bit is added: WESTON_CAP_VIEW_CLIP_MASK. Shells (and all users) of this new API are required to check the capability bit is set before using the API. Otherwise the rendering will not be what they expect. View clips are inherited through the transformation inheritance mechanism. However, there are restrictions. The clip rectangle can be set only on the root view of a transformation inheritance tree. The additional transformations in child views must not rotate the coordinate axes. These restrictions avoid corner cases in clip inheritance, and keep the renderer implementations as simple as they are right now. Renderers only need to do an additional intersection with the clip rectangle which is always aligned to the surface coordinate system. For more details, see the API documentation in the patch. Signed-off-by: Pekka Paalanen <pekka.paalanen@collabora.co.uk> Reviewed-by: Nobuhiko Tanibata <NOBUHIKO_TANIBATA@xddp.denso.co.jp> Tested-by: Nobuhiko Tanibata <NOBUHIKO_TANIBATA@xddp.denso.co.jp> Reviewed-By: Derek Foreman <derekf@osg.samsung.com>
2015-02-16 15:39:11 +03:00
if (!parent)
view->geometry.scissor_enabled = false;
}
Split the geometry information from weston_surface out into weston_view The weston_surface structure is split into two structures: * The weston_surface structure storres everything required for a client-side or server-side surface. This includes buffers; callbacks; backend private data; input, damage, and opaque regions; and a few other bookkeeping bits. * The weston_view structure represents an entity in the scenegraph and storres all of the geometry information. This includes clip region, alpha, position, and the transformation list as well as all of the temporary information derived from the geometry state. Because a view, and not a surface, is a scenegraph element, the view is what is placed in layers and planes. There are a few things worth noting about the surface/view split: 1. This is *not* a modification to the protocol. It is, instead, a modification to Weston's internal scenegraph to allow a single surface to exist in multiple places at a time. Clients are completely unaware of how many views to a particular surface exist. 2. A view is considered a direct child of a surface and is destroyed when the surface is destroyed. Because of this, the view.surface pointer is always valid and non-null. 3. The compositor's surface_list is replaced with a view_list. Due to subsurfaces, building the view list is a little more complicated than it used to be and involves building a tree of views on the fly whenever subsurfaces are used. However, this means that backends can remain completely subsurface-agnostic. 4. Surfaces and views both keep track of which outputs they are on. 5. The weston_surface structure now has width and height fields. These are populated when a new buffer is attached before surface.configure is called. This is because there are many surface-based operations that really require the width and height and digging through the views didn't work well. Signed-off-by: Jason Ekstrand <jason@jlekstrand.net>
2013-10-13 07:38:11 +04:00
view->geometry.parent = parent;
Split the geometry information from weston_surface out into weston_view The weston_surface structure is split into two structures: * The weston_surface structure storres everything required for a client-side or server-side surface. This includes buffers; callbacks; backend private data; input, damage, and opaque regions; and a few other bookkeeping bits. * The weston_view structure represents an entity in the scenegraph and storres all of the geometry information. This includes clip region, alpha, position, and the transformation list as well as all of the temporary information derived from the geometry state. Because a view, and not a surface, is a scenegraph element, the view is what is placed in layers and planes. There are a few things worth noting about the surface/view split: 1. This is *not* a modification to the protocol. It is, instead, a modification to Weston's internal scenegraph to allow a single surface to exist in multiple places at a time. Clients are completely unaware of how many views to a particular surface exist. 2. A view is considered a direct child of a surface and is destroyed when the surface is destroyed. Because of this, the view.surface pointer is always valid and non-null. 3. The compositor's surface_list is replaced with a view_list. Due to subsurfaces, building the view list is a little more complicated than it used to be and involves building a tree of views on the fly whenever subsurfaces are used. However, this means that backends can remain completely subsurface-agnostic. 4. Surfaces and views both keep track of which outputs they are on. 5. The weston_surface structure now has width and height fields. These are populated when a new buffer is attached before surface.configure is called. This is because there are many surface-based operations that really require the width and height and digging through the views didn't work well. Signed-off-by: Jason Ekstrand <jason@jlekstrand.net>
2013-10-13 07:38:11 +04:00
view->geometry.parent_destroy_listener.notify =
transform_parent_handle_parent_destroy;
if (parent) {
wl_signal_add(&parent->destroy_signal,
Split the geometry information from weston_surface out into weston_view The weston_surface structure is split into two structures: * The weston_surface structure storres everything required for a client-side or server-side surface. This includes buffers; callbacks; backend private data; input, damage, and opaque regions; and a few other bookkeeping bits. * The weston_view structure represents an entity in the scenegraph and storres all of the geometry information. This includes clip region, alpha, position, and the transformation list as well as all of the temporary information derived from the geometry state. Because a view, and not a surface, is a scenegraph element, the view is what is placed in layers and planes. There are a few things worth noting about the surface/view split: 1. This is *not* a modification to the protocol. It is, instead, a modification to Weston's internal scenegraph to allow a single surface to exist in multiple places at a time. Clients are completely unaware of how many views to a particular surface exist. 2. A view is considered a direct child of a surface and is destroyed when the surface is destroyed. Because of this, the view.surface pointer is always valid and non-null. 3. The compositor's surface_list is replaced with a view_list. Due to subsurfaces, building the view list is a little more complicated than it used to be and involves building a tree of views on the fly whenever subsurfaces are used. However, this means that backends can remain completely subsurface-agnostic. 4. Surfaces and views both keep track of which outputs they are on. 5. The weston_surface structure now has width and height fields. These are populated when a new buffer is attached before surface.configure is called. This is because there are many surface-based operations that really require the width and height and digging through the views didn't work well. Signed-off-by: Jason Ekstrand <jason@jlekstrand.net>
2013-10-13 07:38:11 +04:00
&view->geometry.parent_destroy_listener);
wl_list_insert(&parent->geometry.child_list,
Split the geometry information from weston_surface out into weston_view The weston_surface structure is split into two structures: * The weston_surface structure storres everything required for a client-side or server-side surface. This includes buffers; callbacks; backend private data; input, damage, and opaque regions; and a few other bookkeeping bits. * The weston_view structure represents an entity in the scenegraph and storres all of the geometry information. This includes clip region, alpha, position, and the transformation list as well as all of the temporary information derived from the geometry state. Because a view, and not a surface, is a scenegraph element, the view is what is placed in layers and planes. There are a few things worth noting about the surface/view split: 1. This is *not* a modification to the protocol. It is, instead, a modification to Weston's internal scenegraph to allow a single surface to exist in multiple places at a time. Clients are completely unaware of how many views to a particular surface exist. 2. A view is considered a direct child of a surface and is destroyed when the surface is destroyed. Because of this, the view.surface pointer is always valid and non-null. 3. The compositor's surface_list is replaced with a view_list. Due to subsurfaces, building the view list is a little more complicated than it used to be and involves building a tree of views on the fly whenever subsurfaces are used. However, this means that backends can remain completely subsurface-agnostic. 4. Surfaces and views both keep track of which outputs they are on. 5. The weston_surface structure now has width and height fields. These are populated when a new buffer is attached before surface.configure is called. This is because there are many surface-based operations that really require the width and height and digging through the views didn't work well. Signed-off-by: Jason Ekstrand <jason@jlekstrand.net>
2013-10-13 07:38:11 +04:00
&view->geometry.parent_link);
}
Split the geometry information from weston_surface out into weston_view The weston_surface structure is split into two structures: * The weston_surface structure storres everything required for a client-side or server-side surface. This includes buffers; callbacks; backend private data; input, damage, and opaque regions; and a few other bookkeeping bits. * The weston_view structure represents an entity in the scenegraph and storres all of the geometry information. This includes clip region, alpha, position, and the transformation list as well as all of the temporary information derived from the geometry state. Because a view, and not a surface, is a scenegraph element, the view is what is placed in layers and planes. There are a few things worth noting about the surface/view split: 1. This is *not* a modification to the protocol. It is, instead, a modification to Weston's internal scenegraph to allow a single surface to exist in multiple places at a time. Clients are completely unaware of how many views to a particular surface exist. 2. A view is considered a direct child of a surface and is destroyed when the surface is destroyed. Because of this, the view.surface pointer is always valid and non-null. 3. The compositor's surface_list is replaced with a view_list. Due to subsurfaces, building the view list is a little more complicated than it used to be and involves building a tree of views on the fly whenever subsurfaces are used. However, this means that backends can remain completely subsurface-agnostic. 4. Surfaces and views both keep track of which outputs they are on. 5. The weston_surface structure now has width and height fields. These are populated when a new buffer is attached before surface.configure is called. This is because there are many surface-based operations that really require the width and height and digging through the views didn't work well. Signed-off-by: Jason Ekstrand <jason@jlekstrand.net>
2013-10-13 07:38:11 +04:00
weston_view_geometry_dirty(view);
}
compositor: add weston_view_set_mask() API and state Add API for setting a clip ('scissor' in the code) rectangle per view, in surface coordinates. Ivi-shell requires this feature to be able to implement the IVI Layer Manager API, which includes clipping of surfaces. The names weston_view_set_mask() and weston_view_set_mask_infinite() mirror the existing weston_layer_set_mask*() functions. This view clipping complements the weston_layer clipping, because view clipping is defined in surface local coordinates, while layer mask/clipping is defined in global coordinates. View clipping requires explicit support from the renderers. Therefore a new Weston capability bit is added: WESTON_CAP_VIEW_CLIP_MASK. Shells (and all users) of this new API are required to check the capability bit is set before using the API. Otherwise the rendering will not be what they expect. View clips are inherited through the transformation inheritance mechanism. However, there are restrictions. The clip rectangle can be set only on the root view of a transformation inheritance tree. The additional transformations in child views must not rotate the coordinate axes. These restrictions avoid corner cases in clip inheritance, and keep the renderer implementations as simple as they are right now. Renderers only need to do an additional intersection with the clip rectangle which is always aligned to the surface coordinate system. For more details, see the API documentation in the patch. Signed-off-by: Pekka Paalanen <pekka.paalanen@collabora.co.uk> Reviewed-by: Nobuhiko Tanibata <NOBUHIKO_TANIBATA@xddp.denso.co.jp> Tested-by: Nobuhiko Tanibata <NOBUHIKO_TANIBATA@xddp.denso.co.jp> Reviewed-By: Derek Foreman <derekf@osg.samsung.com>
2015-02-16 15:39:11 +03:00
/** Set a clip mask rectangle on a view
*
* \param view The view to set the clip mask on.
* \param x Top-left corner X coordinate of the clip rectangle.
* \param y Top-left corner Y coordinate of the clip rectangle.
* \param width Width of the clip rectangle, non-negative.
* \param height Height of the clip rectangle, non-negative.
*
* A shell may set a clip mask rectangle on a view. Everything outside
* the rectangle is cut away for input and output purposes: it is
* not drawn and cannot be hit by hit-test based input like pointer
* motion or touch-downs. Everything inside the rectangle will behave
* normally. Clients are unaware of clipping.
*
* The rectangle is set in surface-local coordinates. Setting a clip
compositor: add weston_view_set_mask() API and state Add API for setting a clip ('scissor' in the code) rectangle per view, in surface coordinates. Ivi-shell requires this feature to be able to implement the IVI Layer Manager API, which includes clipping of surfaces. The names weston_view_set_mask() and weston_view_set_mask_infinite() mirror the existing weston_layer_set_mask*() functions. This view clipping complements the weston_layer clipping, because view clipping is defined in surface local coordinates, while layer mask/clipping is defined in global coordinates. View clipping requires explicit support from the renderers. Therefore a new Weston capability bit is added: WESTON_CAP_VIEW_CLIP_MASK. Shells (and all users) of this new API are required to check the capability bit is set before using the API. Otherwise the rendering will not be what they expect. View clips are inherited through the transformation inheritance mechanism. However, there are restrictions. The clip rectangle can be set only on the root view of a transformation inheritance tree. The additional transformations in child views must not rotate the coordinate axes. These restrictions avoid corner cases in clip inheritance, and keep the renderer implementations as simple as they are right now. Renderers only need to do an additional intersection with the clip rectangle which is always aligned to the surface coordinate system. For more details, see the API documentation in the patch. Signed-off-by: Pekka Paalanen <pekka.paalanen@collabora.co.uk> Reviewed-by: Nobuhiko Tanibata <NOBUHIKO_TANIBATA@xddp.denso.co.jp> Tested-by: Nobuhiko Tanibata <NOBUHIKO_TANIBATA@xddp.denso.co.jp> Reviewed-By: Derek Foreman <derekf@osg.samsung.com>
2015-02-16 15:39:11 +03:00
* mask rectangle does not affect the view position, the view is positioned
* as it would be without a clip. The clip also does not change
* weston_surface::width,height.
*
* The clip mask rectangle is part of transformation inheritance
* (weston_view_set_transform_parent()). A clip set in the root of the
* transformation inheritance tree will affect all views in the tree.
* A clip can be set only on the root view. Attempting to set a clip
* on view that has a transformation parent will fail. Assigning a parent
* to a view that has a clip set will cause the clip to be forgotten.
*
* Because the clip mask is an axis-aligned rectangle, it poses restrictions
* on the additional transformations in the child views. These transformations
* may not rotate the coordinate axes, i.e., only translation and scaling
* are allowed. Violating this restriction causes the clipping to malfunction.
* Furthermore, using scaling may cause rounding errors in child clipping.
*
* The clip mask rectangle is not automatically adjusted based on
* wl_surface.attach dx and dy arguments.
*
* A clip mask rectangle can be set only if the compositor capability
* WESTON_CAP_VIEW_CLIP_MASK is present.
*
* This function sets the clip mask rectangle and schedules a repaint for
* the view.
*/
WL_EXPORT void
weston_view_set_mask(struct weston_view *view,
int x, int y, int width, int height)
{
struct weston_compositor *compositor = view->surface->compositor;
if (!(compositor->capabilities & WESTON_CAP_VIEW_CLIP_MASK)) {
weston_log("%s not allowed without capability!\n", __func__);
return;
}
if (view->geometry.parent) {
weston_log("view %p has a parent, clip forbidden!\n", view);
return;
}
if (width < 0 || height < 0) {
weston_log("%s: illegal args %d, %d, %d, %d\n", __func__,
x, y, width, height);
return;
}
pixman_region32_fini(&view->geometry.scissor);
pixman_region32_init_rect(&view->geometry.scissor, x, y, width, height);
view->geometry.scissor_enabled = true;
weston_view_geometry_dirty(view);
weston_view_schedule_repaint(view);
}
/** Remove the clip mask from a view
*
* \param view The view to remove the clip mask from.
*
* Removed the clip mask rectangle and schedules a repaint.
*
* \sa weston_view_set_mask
*/
WL_EXPORT void
weston_view_set_mask_infinite(struct weston_view *view)
{
view->geometry.scissor_enabled = false;
weston_view_geometry_dirty(view);
weston_view_schedule_repaint(view);
}
/* Check if view should be displayed
*
* The indicator is set manually when assigning
* a view to a surface.
*
* This needs reworking. See the thread starting at:
*
* https://lists.freedesktop.org/archives/wayland-devel/2016-June/029656.html
*/
WL_EXPORT bool
Split the geometry information from weston_surface out into weston_view The weston_surface structure is split into two structures: * The weston_surface structure storres everything required for a client-side or server-side surface. This includes buffers; callbacks; backend private data; input, damage, and opaque regions; and a few other bookkeeping bits. * The weston_view structure represents an entity in the scenegraph and storres all of the geometry information. This includes clip region, alpha, position, and the transformation list as well as all of the temporary information derived from the geometry state. Because a view, and not a surface, is a scenegraph element, the view is what is placed in layers and planes. There are a few things worth noting about the surface/view split: 1. This is *not* a modification to the protocol. It is, instead, a modification to Weston's internal scenegraph to allow a single surface to exist in multiple places at a time. Clients are completely unaware of how many views to a particular surface exist. 2. A view is considered a direct child of a surface and is destroyed when the surface is destroyed. Because of this, the view.surface pointer is always valid and non-null. 3. The compositor's surface_list is replaced with a view_list. Due to subsurfaces, building the view list is a little more complicated than it used to be and involves building a tree of views on the fly whenever subsurfaces are used. However, this means that backends can remain completely subsurface-agnostic. 4. Surfaces and views both keep track of which outputs they are on. 5. The weston_surface structure now has width and height fields. These are populated when a new buffer is attached before surface.configure is called. This is because there are many surface-based operations that really require the width and height and digging through the views didn't work well. Signed-off-by: Jason Ekstrand <jason@jlekstrand.net>
2013-10-13 07:38:11 +04:00
weston_view_is_mapped(struct weston_view *view)
{
return view->is_mapped;
}
/* Check if a surface has a view assigned to it
*
* The indicator is set manually when mapping
* a surface and creating a view for it.
*
* This needs to go. See the thread starting at:
*
* https://lists.freedesktop.org/archives/wayland-devel/2016-June/029656.html
*
*/
WL_EXPORT bool
weston_surface_is_mapped(struct weston_surface *surface)
{
return surface->is_mapped;
}
static void
surface_set_size(struct weston_surface *surface, int32_t width, int32_t height)
{
struct weston_view *view;
if (surface->width == width && surface->height == height)
return;
surface->width = width;
surface->height = height;
wl_list_for_each(view, &surface->views, surface_link)
weston_view_geometry_dirty(view);
}
WL_EXPORT void
weston_surface_set_size(struct weston_surface *surface,
int32_t width, int32_t height)
{
assert(!surface->resource);
surface_set_size(surface, width, height);
}
static int
fixed_round_up_to_int(wl_fixed_t f)
{
return wl_fixed_to_int(wl_fixed_from_int(1) - 1 + f);
}
static void
convert_size_by_transform_scale(int32_t *width_out, int32_t *height_out,
int32_t width, int32_t height,
uint32_t transform,
int32_t scale)
{
assert(scale > 0);
switch (transform) {
case WL_OUTPUT_TRANSFORM_NORMAL:
case WL_OUTPUT_TRANSFORM_180:
case WL_OUTPUT_TRANSFORM_FLIPPED:
case WL_OUTPUT_TRANSFORM_FLIPPED_180:
*width_out = width / scale;
*height_out = height / scale;
break;
case WL_OUTPUT_TRANSFORM_90:
case WL_OUTPUT_TRANSFORM_270:
case WL_OUTPUT_TRANSFORM_FLIPPED_90:
case WL_OUTPUT_TRANSFORM_FLIPPED_270:
*width_out = height / scale;
*height_out = width / scale;
break;
default:
assert(0 && "invalid transform");
}
}
static void
weston_surface_calculate_size_from_buffer(struct weston_surface *surface)
{
struct weston_buffer_viewport *vp = &surface->buffer_viewport;
if (!surface->buffer_ref.buffer) {
surface->width_from_buffer = 0;
surface->height_from_buffer = 0;
return;
}
convert_size_by_transform_scale(&surface->width_from_buffer,
&surface->height_from_buffer,
surface->buffer_ref.buffer->width,
surface->buffer_ref.buffer->height,
vp->buffer.transform,
vp->buffer.scale);
}
static void
weston_surface_update_size(struct weston_surface *surface)
{
struct weston_buffer_viewport *vp = &surface->buffer_viewport;
int32_t width, height;
width = surface->width_from_buffer;
height = surface->height_from_buffer;
if (width != 0 && vp->surface.width != -1) {
surface_set_size(surface,
vp->surface.width, vp->surface.height);
return;
}
if (width != 0 && vp->buffer.src_width != wl_fixed_from_int(-1)) {
int32_t w = fixed_round_up_to_int(vp->buffer.src_width);
int32_t h = fixed_round_up_to_int(vp->buffer.src_height);
surface_set_size(surface, w ?: 1, h ?: 1);
return;
}
surface_set_size(surface, width, height);
}
WL_EXPORT uint32_t
weston_compositor_get_time(void)
{
struct timeval tv;
gettimeofday(&tv, NULL);
return tv.tv_sec * 1000 + tv.tv_usec / 1000;
}
Split the geometry information from weston_surface out into weston_view The weston_surface structure is split into two structures: * The weston_surface structure storres everything required for a client-side or server-side surface. This includes buffers; callbacks; backend private data; input, damage, and opaque regions; and a few other bookkeeping bits. * The weston_view structure represents an entity in the scenegraph and storres all of the geometry information. This includes clip region, alpha, position, and the transformation list as well as all of the temporary information derived from the geometry state. Because a view, and not a surface, is a scenegraph element, the view is what is placed in layers and planes. There are a few things worth noting about the surface/view split: 1. This is *not* a modification to the protocol. It is, instead, a modification to Weston's internal scenegraph to allow a single surface to exist in multiple places at a time. Clients are completely unaware of how many views to a particular surface exist. 2. A view is considered a direct child of a surface and is destroyed when the surface is destroyed. Because of this, the view.surface pointer is always valid and non-null. 3. The compositor's surface_list is replaced with a view_list. Due to subsurfaces, building the view list is a little more complicated than it used to be and involves building a tree of views on the fly whenever subsurfaces are used. However, this means that backends can remain completely subsurface-agnostic. 4. Surfaces and views both keep track of which outputs they are on. 5. The weston_surface structure now has width and height fields. These are populated when a new buffer is attached before surface.configure is called. This is because there are many surface-based operations that really require the width and height and digging through the views didn't work well. Signed-off-by: Jason Ekstrand <jason@jlekstrand.net>
2013-10-13 07:38:11 +04:00
WL_EXPORT struct weston_view *
weston_compositor_pick_view(struct weston_compositor *compositor,
wl_fixed_t x, wl_fixed_t y,
wl_fixed_t *vx, wl_fixed_t *vy)
{
Split the geometry information from weston_surface out into weston_view The weston_surface structure is split into two structures: * The weston_surface structure storres everything required for a client-side or server-side surface. This includes buffers; callbacks; backend private data; input, damage, and opaque regions; and a few other bookkeeping bits. * The weston_view structure represents an entity in the scenegraph and storres all of the geometry information. This includes clip region, alpha, position, and the transformation list as well as all of the temporary information derived from the geometry state. Because a view, and not a surface, is a scenegraph element, the view is what is placed in layers and planes. There are a few things worth noting about the surface/view split: 1. This is *not* a modification to the protocol. It is, instead, a modification to Weston's internal scenegraph to allow a single surface to exist in multiple places at a time. Clients are completely unaware of how many views to a particular surface exist. 2. A view is considered a direct child of a surface and is destroyed when the surface is destroyed. Because of this, the view.surface pointer is always valid and non-null. 3. The compositor's surface_list is replaced with a view_list. Due to subsurfaces, building the view list is a little more complicated than it used to be and involves building a tree of views on the fly whenever subsurfaces are used. However, this means that backends can remain completely subsurface-agnostic. 4. Surfaces and views both keep track of which outputs they are on. 5. The weston_surface structure now has width and height fields. These are populated when a new buffer is attached before surface.configure is called. This is because there are many surface-based operations that really require the width and height and digging through the views didn't work well. Signed-off-by: Jason Ekstrand <jason@jlekstrand.net>
2013-10-13 07:38:11 +04:00
struct weston_view *view;
wl_fixed_t view_x, view_y;
int view_ix, view_iy;
int ix = wl_fixed_to_int(x);
int iy = wl_fixed_to_int(y);
Split the geometry information from weston_surface out into weston_view The weston_surface structure is split into two structures: * The weston_surface structure storres everything required for a client-side or server-side surface. This includes buffers; callbacks; backend private data; input, damage, and opaque regions; and a few other bookkeeping bits. * The weston_view structure represents an entity in the scenegraph and storres all of the geometry information. This includes clip region, alpha, position, and the transformation list as well as all of the temporary information derived from the geometry state. Because a view, and not a surface, is a scenegraph element, the view is what is placed in layers and planes. There are a few things worth noting about the surface/view split: 1. This is *not* a modification to the protocol. It is, instead, a modification to Weston's internal scenegraph to allow a single surface to exist in multiple places at a time. Clients are completely unaware of how many views to a particular surface exist. 2. A view is considered a direct child of a surface and is destroyed when the surface is destroyed. Because of this, the view.surface pointer is always valid and non-null. 3. The compositor's surface_list is replaced with a view_list. Due to subsurfaces, building the view list is a little more complicated than it used to be and involves building a tree of views on the fly whenever subsurfaces are used. However, this means that backends can remain completely subsurface-agnostic. 4. Surfaces and views both keep track of which outputs they are on. 5. The weston_surface structure now has width and height fields. These are populated when a new buffer is attached before surface.configure is called. This is because there are many surface-based operations that really require the width and height and digging through the views didn't work well. Signed-off-by: Jason Ekstrand <jason@jlekstrand.net>
2013-10-13 07:38:11 +04:00
wl_list_for_each(view, &compositor->view_list, link) {
if (!pixman_region32_contains_point(
&view->transform.boundingbox, ix, iy, NULL))
continue;
weston_view_from_global_fixed(view, x, y, &view_x, &view_y);
view_ix = wl_fixed_to_int(view_x);
view_iy = wl_fixed_to_int(view_y);
if (!pixman_region32_contains_point(&view->surface->input,
view_ix, view_iy, NULL))
continue;
compositor: add weston_view_set_mask() API and state Add API for setting a clip ('scissor' in the code) rectangle per view, in surface coordinates. Ivi-shell requires this feature to be able to implement the IVI Layer Manager API, which includes clipping of surfaces. The names weston_view_set_mask() and weston_view_set_mask_infinite() mirror the existing weston_layer_set_mask*() functions. This view clipping complements the weston_layer clipping, because view clipping is defined in surface local coordinates, while layer mask/clipping is defined in global coordinates. View clipping requires explicit support from the renderers. Therefore a new Weston capability bit is added: WESTON_CAP_VIEW_CLIP_MASK. Shells (and all users) of this new API are required to check the capability bit is set before using the API. Otherwise the rendering will not be what they expect. View clips are inherited through the transformation inheritance mechanism. However, there are restrictions. The clip rectangle can be set only on the root view of a transformation inheritance tree. The additional transformations in child views must not rotate the coordinate axes. These restrictions avoid corner cases in clip inheritance, and keep the renderer implementations as simple as they are right now. Renderers only need to do an additional intersection with the clip rectangle which is always aligned to the surface coordinate system. For more details, see the API documentation in the patch. Signed-off-by: Pekka Paalanen <pekka.paalanen@collabora.co.uk> Reviewed-by: Nobuhiko Tanibata <NOBUHIKO_TANIBATA@xddp.denso.co.jp> Tested-by: Nobuhiko Tanibata <NOBUHIKO_TANIBATA@xddp.denso.co.jp> Reviewed-By: Derek Foreman <derekf@osg.samsung.com>
2015-02-16 15:39:11 +03:00
if (view->geometry.scissor_enabled &&
!pixman_region32_contains_point(&view->geometry.scissor,
view_ix, view_iy, NULL))
continue;
*vx = view_x;
*vy = view_y;
return view;
}
*vx = wl_fixed_from_int(-1000000);
*vy = wl_fixed_from_int(-1000000);
return NULL;
}
static void
weston_compositor_repick(struct weston_compositor *compositor)
{
struct weston_seat *seat;
if (!compositor->session_active)
return;
wl_list_for_each(seat, &compositor->seat_list, link)
weston_seat_repick(seat);
}
WL_EXPORT void
Split the geometry information from weston_surface out into weston_view The weston_surface structure is split into two structures: * The weston_surface structure storres everything required for a client-side or server-side surface. This includes buffers; callbacks; backend private data; input, damage, and opaque regions; and a few other bookkeeping bits. * The weston_view structure represents an entity in the scenegraph and storres all of the geometry information. This includes clip region, alpha, position, and the transformation list as well as all of the temporary information derived from the geometry state. Because a view, and not a surface, is a scenegraph element, the view is what is placed in layers and planes. There are a few things worth noting about the surface/view split: 1. This is *not* a modification to the protocol. It is, instead, a modification to Weston's internal scenegraph to allow a single surface to exist in multiple places at a time. Clients are completely unaware of how many views to a particular surface exist. 2. A view is considered a direct child of a surface and is destroyed when the surface is destroyed. Because of this, the view.surface pointer is always valid and non-null. 3. The compositor's surface_list is replaced with a view_list. Due to subsurfaces, building the view list is a little more complicated than it used to be and involves building a tree of views on the fly whenever subsurfaces are used. However, this means that backends can remain completely subsurface-agnostic. 4. Surfaces and views both keep track of which outputs they are on. 5. The weston_surface structure now has width and height fields. These are populated when a new buffer is attached before surface.configure is called. This is because there are many surface-based operations that really require the width and height and digging through the views didn't work well. Signed-off-by: Jason Ekstrand <jason@jlekstrand.net>
2013-10-13 07:38:11 +04:00
weston_view_unmap(struct weston_view *view)
{
struct weston_seat *seat;
Split the geometry information from weston_surface out into weston_view The weston_surface structure is split into two structures: * The weston_surface structure storres everything required for a client-side or server-side surface. This includes buffers; callbacks; backend private data; input, damage, and opaque regions; and a few other bookkeeping bits. * The weston_view structure represents an entity in the scenegraph and storres all of the geometry information. This includes clip region, alpha, position, and the transformation list as well as all of the temporary information derived from the geometry state. Because a view, and not a surface, is a scenegraph element, the view is what is placed in layers and planes. There are a few things worth noting about the surface/view split: 1. This is *not* a modification to the protocol. It is, instead, a modification to Weston's internal scenegraph to allow a single surface to exist in multiple places at a time. Clients are completely unaware of how many views to a particular surface exist. 2. A view is considered a direct child of a surface and is destroyed when the surface is destroyed. Because of this, the view.surface pointer is always valid and non-null. 3. The compositor's surface_list is replaced with a view_list. Due to subsurfaces, building the view list is a little more complicated than it used to be and involves building a tree of views on the fly whenever subsurfaces are used. However, this means that backends can remain completely subsurface-agnostic. 4. Surfaces and views both keep track of which outputs they are on. 5. The weston_surface structure now has width and height fields. These are populated when a new buffer is attached before surface.configure is called. This is because there are many surface-based operations that really require the width and height and digging through the views didn't work well. Signed-off-by: Jason Ekstrand <jason@jlekstrand.net>
2013-10-13 07:38:11 +04:00
if (!weston_view_is_mapped(view))
return;
weston_view_damage_below(view);
view->output = NULL;
view->plane = NULL;
view->is_mapped = false;
weston_layer_entry_remove(&view->layer_link);
Split the geometry information from weston_surface out into weston_view The weston_surface structure is split into two structures: * The weston_surface structure storres everything required for a client-side or server-side surface. This includes buffers; callbacks; backend private data; input, damage, and opaque regions; and a few other bookkeeping bits. * The weston_view structure represents an entity in the scenegraph and storres all of the geometry information. This includes clip region, alpha, position, and the transformation list as well as all of the temporary information derived from the geometry state. Because a view, and not a surface, is a scenegraph element, the view is what is placed in layers and planes. There are a few things worth noting about the surface/view split: 1. This is *not* a modification to the protocol. It is, instead, a modification to Weston's internal scenegraph to allow a single surface to exist in multiple places at a time. Clients are completely unaware of how many views to a particular surface exist. 2. A view is considered a direct child of a surface and is destroyed when the surface is destroyed. Because of this, the view.surface pointer is always valid and non-null. 3. The compositor's surface_list is replaced with a view_list. Due to subsurfaces, building the view list is a little more complicated than it used to be and involves building a tree of views on the fly whenever subsurfaces are used. However, this means that backends can remain completely subsurface-agnostic. 4. Surfaces and views both keep track of which outputs they are on. 5. The weston_surface structure now has width and height fields. These are populated when a new buffer is attached before surface.configure is called. This is because there are many surface-based operations that really require the width and height and digging through the views didn't work well. Signed-off-by: Jason Ekstrand <jason@jlekstrand.net>
2013-10-13 07:38:11 +04:00
wl_list_remove(&view->link);
wl_list_init(&view->link);
view->output_mask = 0;
weston_surface_assign_output(view->surface);
if (weston_surface_is_mapped(view->surface))
return;
Split the geometry information from weston_surface out into weston_view The weston_surface structure is split into two structures: * The weston_surface structure storres everything required for a client-side or server-side surface. This includes buffers; callbacks; backend private data; input, damage, and opaque regions; and a few other bookkeeping bits. * The weston_view structure represents an entity in the scenegraph and storres all of the geometry information. This includes clip region, alpha, position, and the transformation list as well as all of the temporary information derived from the geometry state. Because a view, and not a surface, is a scenegraph element, the view is what is placed in layers and planes. There are a few things worth noting about the surface/view split: 1. This is *not* a modification to the protocol. It is, instead, a modification to Weston's internal scenegraph to allow a single surface to exist in multiple places at a time. Clients are completely unaware of how many views to a particular surface exist. 2. A view is considered a direct child of a surface and is destroyed when the surface is destroyed. Because of this, the view.surface pointer is always valid and non-null. 3. The compositor's surface_list is replaced with a view_list. Due to subsurfaces, building the view list is a little more complicated than it used to be and involves building a tree of views on the fly whenever subsurfaces are used. However, this means that backends can remain completely subsurface-agnostic. 4. Surfaces and views both keep track of which outputs they are on. 5. The weston_surface structure now has width and height fields. These are populated when a new buffer is attached before surface.configure is called. This is because there are many surface-based operations that really require the width and height and digging through the views didn't work well. Signed-off-by: Jason Ekstrand <jason@jlekstrand.net>
2013-10-13 07:38:11 +04:00
wl_list_for_each(seat, &view->surface->compositor->seat_list, link) {
struct weston_touch *touch = weston_seat_get_touch(seat);
struct weston_pointer *pointer = weston_seat_get_pointer(seat);
struct weston_keyboard *keyboard =
weston_seat_get_keyboard(seat);
if (keyboard && keyboard->focus == view->surface)
weston_keyboard_set_focus(keyboard, NULL);
if (pointer && pointer->focus == view)
weston_pointer_clear_focus(pointer);
if (touch && touch->focus == view)
weston_touch_set_focus(touch, NULL);
}
Split the geometry information from weston_surface out into weston_view The weston_surface structure is split into two structures: * The weston_surface structure storres everything required for a client-side or server-side surface. This includes buffers; callbacks; backend private data; input, damage, and opaque regions; and a few other bookkeeping bits. * The weston_view structure represents an entity in the scenegraph and storres all of the geometry information. This includes clip region, alpha, position, and the transformation list as well as all of the temporary information derived from the geometry state. Because a view, and not a surface, is a scenegraph element, the view is what is placed in layers and planes. There are a few things worth noting about the surface/view split: 1. This is *not* a modification to the protocol. It is, instead, a modification to Weston's internal scenegraph to allow a single surface to exist in multiple places at a time. Clients are completely unaware of how many views to a particular surface exist. 2. A view is considered a direct child of a surface and is destroyed when the surface is destroyed. Because of this, the view.surface pointer is always valid and non-null. 3. The compositor's surface_list is replaced with a view_list. Due to subsurfaces, building the view list is a little more complicated than it used to be and involves building a tree of views on the fly whenever subsurfaces are used. However, this means that backends can remain completely subsurface-agnostic. 4. Surfaces and views both keep track of which outputs they are on. 5. The weston_surface structure now has width and height fields. These are populated when a new buffer is attached before surface.configure is called. This is because there are many surface-based operations that really require the width and height and digging through the views didn't work well. Signed-off-by: Jason Ekstrand <jason@jlekstrand.net>
2013-10-13 07:38:11 +04:00
}
Split the geometry information from weston_surface out into weston_view The weston_surface structure is split into two structures: * The weston_surface structure storres everything required for a client-side or server-side surface. This includes buffers; callbacks; backend private data; input, damage, and opaque regions; and a few other bookkeeping bits. * The weston_view structure represents an entity in the scenegraph and storres all of the geometry information. This includes clip region, alpha, position, and the transformation list as well as all of the temporary information derived from the geometry state. Because a view, and not a surface, is a scenegraph element, the view is what is placed in layers and planes. There are a few things worth noting about the surface/view split: 1. This is *not* a modification to the protocol. It is, instead, a modification to Weston's internal scenegraph to allow a single surface to exist in multiple places at a time. Clients are completely unaware of how many views to a particular surface exist. 2. A view is considered a direct child of a surface and is destroyed when the surface is destroyed. Because of this, the view.surface pointer is always valid and non-null. 3. The compositor's surface_list is replaced with a view_list. Due to subsurfaces, building the view list is a little more complicated than it used to be and involves building a tree of views on the fly whenever subsurfaces are used. However, this means that backends can remain completely subsurface-agnostic. 4. Surfaces and views both keep track of which outputs they are on. 5. The weston_surface structure now has width and height fields. These are populated when a new buffer is attached before surface.configure is called. This is because there are many surface-based operations that really require the width and height and digging through the views didn't work well. Signed-off-by: Jason Ekstrand <jason@jlekstrand.net>
2013-10-13 07:38:11 +04:00
WL_EXPORT void
weston_surface_unmap(struct weston_surface *surface)
{
struct weston_view *view;
surface->is_mapped = false;
Split the geometry information from weston_surface out into weston_view The weston_surface structure is split into two structures: * The weston_surface structure storres everything required for a client-side or server-side surface. This includes buffers; callbacks; backend private data; input, damage, and opaque regions; and a few other bookkeeping bits. * The weston_view structure represents an entity in the scenegraph and storres all of the geometry information. This includes clip region, alpha, position, and the transformation list as well as all of the temporary information derived from the geometry state. Because a view, and not a surface, is a scenegraph element, the view is what is placed in layers and planes. There are a few things worth noting about the surface/view split: 1. This is *not* a modification to the protocol. It is, instead, a modification to Weston's internal scenegraph to allow a single surface to exist in multiple places at a time. Clients are completely unaware of how many views to a particular surface exist. 2. A view is considered a direct child of a surface and is destroyed when the surface is destroyed. Because of this, the view.surface pointer is always valid and non-null. 3. The compositor's surface_list is replaced with a view_list. Due to subsurfaces, building the view list is a little more complicated than it used to be and involves building a tree of views on the fly whenever subsurfaces are used. However, this means that backends can remain completely subsurface-agnostic. 4. Surfaces and views both keep track of which outputs they are on. 5. The weston_surface structure now has width and height fields. These are populated when a new buffer is attached before surface.configure is called. This is because there are many surface-based operations that really require the width and height and digging through the views didn't work well. Signed-off-by: Jason Ekstrand <jason@jlekstrand.net>
2013-10-13 07:38:11 +04:00
wl_list_for_each(view, &surface->views, surface_link)
weston_view_unmap(view);
surface->output = NULL;
}
static void
weston_surface_reset_pending_buffer(struct weston_surface *surface)
{
weston_surface_state_set_buffer(&surface->pending, NULL);
surface->pending.sx = 0;
surface->pending.sy = 0;
surface->pending.newly_attached = 0;
surface->pending.buffer_viewport.changed = 0;
}
Split the geometry information from weston_surface out into weston_view The weston_surface structure is split into two structures: * The weston_surface structure storres everything required for a client-side or server-side surface. This includes buffers; callbacks; backend private data; input, damage, and opaque regions; and a few other bookkeeping bits. * The weston_view structure represents an entity in the scenegraph and storres all of the geometry information. This includes clip region, alpha, position, and the transformation list as well as all of the temporary information derived from the geometry state. Because a view, and not a surface, is a scenegraph element, the view is what is placed in layers and planes. There are a few things worth noting about the surface/view split: 1. This is *not* a modification to the protocol. It is, instead, a modification to Weston's internal scenegraph to allow a single surface to exist in multiple places at a time. Clients are completely unaware of how many views to a particular surface exist. 2. A view is considered a direct child of a surface and is destroyed when the surface is destroyed. Because of this, the view.surface pointer is always valid and non-null. 3. The compositor's surface_list is replaced with a view_list. Due to subsurfaces, building the view list is a little more complicated than it used to be and involves building a tree of views on the fly whenever subsurfaces are used. However, this means that backends can remain completely subsurface-agnostic. 4. Surfaces and views both keep track of which outputs they are on. 5. The weston_surface structure now has width and height fields. These are populated when a new buffer is attached before surface.configure is called. This is because there are many surface-based operations that really require the width and height and digging through the views didn't work well. Signed-off-by: Jason Ekstrand <jason@jlekstrand.net>
2013-10-13 07:38:11 +04:00
WL_EXPORT void
weston_view_destroy(struct weston_view *view)
{
wl_signal_emit(&view->destroy_signal, view);
assert(wl_list_empty(&view->geometry.child_list));
if (weston_view_is_mapped(view)) {
weston_view_unmap(view);
weston_compositor_build_view_list(view->surface->compositor);
}
wl_list_remove(&view->link);
weston_layer_entry_remove(&view->layer_link);
Split the geometry information from weston_surface out into weston_view The weston_surface structure is split into two structures: * The weston_surface structure storres everything required for a client-side or server-side surface. This includes buffers; callbacks; backend private data; input, damage, and opaque regions; and a few other bookkeeping bits. * The weston_view structure represents an entity in the scenegraph and storres all of the geometry information. This includes clip region, alpha, position, and the transformation list as well as all of the temporary information derived from the geometry state. Because a view, and not a surface, is a scenegraph element, the view is what is placed in layers and planes. There are a few things worth noting about the surface/view split: 1. This is *not* a modification to the protocol. It is, instead, a modification to Weston's internal scenegraph to allow a single surface to exist in multiple places at a time. Clients are completely unaware of how many views to a particular surface exist. 2. A view is considered a direct child of a surface and is destroyed when the surface is destroyed. Because of this, the view.surface pointer is always valid and non-null. 3. The compositor's surface_list is replaced with a view_list. Due to subsurfaces, building the view list is a little more complicated than it used to be and involves building a tree of views on the fly whenever subsurfaces are used. However, this means that backends can remain completely subsurface-agnostic. 4. Surfaces and views both keep track of which outputs they are on. 5. The weston_surface structure now has width and height fields. These are populated when a new buffer is attached before surface.configure is called. This is because there are many surface-based operations that really require the width and height and digging through the views didn't work well. Signed-off-by: Jason Ekstrand <jason@jlekstrand.net>
2013-10-13 07:38:11 +04:00
pixman_region32_fini(&view->clip);
compositor: add weston_view_set_mask() API and state Add API for setting a clip ('scissor' in the code) rectangle per view, in surface coordinates. Ivi-shell requires this feature to be able to implement the IVI Layer Manager API, which includes clipping of surfaces. The names weston_view_set_mask() and weston_view_set_mask_infinite() mirror the existing weston_layer_set_mask*() functions. This view clipping complements the weston_layer clipping, because view clipping is defined in surface local coordinates, while layer mask/clipping is defined in global coordinates. View clipping requires explicit support from the renderers. Therefore a new Weston capability bit is added: WESTON_CAP_VIEW_CLIP_MASK. Shells (and all users) of this new API are required to check the capability bit is set before using the API. Otherwise the rendering will not be what they expect. View clips are inherited through the transformation inheritance mechanism. However, there are restrictions. The clip rectangle can be set only on the root view of a transformation inheritance tree. The additional transformations in child views must not rotate the coordinate axes. These restrictions avoid corner cases in clip inheritance, and keep the renderer implementations as simple as they are right now. Renderers only need to do an additional intersection with the clip rectangle which is always aligned to the surface coordinate system. For more details, see the API documentation in the patch. Signed-off-by: Pekka Paalanen <pekka.paalanen@collabora.co.uk> Reviewed-by: Nobuhiko Tanibata <NOBUHIKO_TANIBATA@xddp.denso.co.jp> Tested-by: Nobuhiko Tanibata <NOBUHIKO_TANIBATA@xddp.denso.co.jp> Reviewed-By: Derek Foreman <derekf@osg.samsung.com>
2015-02-16 15:39:11 +03:00
pixman_region32_fini(&view->geometry.scissor);
Split the geometry information from weston_surface out into weston_view The weston_surface structure is split into two structures: * The weston_surface structure storres everything required for a client-side or server-side surface. This includes buffers; callbacks; backend private data; input, damage, and opaque regions; and a few other bookkeeping bits. * The weston_view structure represents an entity in the scenegraph and storres all of the geometry information. This includes clip region, alpha, position, and the transformation list as well as all of the temporary information derived from the geometry state. Because a view, and not a surface, is a scenegraph element, the view is what is placed in layers and planes. There are a few things worth noting about the surface/view split: 1. This is *not* a modification to the protocol. It is, instead, a modification to Weston's internal scenegraph to allow a single surface to exist in multiple places at a time. Clients are completely unaware of how many views to a particular surface exist. 2. A view is considered a direct child of a surface and is destroyed when the surface is destroyed. Because of this, the view.surface pointer is always valid and non-null. 3. The compositor's surface_list is replaced with a view_list. Due to subsurfaces, building the view list is a little more complicated than it used to be and involves building a tree of views on the fly whenever subsurfaces are used. However, this means that backends can remain completely subsurface-agnostic. 4. Surfaces and views both keep track of which outputs they are on. 5. The weston_surface structure now has width and height fields. These are populated when a new buffer is attached before surface.configure is called. This is because there are many surface-based operations that really require the width and height and digging through the views didn't work well. Signed-off-by: Jason Ekstrand <jason@jlekstrand.net>
2013-10-13 07:38:11 +04:00
pixman_region32_fini(&view->transform.boundingbox);
pixman_region32_fini(&view->transform.opaque);
Split the geometry information from weston_surface out into weston_view The weston_surface structure is split into two structures: * The weston_surface structure storres everything required for a client-side or server-side surface. This includes buffers; callbacks; backend private data; input, damage, and opaque regions; and a few other bookkeeping bits. * The weston_view structure represents an entity in the scenegraph and storres all of the geometry information. This includes clip region, alpha, position, and the transformation list as well as all of the temporary information derived from the geometry state. Because a view, and not a surface, is a scenegraph element, the view is what is placed in layers and planes. There are a few things worth noting about the surface/view split: 1. This is *not* a modification to the protocol. It is, instead, a modification to Weston's internal scenegraph to allow a single surface to exist in multiple places at a time. Clients are completely unaware of how many views to a particular surface exist. 2. A view is considered a direct child of a surface and is destroyed when the surface is destroyed. Because of this, the view.surface pointer is always valid and non-null. 3. The compositor's surface_list is replaced with a view_list. Due to subsurfaces, building the view list is a little more complicated than it used to be and involves building a tree of views on the fly whenever subsurfaces are used. However, this means that backends can remain completely subsurface-agnostic. 4. Surfaces and views both keep track of which outputs they are on. 5. The weston_surface structure now has width and height fields. These are populated when a new buffer is attached before surface.configure is called. This is because there are many surface-based operations that really require the width and height and digging through the views didn't work well. Signed-off-by: Jason Ekstrand <jason@jlekstrand.net>
2013-10-13 07:38:11 +04:00
weston_view_set_transform_parent(view, NULL);
wl_list_remove(&view->surface_link);
free(view);
}
WL_EXPORT void
weston_surface_destroy(struct weston_surface *surface)
{
struct weston_frame_callback *cb, *next;
Split the geometry information from weston_surface out into weston_view The weston_surface structure is split into two structures: * The weston_surface structure storres everything required for a client-side or server-side surface. This includes buffers; callbacks; backend private data; input, damage, and opaque regions; and a few other bookkeeping bits. * The weston_view structure represents an entity in the scenegraph and storres all of the geometry information. This includes clip region, alpha, position, and the transformation list as well as all of the temporary information derived from the geometry state. Because a view, and not a surface, is a scenegraph element, the view is what is placed in layers and planes. There are a few things worth noting about the surface/view split: 1. This is *not* a modification to the protocol. It is, instead, a modification to Weston's internal scenegraph to allow a single surface to exist in multiple places at a time. Clients are completely unaware of how many views to a particular surface exist. 2. A view is considered a direct child of a surface and is destroyed when the surface is destroyed. Because of this, the view.surface pointer is always valid and non-null. 3. The compositor's surface_list is replaced with a view_list. Due to subsurfaces, building the view list is a little more complicated than it used to be and involves building a tree of views on the fly whenever subsurfaces are used. However, this means that backends can remain completely subsurface-agnostic. 4. Surfaces and views both keep track of which outputs they are on. 5. The weston_surface structure now has width and height fields. These are populated when a new buffer is attached before surface.configure is called. This is because there are many surface-based operations that really require the width and height and digging through the views didn't work well. Signed-off-by: Jason Ekstrand <jason@jlekstrand.net>
2013-10-13 07:38:11 +04:00
struct weston_view *ev, *nv;
struct weston_pointer_constraint *constraint, *next_constraint;
if (--surface->ref_count > 0)
return;
assert(surface->resource == NULL);
wl_signal_emit(&surface->destroy_signal, surface);
compositor: introduce sub-surfaces Implement the basic protocol for sub-surfaces: - expose wl_subcompositor global interface - error checking on protocol calls - associate a parent wl_surface to a sub-surface - introduce the sub-surface role, which is exclusive - an implementation of the wl_subsurface interface - allow nesting of sub-surfaces - proper surface transformation inheritance from parent to sub-surfaces - two different modes of wl_surface.commit for sub-surfaces - hook sub-surfaces up to repaint by modifying the repaint list code Struct weston_subsurface is dynamically allocated. For sub-surfaces, it is completely populated. For parent surfaces, weston_subsurface acts only as a link for stacking order purposes. The wl_resource is unused, parent_destroy_listener is not registered, the transform is not linked, etc. Sub-surfaces are not added directly into layers for display or input. Instead, they are hooked up via the sub-surface list present in parent weston_surface. This way sub-surfaces are inherently linked to the parent surface, and cannot be displayed unless the parent is mapped, too. This also eases restacking, as only the parent will be in a layer list. Also, only the main surface should be subject to shell actions. The surface list rebuilding in weston_output_repaint() is modified to process sub-surface lists, if they are non-empty. The sub-surface list always contains the parent, too, unless empty. The collection of frame_callback_list is moved to a later loop, to streamline the surface list rebuild functions. Features still lacking are: - full-surface alpha support for compound windows Changes in v2: - fix a bug in surface mapping: commit a sub-surface would cause the main surface to never be mapped. - remove debug printfs - detect attempt of making a surface its own parent - always zero-alloc weston_subsurface - apply wl_subsurface.set_position in commit, not immediately - add weston_surface_to_subsurface() - implement sub-surface commit modes parent-cached and independent - implement wl_subcompositor.destroy and wl_subsurface.destroy Changes in v3: - rebased, and use the new transform inheritance code - squashed the commit "add sub-surfaces to repaint list" - fixed a buffer reference leak in commit_from_cache() - Rewrite the sub-surface destructor code, and make it leave the wl_subsurface protocol object inert, if one destroys the corresponding wl_surface. - replaced set_commit_mode with set_sync and set_desync - allowed sub-surface nesting, and fixed repaint accordingly - implemented nested sub-surface commit modes - Made the sub-surface order changes from wl_subsurface.place_above and .place_below to be applied when the parent surface state is applied, instead of immediately. This conforms with the protocol specification now. Signed-off-by: Pekka Paalanen <ppaalanen@gmail.com>
2013-04-25 14:57:42 +04:00
assert(wl_list_empty(&surface->subsurface_list_pending));
assert(wl_list_empty(&surface->subsurface_list));
Split the geometry information from weston_surface out into weston_view The weston_surface structure is split into two structures: * The weston_surface structure storres everything required for a client-side or server-side surface. This includes buffers; callbacks; backend private data; input, damage, and opaque regions; and a few other bookkeeping bits. * The weston_view structure represents an entity in the scenegraph and storres all of the geometry information. This includes clip region, alpha, position, and the transformation list as well as all of the temporary information derived from the geometry state. Because a view, and not a surface, is a scenegraph element, the view is what is placed in layers and planes. There are a few things worth noting about the surface/view split: 1. This is *not* a modification to the protocol. It is, instead, a modification to Weston's internal scenegraph to allow a single surface to exist in multiple places at a time. Clients are completely unaware of how many views to a particular surface exist. 2. A view is considered a direct child of a surface and is destroyed when the surface is destroyed. Because of this, the view.surface pointer is always valid and non-null. 3. The compositor's surface_list is replaced with a view_list. Due to subsurfaces, building the view list is a little more complicated than it used to be and involves building a tree of views on the fly whenever subsurfaces are used. However, this means that backends can remain completely subsurface-agnostic. 4. Surfaces and views both keep track of which outputs they are on. 5. The weston_surface structure now has width and height fields. These are populated when a new buffer is attached before surface.configure is called. This is because there are many surface-based operations that really require the width and height and digging through the views didn't work well. Signed-off-by: Jason Ekstrand <jason@jlekstrand.net>
2013-10-13 07:38:11 +04:00
wl_list_for_each_safe(ev, nv, &surface->views, surface_link)
weston_view_destroy(ev);
weston_surface_state_fini(&surface->pending);
weston_buffer_reference(&surface->buffer_ref, NULL);
pixman_region32_fini(&surface->damage);
pixman_region32_fini(&surface->opaque);
pixman_region32_fini(&surface->input);
wl_list_for_each_safe(cb, next, &surface->frame_callback_list, link)
wl_resource_destroy(cb->resource);
weston_presentation_feedback_discard_list(&surface->feedback_list);
wl_list_for_each_safe(constraint, next_constraint,
&surface->pointer_constraints,
link)
weston_pointer_constraint_destroy(constraint);
free(surface);
}
static void
destroy_surface(struct wl_resource *resource)
{
struct weston_surface *surface = wl_resource_get_user_data(resource);
assert(surface);
/* Set the resource to NULL, since we don't want to leave a
* dangling pointer if the surface was refcounted and survives
* the weston_surface_destroy() call. */
surface->resource = NULL;
if (surface->viewport_resource)
wl_resource_set_user_data(surface->viewport_resource, NULL);
weston_surface_destroy(surface);
}
static void
weston_buffer_destroy_handler(struct wl_listener *listener, void *data)
{
struct weston_buffer *buffer =
container_of(listener, struct weston_buffer, destroy_listener);
wl_signal_emit(&buffer->destroy_signal, buffer);
free(buffer);
}
WL_EXPORT struct weston_buffer *
weston_buffer_from_resource(struct wl_resource *resource)
{
struct weston_buffer *buffer;
struct wl_listener *listener;
listener = wl_resource_get_destroy_listener(resource,
weston_buffer_destroy_handler);
if (listener)
return container_of(listener, struct weston_buffer,
destroy_listener);
buffer = zalloc(sizeof *buffer);
if (buffer == NULL)
return NULL;
buffer->resource = resource;
wl_signal_init(&buffer->destroy_signal);
buffer->destroy_listener.notify = weston_buffer_destroy_handler;
buffer->y_inverted = 1;
wl_resource_add_destroy_listener(resource, &buffer->destroy_listener);
return buffer;
}
static void
weston_buffer_reference_handle_destroy(struct wl_listener *listener,
void *data)
{
struct weston_buffer_reference *ref =
container_of(listener, struct weston_buffer_reference,
destroy_listener);
assert((struct weston_buffer *)data == ref->buffer);
ref->buffer = NULL;
}
WL_EXPORT void
weston_buffer_reference(struct weston_buffer_reference *ref,
struct weston_buffer *buffer)
{
if (ref->buffer && buffer != ref->buffer) {
ref->buffer->busy_count--;
if (ref->buffer->busy_count == 0) {
assert(wl_resource_get_client(ref->buffer->resource));
wl_resource_queue_event(ref->buffer->resource,
WL_BUFFER_RELEASE);
}
wl_list_remove(&ref->destroy_listener.link);
}
if (buffer && buffer != ref->buffer) {
buffer->busy_count++;
wl_signal_add(&buffer->destroy_signal,
&ref->destroy_listener);
}
ref->buffer = buffer;
ref->destroy_listener.notify = weston_buffer_reference_handle_destroy;
}
static void
weston_surface_attach(struct weston_surface *surface,
struct weston_buffer *buffer)
{
weston_buffer_reference(&surface->buffer_ref, buffer);
if (!buffer) {
if (weston_surface_is_mapped(surface))
weston_surface_unmap(surface);
}
surface->compositor->renderer->attach(surface, buffer);
weston_surface_calculate_size_from_buffer(surface);
weston_presentation_feedback_discard_list(&surface->feedback_list);
}
WL_EXPORT void
weston_compositor_damage_all(struct weston_compositor *compositor)
{
struct weston_output *output;
wl_list_for_each(output, &compositor->output_list, link)
weston_output_damage(output);
}
WL_EXPORT void
weston_output_damage(struct weston_output *output)
{
struct weston_compositor *compositor = output->compositor;
pixman_region32_union(&compositor->primary_plane.damage,
&compositor->primary_plane.damage,
&output->region);
weston_output_schedule_repaint(output);
}
static void
Split the geometry information from weston_surface out into weston_view The weston_surface structure is split into two structures: * The weston_surface structure storres everything required for a client-side or server-side surface. This includes buffers; callbacks; backend private data; input, damage, and opaque regions; and a few other bookkeeping bits. * The weston_view structure represents an entity in the scenegraph and storres all of the geometry information. This includes clip region, alpha, position, and the transformation list as well as all of the temporary information derived from the geometry state. Because a view, and not a surface, is a scenegraph element, the view is what is placed in layers and planes. There are a few things worth noting about the surface/view split: 1. This is *not* a modification to the protocol. It is, instead, a modification to Weston's internal scenegraph to allow a single surface to exist in multiple places at a time. Clients are completely unaware of how many views to a particular surface exist. 2. A view is considered a direct child of a surface and is destroyed when the surface is destroyed. Because of this, the view.surface pointer is always valid and non-null. 3. The compositor's surface_list is replaced with a view_list. Due to subsurfaces, building the view list is a little more complicated than it used to be and involves building a tree of views on the fly whenever subsurfaces are used. However, this means that backends can remain completely subsurface-agnostic. 4. Surfaces and views both keep track of which outputs they are on. 5. The weston_surface structure now has width and height fields. These are populated when a new buffer is attached before surface.configure is called. This is because there are many surface-based operations that really require the width and height and digging through the views didn't work well. Signed-off-by: Jason Ekstrand <jason@jlekstrand.net>
2013-10-13 07:38:11 +04:00
surface_flush_damage(struct weston_surface *surface)
{
if (surface->buffer_ref.buffer &&
wl_shm_buffer_get(surface->buffer_ref.buffer->resource))
surface->compositor->renderer->flush_damage(surface);
compositor: Implement JSON-timeline logging Logging is activated and deactivated with the debug key binding 't'. When activated, it creates a new log file, where it records the events. The log file contains events and detailed object information entries in JSON format, and is meant to be parsed in sequence from beginning to the end. The emitted events are mostly related to the output repaint cycle, like when repaint begins, is submitted to GPU, and when it completes on a vblank. This is recorded per-output. Also some per-surface events are recorded, including when surface damage is flushed. To reduce the log size, events refer to objects like outputs and surfaces by id numbers. Detailed object information is emitted only as needed: on the first object occurrence, and afterwards only if weston_timeline_object::force_refresh asks for it. The detailed information for surfaces includes the string returned by weston_surface::get_label. Therefore it is important to set weston_timeline_object::force_refresh = 1 whenever the string would change, so that the new details get recorded. A rudimentary parser and SVG generator can be found at: https://github.com/ppaalanen/wesgr The timeline logs can answer questions including: - How does the compositor repaint cycle work timing-wise? - When was the vblank deadline missed? - What is the latency from surface commit to showing the new content on screen? - How long does it take to process the scenegraph? v2: weston_surface::get_description renamed to get_label. v3: reafctor a bit into fprint_quoted_string(). Signed-off-by: Pekka Paalanen <pekka.paalanen@collabora.co.uk>
2014-11-12 16:09:24 +03:00
if (weston_timeline_enabled_ &&
pixman_region32_not_empty(&surface->damage))
TL_POINT("core_flush_damage", TLP_SURFACE(surface),
TLP_OUTPUT(surface->output), TLP_END);
pixman_region32_clear(&surface->damage);
Split the geometry information from weston_surface out into weston_view The weston_surface structure is split into two structures: * The weston_surface structure storres everything required for a client-side or server-side surface. This includes buffers; callbacks; backend private data; input, damage, and opaque regions; and a few other bookkeeping bits. * The weston_view structure represents an entity in the scenegraph and storres all of the geometry information. This includes clip region, alpha, position, and the transformation list as well as all of the temporary information derived from the geometry state. Because a view, and not a surface, is a scenegraph element, the view is what is placed in layers and planes. There are a few things worth noting about the surface/view split: 1. This is *not* a modification to the protocol. It is, instead, a modification to Weston's internal scenegraph to allow a single surface to exist in multiple places at a time. Clients are completely unaware of how many views to a particular surface exist. 2. A view is considered a direct child of a surface and is destroyed when the surface is destroyed. Because of this, the view.surface pointer is always valid and non-null. 3. The compositor's surface_list is replaced with a view_list. Due to subsurfaces, building the view list is a little more complicated than it used to be and involves building a tree of views on the fly whenever subsurfaces are used. However, this means that backends can remain completely subsurface-agnostic. 4. Surfaces and views both keep track of which outputs they are on. 5. The weston_surface structure now has width and height fields. These are populated when a new buffer is attached before surface.configure is called. This is because there are many surface-based operations that really require the width and height and digging through the views didn't work well. Signed-off-by: Jason Ekstrand <jason@jlekstrand.net>
2013-10-13 07:38:11 +04:00
}
static void
view_accumulate_damage(struct weston_view *view,
pixman_region32_t *opaque)
{
pixman_region32_t damage;
pixman_region32_init(&damage);
if (view->transform.enabled) {
pixman_box32_t *extents;
Split the geometry information from weston_surface out into weston_view The weston_surface structure is split into two structures: * The weston_surface structure storres everything required for a client-side or server-side surface. This includes buffers; callbacks; backend private data; input, damage, and opaque regions; and a few other bookkeeping bits. * The weston_view structure represents an entity in the scenegraph and storres all of the geometry information. This includes clip region, alpha, position, and the transformation list as well as all of the temporary information derived from the geometry state. Because a view, and not a surface, is a scenegraph element, the view is what is placed in layers and planes. There are a few things worth noting about the surface/view split: 1. This is *not* a modification to the protocol. It is, instead, a modification to Weston's internal scenegraph to allow a single surface to exist in multiple places at a time. Clients are completely unaware of how many views to a particular surface exist. 2. A view is considered a direct child of a surface and is destroyed when the surface is destroyed. Because of this, the view.surface pointer is always valid and non-null. 3. The compositor's surface_list is replaced with a view_list. Due to subsurfaces, building the view list is a little more complicated than it used to be and involves building a tree of views on the fly whenever subsurfaces are used. However, this means that backends can remain completely subsurface-agnostic. 4. Surfaces and views both keep track of which outputs they are on. 5. The weston_surface structure now has width and height fields. These are populated when a new buffer is attached before surface.configure is called. This is because there are many surface-based operations that really require the width and height and digging through the views didn't work well. Signed-off-by: Jason Ekstrand <jason@jlekstrand.net>
2013-10-13 07:38:11 +04:00
extents = pixman_region32_extents(&view->surface->damage);
view_compute_bbox(view, extents, &damage);
} else {
Split the geometry information from weston_surface out into weston_view The weston_surface structure is split into two structures: * The weston_surface structure storres everything required for a client-side or server-side surface. This includes buffers; callbacks; backend private data; input, damage, and opaque regions; and a few other bookkeeping bits. * The weston_view structure represents an entity in the scenegraph and storres all of the geometry information. This includes clip region, alpha, position, and the transformation list as well as all of the temporary information derived from the geometry state. Because a view, and not a surface, is a scenegraph element, the view is what is placed in layers and planes. There are a few things worth noting about the surface/view split: 1. This is *not* a modification to the protocol. It is, instead, a modification to Weston's internal scenegraph to allow a single surface to exist in multiple places at a time. Clients are completely unaware of how many views to a particular surface exist. 2. A view is considered a direct child of a surface and is destroyed when the surface is destroyed. Because of this, the view.surface pointer is always valid and non-null. 3. The compositor's surface_list is replaced with a view_list. Due to subsurfaces, building the view list is a little more complicated than it used to be and involves building a tree of views on the fly whenever subsurfaces are used. However, this means that backends can remain completely subsurface-agnostic. 4. Surfaces and views both keep track of which outputs they are on. 5. The weston_surface structure now has width and height fields. These are populated when a new buffer is attached before surface.configure is called. This is because there are many surface-based operations that really require the width and height and digging through the views didn't work well. Signed-off-by: Jason Ekstrand <jason@jlekstrand.net>
2013-10-13 07:38:11 +04:00
pixman_region32_copy(&damage, &view->surface->damage);
pixman_region32_translate(&damage,
view->geometry.x, view->geometry.y);
}
compositor: add weston_view_set_mask() API and state Add API for setting a clip ('scissor' in the code) rectangle per view, in surface coordinates. Ivi-shell requires this feature to be able to implement the IVI Layer Manager API, which includes clipping of surfaces. The names weston_view_set_mask() and weston_view_set_mask_infinite() mirror the existing weston_layer_set_mask*() functions. This view clipping complements the weston_layer clipping, because view clipping is defined in surface local coordinates, while layer mask/clipping is defined in global coordinates. View clipping requires explicit support from the renderers. Therefore a new Weston capability bit is added: WESTON_CAP_VIEW_CLIP_MASK. Shells (and all users) of this new API are required to check the capability bit is set before using the API. Otherwise the rendering will not be what they expect. View clips are inherited through the transformation inheritance mechanism. However, there are restrictions. The clip rectangle can be set only on the root view of a transformation inheritance tree. The additional transformations in child views must not rotate the coordinate axes. These restrictions avoid corner cases in clip inheritance, and keep the renderer implementations as simple as they are right now. Renderers only need to do an additional intersection with the clip rectangle which is always aligned to the surface coordinate system. For more details, see the API documentation in the patch. Signed-off-by: Pekka Paalanen <pekka.paalanen@collabora.co.uk> Reviewed-by: Nobuhiko Tanibata <NOBUHIKO_TANIBATA@xddp.denso.co.jp> Tested-by: Nobuhiko Tanibata <NOBUHIKO_TANIBATA@xddp.denso.co.jp> Reviewed-By: Derek Foreman <derekf@osg.samsung.com>
2015-02-16 15:39:11 +03:00
pixman_region32_intersect(&damage, &damage,
&view->transform.boundingbox);
Split the geometry information from weston_surface out into weston_view The weston_surface structure is split into two structures: * The weston_surface structure storres everything required for a client-side or server-side surface. This includes buffers; callbacks; backend private data; input, damage, and opaque regions; and a few other bookkeeping bits. * The weston_view structure represents an entity in the scenegraph and storres all of the geometry information. This includes clip region, alpha, position, and the transformation list as well as all of the temporary information derived from the geometry state. Because a view, and not a surface, is a scenegraph element, the view is what is placed in layers and planes. There are a few things worth noting about the surface/view split: 1. This is *not* a modification to the protocol. It is, instead, a modification to Weston's internal scenegraph to allow a single surface to exist in multiple places at a time. Clients are completely unaware of how many views to a particular surface exist. 2. A view is considered a direct child of a surface and is destroyed when the surface is destroyed. Because of this, the view.surface pointer is always valid and non-null. 3. The compositor's surface_list is replaced with a view_list. Due to subsurfaces, building the view list is a little more complicated than it used to be and involves building a tree of views on the fly whenever subsurfaces are used. However, this means that backends can remain completely subsurface-agnostic. 4. Surfaces and views both keep track of which outputs they are on. 5. The weston_surface structure now has width and height fields. These are populated when a new buffer is attached before surface.configure is called. This is because there are many surface-based operations that really require the width and height and digging through the views didn't work well. Signed-off-by: Jason Ekstrand <jason@jlekstrand.net>
2013-10-13 07:38:11 +04:00
pixman_region32_subtract(&damage, &damage, opaque);
pixman_region32_union(&view->plane->damage,
&view->plane->damage, &damage);
pixman_region32_fini(&damage);
pixman_region32_copy(&view->clip, opaque);
pixman_region32_union(opaque, opaque, &view->transform.opaque);
}
static void
compositor_accumulate_damage(struct weston_compositor *ec)
{
struct weston_plane *plane;
Split the geometry information from weston_surface out into weston_view The weston_surface structure is split into two structures: * The weston_surface structure storres everything required for a client-side or server-side surface. This includes buffers; callbacks; backend private data; input, damage, and opaque regions; and a few other bookkeeping bits. * The weston_view structure represents an entity in the scenegraph and storres all of the geometry information. This includes clip region, alpha, position, and the transformation list as well as all of the temporary information derived from the geometry state. Because a view, and not a surface, is a scenegraph element, the view is what is placed in layers and planes. There are a few things worth noting about the surface/view split: 1. This is *not* a modification to the protocol. It is, instead, a modification to Weston's internal scenegraph to allow a single surface to exist in multiple places at a time. Clients are completely unaware of how many views to a particular surface exist. 2. A view is considered a direct child of a surface and is destroyed when the surface is destroyed. Because of this, the view.surface pointer is always valid and non-null. 3. The compositor's surface_list is replaced with a view_list. Due to subsurfaces, building the view list is a little more complicated than it used to be and involves building a tree of views on the fly whenever subsurfaces are used. However, this means that backends can remain completely subsurface-agnostic. 4. Surfaces and views both keep track of which outputs they are on. 5. The weston_surface structure now has width and height fields. These are populated when a new buffer is attached before surface.configure is called. This is because there are many surface-based operations that really require the width and height and digging through the views didn't work well. Signed-off-by: Jason Ekstrand <jason@jlekstrand.net>
2013-10-13 07:38:11 +04:00
struct weston_view *ev;
pixman_region32_t opaque, clip;
pixman_region32_init(&clip);
wl_list_for_each(plane, &ec->plane_list, link) {
pixman_region32_copy(&plane->clip, &clip);
pixman_region32_init(&opaque);
Split the geometry information from weston_surface out into weston_view The weston_surface structure is split into two structures: * The weston_surface structure storres everything required for a client-side or server-side surface. This includes buffers; callbacks; backend private data; input, damage, and opaque regions; and a few other bookkeeping bits. * The weston_view structure represents an entity in the scenegraph and storres all of the geometry information. This includes clip region, alpha, position, and the transformation list as well as all of the temporary information derived from the geometry state. Because a view, and not a surface, is a scenegraph element, the view is what is placed in layers and planes. There are a few things worth noting about the surface/view split: 1. This is *not* a modification to the protocol. It is, instead, a modification to Weston's internal scenegraph to allow a single surface to exist in multiple places at a time. Clients are completely unaware of how many views to a particular surface exist. 2. A view is considered a direct child of a surface and is destroyed when the surface is destroyed. Because of this, the view.surface pointer is always valid and non-null. 3. The compositor's surface_list is replaced with a view_list. Due to subsurfaces, building the view list is a little more complicated than it used to be and involves building a tree of views on the fly whenever subsurfaces are used. However, this means that backends can remain completely subsurface-agnostic. 4. Surfaces and views both keep track of which outputs they are on. 5. The weston_surface structure now has width and height fields. These are populated when a new buffer is attached before surface.configure is called. This is because there are many surface-based operations that really require the width and height and digging through the views didn't work well. Signed-off-by: Jason Ekstrand <jason@jlekstrand.net>
2013-10-13 07:38:11 +04:00
wl_list_for_each(ev, &ec->view_list, link) {
if (ev->plane != plane)
continue;
Split the geometry information from weston_surface out into weston_view The weston_surface structure is split into two structures: * The weston_surface structure storres everything required for a client-side or server-side surface. This includes buffers; callbacks; backend private data; input, damage, and opaque regions; and a few other bookkeeping bits. * The weston_view structure represents an entity in the scenegraph and storres all of the geometry information. This includes clip region, alpha, position, and the transformation list as well as all of the temporary information derived from the geometry state. Because a view, and not a surface, is a scenegraph element, the view is what is placed in layers and planes. There are a few things worth noting about the surface/view split: 1. This is *not* a modification to the protocol. It is, instead, a modification to Weston's internal scenegraph to allow a single surface to exist in multiple places at a time. Clients are completely unaware of how many views to a particular surface exist. 2. A view is considered a direct child of a surface and is destroyed when the surface is destroyed. Because of this, the view.surface pointer is always valid and non-null. 3. The compositor's surface_list is replaced with a view_list. Due to subsurfaces, building the view list is a little more complicated than it used to be and involves building a tree of views on the fly whenever subsurfaces are used. However, this means that backends can remain completely subsurface-agnostic. 4. Surfaces and views both keep track of which outputs they are on. 5. The weston_surface structure now has width and height fields. These are populated when a new buffer is attached before surface.configure is called. This is because there are many surface-based operations that really require the width and height and digging through the views didn't work well. Signed-off-by: Jason Ekstrand <jason@jlekstrand.net>
2013-10-13 07:38:11 +04:00
view_accumulate_damage(ev, &opaque);
}
pixman_region32_union(&clip, &clip, &opaque);
pixman_region32_fini(&opaque);
}
pixman_region32_fini(&clip);
Split the geometry information from weston_surface out into weston_view The weston_surface structure is split into two structures: * The weston_surface structure storres everything required for a client-side or server-side surface. This includes buffers; callbacks; backend private data; input, damage, and opaque regions; and a few other bookkeeping bits. * The weston_view structure represents an entity in the scenegraph and storres all of the geometry information. This includes clip region, alpha, position, and the transformation list as well as all of the temporary information derived from the geometry state. Because a view, and not a surface, is a scenegraph element, the view is what is placed in layers and planes. There are a few things worth noting about the surface/view split: 1. This is *not* a modification to the protocol. It is, instead, a modification to Weston's internal scenegraph to allow a single surface to exist in multiple places at a time. Clients are completely unaware of how many views to a particular surface exist. 2. A view is considered a direct child of a surface and is destroyed when the surface is destroyed. Because of this, the view.surface pointer is always valid and non-null. 3. The compositor's surface_list is replaced with a view_list. Due to subsurfaces, building the view list is a little more complicated than it used to be and involves building a tree of views on the fly whenever subsurfaces are used. However, this means that backends can remain completely subsurface-agnostic. 4. Surfaces and views both keep track of which outputs they are on. 5. The weston_surface structure now has width and height fields. These are populated when a new buffer is attached before surface.configure is called. This is because there are many surface-based operations that really require the width and height and digging through the views didn't work well. Signed-off-by: Jason Ekstrand <jason@jlekstrand.net>
2013-10-13 07:38:11 +04:00
wl_list_for_each(ev, &ec->view_list, link)
ev->surface->touched = false;
Split the geometry information from weston_surface out into weston_view The weston_surface structure is split into two structures: * The weston_surface structure storres everything required for a client-side or server-side surface. This includes buffers; callbacks; backend private data; input, damage, and opaque regions; and a few other bookkeeping bits. * The weston_view structure represents an entity in the scenegraph and storres all of the geometry information. This includes clip region, alpha, position, and the transformation list as well as all of the temporary information derived from the geometry state. Because a view, and not a surface, is a scenegraph element, the view is what is placed in layers and planes. There are a few things worth noting about the surface/view split: 1. This is *not* a modification to the protocol. It is, instead, a modification to Weston's internal scenegraph to allow a single surface to exist in multiple places at a time. Clients are completely unaware of how many views to a particular surface exist. 2. A view is considered a direct child of a surface and is destroyed when the surface is destroyed. Because of this, the view.surface pointer is always valid and non-null. 3. The compositor's surface_list is replaced with a view_list. Due to subsurfaces, building the view list is a little more complicated than it used to be and involves building a tree of views on the fly whenever subsurfaces are used. However, this means that backends can remain completely subsurface-agnostic. 4. Surfaces and views both keep track of which outputs they are on. 5. The weston_surface structure now has width and height fields. These are populated when a new buffer is attached before surface.configure is called. This is because there are many surface-based operations that really require the width and height and digging through the views didn't work well. Signed-off-by: Jason Ekstrand <jason@jlekstrand.net>
2013-10-13 07:38:11 +04:00
wl_list_for_each(ev, &ec->view_list, link) {
if (ev->surface->touched)
continue;
ev->surface->touched = true;
Split the geometry information from weston_surface out into weston_view The weston_surface structure is split into two structures: * The weston_surface structure storres everything required for a client-side or server-side surface. This includes buffers; callbacks; backend private data; input, damage, and opaque regions; and a few other bookkeeping bits. * The weston_view structure represents an entity in the scenegraph and storres all of the geometry information. This includes clip region, alpha, position, and the transformation list as well as all of the temporary information derived from the geometry state. Because a view, and not a surface, is a scenegraph element, the view is what is placed in layers and planes. There are a few things worth noting about the surface/view split: 1. This is *not* a modification to the protocol. It is, instead, a modification to Weston's internal scenegraph to allow a single surface to exist in multiple places at a time. Clients are completely unaware of how many views to a particular surface exist. 2. A view is considered a direct child of a surface and is destroyed when the surface is destroyed. Because of this, the view.surface pointer is always valid and non-null. 3. The compositor's surface_list is replaced with a view_list. Due to subsurfaces, building the view list is a little more complicated than it used to be and involves building a tree of views on the fly whenever subsurfaces are used. However, this means that backends can remain completely subsurface-agnostic. 4. Surfaces and views both keep track of which outputs they are on. 5. The weston_surface structure now has width and height fields. These are populated when a new buffer is attached before surface.configure is called. This is because there are many surface-based operations that really require the width and height and digging through the views didn't work well. Signed-off-by: Jason Ekstrand <jason@jlekstrand.net>
2013-10-13 07:38:11 +04:00
surface_flush_damage(ev->surface);
/* Both the renderer and the backend have seen the buffer
* by now. If renderer needs the buffer, it has its own
* reference set. If the backend wants to keep the buffer
* around for migrating the surface into a non-primary plane
* later, keep_buffer is true. Otherwise, drop the core
* reference now, and allow early buffer release. This enables
* clients to use single-buffering.
*/
Split the geometry information from weston_surface out into weston_view The weston_surface structure is split into two structures: * The weston_surface structure storres everything required for a client-side or server-side surface. This includes buffers; callbacks; backend private data; input, damage, and opaque regions; and a few other bookkeeping bits. * The weston_view structure represents an entity in the scenegraph and storres all of the geometry information. This includes clip region, alpha, position, and the transformation list as well as all of the temporary information derived from the geometry state. Because a view, and not a surface, is a scenegraph element, the view is what is placed in layers and planes. There are a few things worth noting about the surface/view split: 1. This is *not* a modification to the protocol. It is, instead, a modification to Weston's internal scenegraph to allow a single surface to exist in multiple places at a time. Clients are completely unaware of how many views to a particular surface exist. 2. A view is considered a direct child of a surface and is destroyed when the surface is destroyed. Because of this, the view.surface pointer is always valid and non-null. 3. The compositor's surface_list is replaced with a view_list. Due to subsurfaces, building the view list is a little more complicated than it used to be and involves building a tree of views on the fly whenever subsurfaces are used. However, this means that backends can remain completely subsurface-agnostic. 4. Surfaces and views both keep track of which outputs they are on. 5. The weston_surface structure now has width and height fields. These are populated when a new buffer is attached before surface.configure is called. This is because there are many surface-based operations that really require the width and height and digging through the views didn't work well. Signed-off-by: Jason Ekstrand <jason@jlekstrand.net>
2013-10-13 07:38:11 +04:00
if (!ev->surface->keep_buffer)
weston_buffer_reference(&ev->surface->buffer_ref, NULL);
}
}
compositor: introduce sub-surfaces Implement the basic protocol for sub-surfaces: - expose wl_subcompositor global interface - error checking on protocol calls - associate a parent wl_surface to a sub-surface - introduce the sub-surface role, which is exclusive - an implementation of the wl_subsurface interface - allow nesting of sub-surfaces - proper surface transformation inheritance from parent to sub-surfaces - two different modes of wl_surface.commit for sub-surfaces - hook sub-surfaces up to repaint by modifying the repaint list code Struct weston_subsurface is dynamically allocated. For sub-surfaces, it is completely populated. For parent surfaces, weston_subsurface acts only as a link for stacking order purposes. The wl_resource is unused, parent_destroy_listener is not registered, the transform is not linked, etc. Sub-surfaces are not added directly into layers for display or input. Instead, they are hooked up via the sub-surface list present in parent weston_surface. This way sub-surfaces are inherently linked to the parent surface, and cannot be displayed unless the parent is mapped, too. This also eases restacking, as only the parent will be in a layer list. Also, only the main surface should be subject to shell actions. The surface list rebuilding in weston_output_repaint() is modified to process sub-surface lists, if they are non-empty. The sub-surface list always contains the parent, too, unless empty. The collection of frame_callback_list is moved to a later loop, to streamline the surface list rebuild functions. Features still lacking are: - full-surface alpha support for compound windows Changes in v2: - fix a bug in surface mapping: commit a sub-surface would cause the main surface to never be mapped. - remove debug printfs - detect attempt of making a surface its own parent - always zero-alloc weston_subsurface - apply wl_subsurface.set_position in commit, not immediately - add weston_surface_to_subsurface() - implement sub-surface commit modes parent-cached and independent - implement wl_subcompositor.destroy and wl_subsurface.destroy Changes in v3: - rebased, and use the new transform inheritance code - squashed the commit "add sub-surfaces to repaint list" - fixed a buffer reference leak in commit_from_cache() - Rewrite the sub-surface destructor code, and make it leave the wl_subsurface protocol object inert, if one destroys the corresponding wl_surface. - replaced set_commit_mode with set_sync and set_desync - allowed sub-surface nesting, and fixed repaint accordingly - implemented nested sub-surface commit modes - Made the sub-surface order changes from wl_subsurface.place_above and .place_below to be applied when the parent surface state is applied, instead of immediately. This conforms with the protocol specification now. Signed-off-by: Pekka Paalanen <ppaalanen@gmail.com>
2013-04-25 14:57:42 +04:00
static void
Split the geometry information from weston_surface out into weston_view The weston_surface structure is split into two structures: * The weston_surface structure storres everything required for a client-side or server-side surface. This includes buffers; callbacks; backend private data; input, damage, and opaque regions; and a few other bookkeeping bits. * The weston_view structure represents an entity in the scenegraph and storres all of the geometry information. This includes clip region, alpha, position, and the transformation list as well as all of the temporary information derived from the geometry state. Because a view, and not a surface, is a scenegraph element, the view is what is placed in layers and planes. There are a few things worth noting about the surface/view split: 1. This is *not* a modification to the protocol. It is, instead, a modification to Weston's internal scenegraph to allow a single surface to exist in multiple places at a time. Clients are completely unaware of how many views to a particular surface exist. 2. A view is considered a direct child of a surface and is destroyed when the surface is destroyed. Because of this, the view.surface pointer is always valid and non-null. 3. The compositor's surface_list is replaced with a view_list. Due to subsurfaces, building the view list is a little more complicated than it used to be and involves building a tree of views on the fly whenever subsurfaces are used. However, this means that backends can remain completely subsurface-agnostic. 4. Surfaces and views both keep track of which outputs they are on. 5. The weston_surface structure now has width and height fields. These are populated when a new buffer is attached before surface.configure is called. This is because there are many surface-based operations that really require the width and height and digging through the views didn't work well. Signed-off-by: Jason Ekstrand <jason@jlekstrand.net>
2013-10-13 07:38:11 +04:00
surface_stash_subsurface_views(struct weston_surface *surface)
compositor: introduce sub-surfaces Implement the basic protocol for sub-surfaces: - expose wl_subcompositor global interface - error checking on protocol calls - associate a parent wl_surface to a sub-surface - introduce the sub-surface role, which is exclusive - an implementation of the wl_subsurface interface - allow nesting of sub-surfaces - proper surface transformation inheritance from parent to sub-surfaces - two different modes of wl_surface.commit for sub-surfaces - hook sub-surfaces up to repaint by modifying the repaint list code Struct weston_subsurface is dynamically allocated. For sub-surfaces, it is completely populated. For parent surfaces, weston_subsurface acts only as a link for stacking order purposes. The wl_resource is unused, parent_destroy_listener is not registered, the transform is not linked, etc. Sub-surfaces are not added directly into layers for display or input. Instead, they are hooked up via the sub-surface list present in parent weston_surface. This way sub-surfaces are inherently linked to the parent surface, and cannot be displayed unless the parent is mapped, too. This also eases restacking, as only the parent will be in a layer list. Also, only the main surface should be subject to shell actions. The surface list rebuilding in weston_output_repaint() is modified to process sub-surface lists, if they are non-empty. The sub-surface list always contains the parent, too, unless empty. The collection of frame_callback_list is moved to a later loop, to streamline the surface list rebuild functions. Features still lacking are: - full-surface alpha support for compound windows Changes in v2: - fix a bug in surface mapping: commit a sub-surface would cause the main surface to never be mapped. - remove debug printfs - detect attempt of making a surface its own parent - always zero-alloc weston_subsurface - apply wl_subsurface.set_position in commit, not immediately - add weston_surface_to_subsurface() - implement sub-surface commit modes parent-cached and independent - implement wl_subcompositor.destroy and wl_subsurface.destroy Changes in v3: - rebased, and use the new transform inheritance code - squashed the commit "add sub-surfaces to repaint list" - fixed a buffer reference leak in commit_from_cache() - Rewrite the sub-surface destructor code, and make it leave the wl_subsurface protocol object inert, if one destroys the corresponding wl_surface. - replaced set_commit_mode with set_sync and set_desync - allowed sub-surface nesting, and fixed repaint accordingly - implemented nested sub-surface commit modes - Made the sub-surface order changes from wl_subsurface.place_above and .place_below to be applied when the parent surface state is applied, instead of immediately. This conforms with the protocol specification now. Signed-off-by: Pekka Paalanen <ppaalanen@gmail.com>
2013-04-25 14:57:42 +04:00
{
struct weston_subsurface *sub;
Split the geometry information from weston_surface out into weston_view The weston_surface structure is split into two structures: * The weston_surface structure storres everything required for a client-side or server-side surface. This includes buffers; callbacks; backend private data; input, damage, and opaque regions; and a few other bookkeeping bits. * The weston_view structure represents an entity in the scenegraph and storres all of the geometry information. This includes clip region, alpha, position, and the transformation list as well as all of the temporary information derived from the geometry state. Because a view, and not a surface, is a scenegraph element, the view is what is placed in layers and planes. There are a few things worth noting about the surface/view split: 1. This is *not* a modification to the protocol. It is, instead, a modification to Weston's internal scenegraph to allow a single surface to exist in multiple places at a time. Clients are completely unaware of how many views to a particular surface exist. 2. A view is considered a direct child of a surface and is destroyed when the surface is destroyed. Because of this, the view.surface pointer is always valid and non-null. 3. The compositor's surface_list is replaced with a view_list. Due to subsurfaces, building the view list is a little more complicated than it used to be and involves building a tree of views on the fly whenever subsurfaces are used. However, this means that backends can remain completely subsurface-agnostic. 4. Surfaces and views both keep track of which outputs they are on. 5. The weston_surface structure now has width and height fields. These are populated when a new buffer is attached before surface.configure is called. This is because there are many surface-based operations that really require the width and height and digging through the views didn't work well. Signed-off-by: Jason Ekstrand <jason@jlekstrand.net>
2013-10-13 07:38:11 +04:00
wl_list_for_each(sub, &surface->subsurface_list, parent_link) {
if (sub->surface == surface)
continue;
wl_list_insert_list(&sub->unused_views, &sub->surface->views);
wl_list_init(&sub->surface->views);
surface_stash_subsurface_views(sub->surface);
compositor: introduce sub-surfaces Implement the basic protocol for sub-surfaces: - expose wl_subcompositor global interface - error checking on protocol calls - associate a parent wl_surface to a sub-surface - introduce the sub-surface role, which is exclusive - an implementation of the wl_subsurface interface - allow nesting of sub-surfaces - proper surface transformation inheritance from parent to sub-surfaces - two different modes of wl_surface.commit for sub-surfaces - hook sub-surfaces up to repaint by modifying the repaint list code Struct weston_subsurface is dynamically allocated. For sub-surfaces, it is completely populated. For parent surfaces, weston_subsurface acts only as a link for stacking order purposes. The wl_resource is unused, parent_destroy_listener is not registered, the transform is not linked, etc. Sub-surfaces are not added directly into layers for display or input. Instead, they are hooked up via the sub-surface list present in parent weston_surface. This way sub-surfaces are inherently linked to the parent surface, and cannot be displayed unless the parent is mapped, too. This also eases restacking, as only the parent will be in a layer list. Also, only the main surface should be subject to shell actions. The surface list rebuilding in weston_output_repaint() is modified to process sub-surface lists, if they are non-empty. The sub-surface list always contains the parent, too, unless empty. The collection of frame_callback_list is moved to a later loop, to streamline the surface list rebuild functions. Features still lacking are: - full-surface alpha support for compound windows Changes in v2: - fix a bug in surface mapping: commit a sub-surface would cause the main surface to never be mapped. - remove debug printfs - detect attempt of making a surface its own parent - always zero-alloc weston_subsurface - apply wl_subsurface.set_position in commit, not immediately - add weston_surface_to_subsurface() - implement sub-surface commit modes parent-cached and independent - implement wl_subcompositor.destroy and wl_subsurface.destroy Changes in v3: - rebased, and use the new transform inheritance code - squashed the commit "add sub-surfaces to repaint list" - fixed a buffer reference leak in commit_from_cache() - Rewrite the sub-surface destructor code, and make it leave the wl_subsurface protocol object inert, if one destroys the corresponding wl_surface. - replaced set_commit_mode with set_sync and set_desync - allowed sub-surface nesting, and fixed repaint accordingly - implemented nested sub-surface commit modes - Made the sub-surface order changes from wl_subsurface.place_above and .place_below to be applied when the parent surface state is applied, instead of immediately. This conforms with the protocol specification now. Signed-off-by: Pekka Paalanen <ppaalanen@gmail.com>
2013-04-25 14:57:42 +04:00
}
Split the geometry information from weston_surface out into weston_view The weston_surface structure is split into two structures: * The weston_surface structure storres everything required for a client-side or server-side surface. This includes buffers; callbacks; backend private data; input, damage, and opaque regions; and a few other bookkeeping bits. * The weston_view structure represents an entity in the scenegraph and storres all of the geometry information. This includes clip region, alpha, position, and the transformation list as well as all of the temporary information derived from the geometry state. Because a view, and not a surface, is a scenegraph element, the view is what is placed in layers and planes. There are a few things worth noting about the surface/view split: 1. This is *not* a modification to the protocol. It is, instead, a modification to Weston's internal scenegraph to allow a single surface to exist in multiple places at a time. Clients are completely unaware of how many views to a particular surface exist. 2. A view is considered a direct child of a surface and is destroyed when the surface is destroyed. Because of this, the view.surface pointer is always valid and non-null. 3. The compositor's surface_list is replaced with a view_list. Due to subsurfaces, building the view list is a little more complicated than it used to be and involves building a tree of views on the fly whenever subsurfaces are used. However, this means that backends can remain completely subsurface-agnostic. 4. Surfaces and views both keep track of which outputs they are on. 5. The weston_surface structure now has width and height fields. These are populated when a new buffer is attached before surface.configure is called. This is because there are many surface-based operations that really require the width and height and digging through the views didn't work well. Signed-off-by: Jason Ekstrand <jason@jlekstrand.net>
2013-10-13 07:38:11 +04:00
}
static void
surface_free_unused_subsurface_views(struct weston_surface *surface)
{
struct weston_subsurface *sub;
struct weston_view *view, *nv;
compositor: introduce sub-surfaces Implement the basic protocol for sub-surfaces: - expose wl_subcompositor global interface - error checking on protocol calls - associate a parent wl_surface to a sub-surface - introduce the sub-surface role, which is exclusive - an implementation of the wl_subsurface interface - allow nesting of sub-surfaces - proper surface transformation inheritance from parent to sub-surfaces - two different modes of wl_surface.commit for sub-surfaces - hook sub-surfaces up to repaint by modifying the repaint list code Struct weston_subsurface is dynamically allocated. For sub-surfaces, it is completely populated. For parent surfaces, weston_subsurface acts only as a link for stacking order purposes. The wl_resource is unused, parent_destroy_listener is not registered, the transform is not linked, etc. Sub-surfaces are not added directly into layers for display or input. Instead, they are hooked up via the sub-surface list present in parent weston_surface. This way sub-surfaces are inherently linked to the parent surface, and cannot be displayed unless the parent is mapped, too. This also eases restacking, as only the parent will be in a layer list. Also, only the main surface should be subject to shell actions. The surface list rebuilding in weston_output_repaint() is modified to process sub-surface lists, if they are non-empty. The sub-surface list always contains the parent, too, unless empty. The collection of frame_callback_list is moved to a later loop, to streamline the surface list rebuild functions. Features still lacking are: - full-surface alpha support for compound windows Changes in v2: - fix a bug in surface mapping: commit a sub-surface would cause the main surface to never be mapped. - remove debug printfs - detect attempt of making a surface its own parent - always zero-alloc weston_subsurface - apply wl_subsurface.set_position in commit, not immediately - add weston_surface_to_subsurface() - implement sub-surface commit modes parent-cached and independent - implement wl_subcompositor.destroy and wl_subsurface.destroy Changes in v3: - rebased, and use the new transform inheritance code - squashed the commit "add sub-surfaces to repaint list" - fixed a buffer reference leak in commit_from_cache() - Rewrite the sub-surface destructor code, and make it leave the wl_subsurface protocol object inert, if one destroys the corresponding wl_surface. - replaced set_commit_mode with set_sync and set_desync - allowed sub-surface nesting, and fixed repaint accordingly - implemented nested sub-surface commit modes - Made the sub-surface order changes from wl_subsurface.place_above and .place_below to be applied when the parent surface state is applied, instead of immediately. This conforms with the protocol specification now. Signed-off-by: Pekka Paalanen <ppaalanen@gmail.com>
2013-04-25 14:57:42 +04:00
wl_list_for_each(sub, &surface->subsurface_list, parent_link) {
Split the geometry information from weston_surface out into weston_view The weston_surface structure is split into two structures: * The weston_surface structure storres everything required for a client-side or server-side surface. This includes buffers; callbacks; backend private data; input, damage, and opaque regions; and a few other bookkeeping bits. * The weston_view structure represents an entity in the scenegraph and storres all of the geometry information. This includes clip region, alpha, position, and the transformation list as well as all of the temporary information derived from the geometry state. Because a view, and not a surface, is a scenegraph element, the view is what is placed in layers and planes. There are a few things worth noting about the surface/view split: 1. This is *not* a modification to the protocol. It is, instead, a modification to Weston's internal scenegraph to allow a single surface to exist in multiple places at a time. Clients are completely unaware of how many views to a particular surface exist. 2. A view is considered a direct child of a surface and is destroyed when the surface is destroyed. Because of this, the view.surface pointer is always valid and non-null. 3. The compositor's surface_list is replaced with a view_list. Due to subsurfaces, building the view list is a little more complicated than it used to be and involves building a tree of views on the fly whenever subsurfaces are used. However, this means that backends can remain completely subsurface-agnostic. 4. Surfaces and views both keep track of which outputs they are on. 5. The weston_surface structure now has width and height fields. These are populated when a new buffer is attached before surface.configure is called. This is because there are many surface-based operations that really require the width and height and digging through the views didn't work well. Signed-off-by: Jason Ekstrand <jason@jlekstrand.net>
2013-10-13 07:38:11 +04:00
if (sub->surface == surface)
compositor: introduce sub-surfaces Implement the basic protocol for sub-surfaces: - expose wl_subcompositor global interface - error checking on protocol calls - associate a parent wl_surface to a sub-surface - introduce the sub-surface role, which is exclusive - an implementation of the wl_subsurface interface - allow nesting of sub-surfaces - proper surface transformation inheritance from parent to sub-surfaces - two different modes of wl_surface.commit for sub-surfaces - hook sub-surfaces up to repaint by modifying the repaint list code Struct weston_subsurface is dynamically allocated. For sub-surfaces, it is completely populated. For parent surfaces, weston_subsurface acts only as a link for stacking order purposes. The wl_resource is unused, parent_destroy_listener is not registered, the transform is not linked, etc. Sub-surfaces are not added directly into layers for display or input. Instead, they are hooked up via the sub-surface list present in parent weston_surface. This way sub-surfaces are inherently linked to the parent surface, and cannot be displayed unless the parent is mapped, too. This also eases restacking, as only the parent will be in a layer list. Also, only the main surface should be subject to shell actions. The surface list rebuilding in weston_output_repaint() is modified to process sub-surface lists, if they are non-empty. The sub-surface list always contains the parent, too, unless empty. The collection of frame_callback_list is moved to a later loop, to streamline the surface list rebuild functions. Features still lacking are: - full-surface alpha support for compound windows Changes in v2: - fix a bug in surface mapping: commit a sub-surface would cause the main surface to never be mapped. - remove debug printfs - detect attempt of making a surface its own parent - always zero-alloc weston_subsurface - apply wl_subsurface.set_position in commit, not immediately - add weston_surface_to_subsurface() - implement sub-surface commit modes parent-cached and independent - implement wl_subcompositor.destroy and wl_subsurface.destroy Changes in v3: - rebased, and use the new transform inheritance code - squashed the commit "add sub-surfaces to repaint list" - fixed a buffer reference leak in commit_from_cache() - Rewrite the sub-surface destructor code, and make it leave the wl_subsurface protocol object inert, if one destroys the corresponding wl_surface. - replaced set_commit_mode with set_sync and set_desync - allowed sub-surface nesting, and fixed repaint accordingly - implemented nested sub-surface commit modes - Made the sub-surface order changes from wl_subsurface.place_above and .place_below to be applied when the parent surface state is applied, instead of immediately. This conforms with the protocol specification now. Signed-off-by: Pekka Paalanen <ppaalanen@gmail.com>
2013-04-25 14:57:42 +04:00
continue;
wl_list_for_each_safe(view, nv, &sub->unused_views, surface_link) {
weston_view_unmap (view);
Split the geometry information from weston_surface out into weston_view The weston_surface structure is split into two structures: * The weston_surface structure storres everything required for a client-side or server-side surface. This includes buffers; callbacks; backend private data; input, damage, and opaque regions; and a few other bookkeeping bits. * The weston_view structure represents an entity in the scenegraph and storres all of the geometry information. This includes clip region, alpha, position, and the transformation list as well as all of the temporary information derived from the geometry state. Because a view, and not a surface, is a scenegraph element, the view is what is placed in layers and planes. There are a few things worth noting about the surface/view split: 1. This is *not* a modification to the protocol. It is, instead, a modification to Weston's internal scenegraph to allow a single surface to exist in multiple places at a time. Clients are completely unaware of how many views to a particular surface exist. 2. A view is considered a direct child of a surface and is destroyed when the surface is destroyed. Because of this, the view.surface pointer is always valid and non-null. 3. The compositor's surface_list is replaced with a view_list. Due to subsurfaces, building the view list is a little more complicated than it used to be and involves building a tree of views on the fly whenever subsurfaces are used. However, this means that backends can remain completely subsurface-agnostic. 4. Surfaces and views both keep track of which outputs they are on. 5. The weston_surface structure now has width and height fields. These are populated when a new buffer is attached before surface.configure is called. This is because there are many surface-based operations that really require the width and height and digging through the views didn't work well. Signed-off-by: Jason Ekstrand <jason@jlekstrand.net>
2013-10-13 07:38:11 +04:00
weston_view_destroy(view);
}
Split the geometry information from weston_surface out into weston_view The weston_surface structure is split into two structures: * The weston_surface structure storres everything required for a client-side or server-side surface. This includes buffers; callbacks; backend private data; input, damage, and opaque regions; and a few other bookkeeping bits. * The weston_view structure represents an entity in the scenegraph and storres all of the geometry information. This includes clip region, alpha, position, and the transformation list as well as all of the temporary information derived from the geometry state. Because a view, and not a surface, is a scenegraph element, the view is what is placed in layers and planes. There are a few things worth noting about the surface/view split: 1. This is *not* a modification to the protocol. It is, instead, a modification to Weston's internal scenegraph to allow a single surface to exist in multiple places at a time. Clients are completely unaware of how many views to a particular surface exist. 2. A view is considered a direct child of a surface and is destroyed when the surface is destroyed. Because of this, the view.surface pointer is always valid and non-null. 3. The compositor's surface_list is replaced with a view_list. Due to subsurfaces, building the view list is a little more complicated than it used to be and involves building a tree of views on the fly whenever subsurfaces are used. However, this means that backends can remain completely subsurface-agnostic. 4. Surfaces and views both keep track of which outputs they are on. 5. The weston_surface structure now has width and height fields. These are populated when a new buffer is attached before surface.configure is called. This is because there are many surface-based operations that really require the width and height and digging through the views didn't work well. Signed-off-by: Jason Ekstrand <jason@jlekstrand.net>
2013-10-13 07:38:11 +04:00
surface_free_unused_subsurface_views(sub->surface);
}
}
static void
view_list_add_subsurface_view(struct weston_compositor *compositor,
struct weston_subsurface *sub,
struct weston_view *parent)
{
struct weston_subsurface *child;
struct weston_view *view = NULL, *iv;
2014-07-28 13:49:24 +04:00
if (!weston_surface_is_mapped(sub->surface))
return;
Split the geometry information from weston_surface out into weston_view The weston_surface structure is split into two structures: * The weston_surface structure storres everything required for a client-side or server-side surface. This includes buffers; callbacks; backend private data; input, damage, and opaque regions; and a few other bookkeeping bits. * The weston_view structure represents an entity in the scenegraph and storres all of the geometry information. This includes clip region, alpha, position, and the transformation list as well as all of the temporary information derived from the geometry state. Because a view, and not a surface, is a scenegraph element, the view is what is placed in layers and planes. There are a few things worth noting about the surface/view split: 1. This is *not* a modification to the protocol. It is, instead, a modification to Weston's internal scenegraph to allow a single surface to exist in multiple places at a time. Clients are completely unaware of how many views to a particular surface exist. 2. A view is considered a direct child of a surface and is destroyed when the surface is destroyed. Because of this, the view.surface pointer is always valid and non-null. 3. The compositor's surface_list is replaced with a view_list. Due to subsurfaces, building the view list is a little more complicated than it used to be and involves building a tree of views on the fly whenever subsurfaces are used. However, this means that backends can remain completely subsurface-agnostic. 4. Surfaces and views both keep track of which outputs they are on. 5. The weston_surface structure now has width and height fields. These are populated when a new buffer is attached before surface.configure is called. This is because there are many surface-based operations that really require the width and height and digging through the views didn't work well. Signed-off-by: Jason Ekstrand <jason@jlekstrand.net>
2013-10-13 07:38:11 +04:00
wl_list_for_each(iv, &sub->unused_views, surface_link) {
if (iv->geometry.parent == parent) {
view = iv;
break;
compositor: introduce sub-surfaces Implement the basic protocol for sub-surfaces: - expose wl_subcompositor global interface - error checking on protocol calls - associate a parent wl_surface to a sub-surface - introduce the sub-surface role, which is exclusive - an implementation of the wl_subsurface interface - allow nesting of sub-surfaces - proper surface transformation inheritance from parent to sub-surfaces - two different modes of wl_surface.commit for sub-surfaces - hook sub-surfaces up to repaint by modifying the repaint list code Struct weston_subsurface is dynamically allocated. For sub-surfaces, it is completely populated. For parent surfaces, weston_subsurface acts only as a link for stacking order purposes. The wl_resource is unused, parent_destroy_listener is not registered, the transform is not linked, etc. Sub-surfaces are not added directly into layers for display or input. Instead, they are hooked up via the sub-surface list present in parent weston_surface. This way sub-surfaces are inherently linked to the parent surface, and cannot be displayed unless the parent is mapped, too. This also eases restacking, as only the parent will be in a layer list. Also, only the main surface should be subject to shell actions. The surface list rebuilding in weston_output_repaint() is modified to process sub-surface lists, if they are non-empty. The sub-surface list always contains the parent, too, unless empty. The collection of frame_callback_list is moved to a later loop, to streamline the surface list rebuild functions. Features still lacking are: - full-surface alpha support for compound windows Changes in v2: - fix a bug in surface mapping: commit a sub-surface would cause the main surface to never be mapped. - remove debug printfs - detect attempt of making a surface its own parent - always zero-alloc weston_subsurface - apply wl_subsurface.set_position in commit, not immediately - add weston_surface_to_subsurface() - implement sub-surface commit modes parent-cached and independent - implement wl_subcompositor.destroy and wl_subsurface.destroy Changes in v3: - rebased, and use the new transform inheritance code - squashed the commit "add sub-surfaces to repaint list" - fixed a buffer reference leak in commit_from_cache() - Rewrite the sub-surface destructor code, and make it leave the wl_subsurface protocol object inert, if one destroys the corresponding wl_surface. - replaced set_commit_mode with set_sync and set_desync - allowed sub-surface nesting, and fixed repaint accordingly - implemented nested sub-surface commit modes - Made the sub-surface order changes from wl_subsurface.place_above and .place_below to be applied when the parent surface state is applied, instead of immediately. This conforms with the protocol specification now. Signed-off-by: Pekka Paalanen <ppaalanen@gmail.com>
2013-04-25 14:57:42 +04:00
}
}
Split the geometry information from weston_surface out into weston_view The weston_surface structure is split into two structures: * The weston_surface structure storres everything required for a client-side or server-side surface. This includes buffers; callbacks; backend private data; input, damage, and opaque regions; and a few other bookkeeping bits. * The weston_view structure represents an entity in the scenegraph and storres all of the geometry information. This includes clip region, alpha, position, and the transformation list as well as all of the temporary information derived from the geometry state. Because a view, and not a surface, is a scenegraph element, the view is what is placed in layers and planes. There are a few things worth noting about the surface/view split: 1. This is *not* a modification to the protocol. It is, instead, a modification to Weston's internal scenegraph to allow a single surface to exist in multiple places at a time. Clients are completely unaware of how many views to a particular surface exist. 2. A view is considered a direct child of a surface and is destroyed when the surface is destroyed. Because of this, the view.surface pointer is always valid and non-null. 3. The compositor's surface_list is replaced with a view_list. Due to subsurfaces, building the view list is a little more complicated than it used to be and involves building a tree of views on the fly whenever subsurfaces are used. However, this means that backends can remain completely subsurface-agnostic. 4. Surfaces and views both keep track of which outputs they are on. 5. The weston_surface structure now has width and height fields. These are populated when a new buffer is attached before surface.configure is called. This is because there are many surface-based operations that really require the width and height and digging through the views didn't work well. Signed-off-by: Jason Ekstrand <jason@jlekstrand.net>
2013-10-13 07:38:11 +04:00
if (view) {
/* Put it back in the surface's list of views */
wl_list_remove(&view->surface_link);
wl_list_insert(&sub->surface->views, &view->surface_link);
} else {
view = weston_view_create(sub->surface);
weston_view_set_position(view,
sub->position.x,
sub->position.y);
Split the geometry information from weston_surface out into weston_view The weston_surface structure is split into two structures: * The weston_surface structure storres everything required for a client-side or server-side surface. This includes buffers; callbacks; backend private data; input, damage, and opaque regions; and a few other bookkeeping bits. * The weston_view structure represents an entity in the scenegraph and storres all of the geometry information. This includes clip region, alpha, position, and the transformation list as well as all of the temporary information derived from the geometry state. Because a view, and not a surface, is a scenegraph element, the view is what is placed in layers and planes. There are a few things worth noting about the surface/view split: 1. This is *not* a modification to the protocol. It is, instead, a modification to Weston's internal scenegraph to allow a single surface to exist in multiple places at a time. Clients are completely unaware of how many views to a particular surface exist. 2. A view is considered a direct child of a surface and is destroyed when the surface is destroyed. Because of this, the view.surface pointer is always valid and non-null. 3. The compositor's surface_list is replaced with a view_list. Due to subsurfaces, building the view list is a little more complicated than it used to be and involves building a tree of views on the fly whenever subsurfaces are used. However, this means that backends can remain completely subsurface-agnostic. 4. Surfaces and views both keep track of which outputs they are on. 5. The weston_surface structure now has width and height fields. These are populated when a new buffer is attached before surface.configure is called. This is because there are many surface-based operations that really require the width and height and digging through the views didn't work well. Signed-off-by: Jason Ekstrand <jason@jlekstrand.net>
2013-10-13 07:38:11 +04:00
weston_view_set_transform_parent(view, parent);
}
view->parent_view = parent;
Split the geometry information from weston_surface out into weston_view The weston_surface structure is split into two structures: * The weston_surface structure storres everything required for a client-side or server-side surface. This includes buffers; callbacks; backend private data; input, damage, and opaque regions; and a few other bookkeeping bits. * The weston_view structure represents an entity in the scenegraph and storres all of the geometry information. This includes clip region, alpha, position, and the transformation list as well as all of the temporary information derived from the geometry state. Because a view, and not a surface, is a scenegraph element, the view is what is placed in layers and planes. There are a few things worth noting about the surface/view split: 1. This is *not* a modification to the protocol. It is, instead, a modification to Weston's internal scenegraph to allow a single surface to exist in multiple places at a time. Clients are completely unaware of how many views to a particular surface exist. 2. A view is considered a direct child of a surface and is destroyed when the surface is destroyed. Because of this, the view.surface pointer is always valid and non-null. 3. The compositor's surface_list is replaced with a view_list. Due to subsurfaces, building the view list is a little more complicated than it used to be and involves building a tree of views on the fly whenever subsurfaces are used. However, this means that backends can remain completely subsurface-agnostic. 4. Surfaces and views both keep track of which outputs they are on. 5. The weston_surface structure now has width and height fields. These are populated when a new buffer is attached before surface.configure is called. This is because there are many surface-based operations that really require the width and height and digging through the views didn't work well. Signed-off-by: Jason Ekstrand <jason@jlekstrand.net>
2013-10-13 07:38:11 +04:00
weston_view_update_transform(view);
view->is_mapped = true;
Split the geometry information from weston_surface out into weston_view The weston_surface structure is split into two structures: * The weston_surface structure storres everything required for a client-side or server-side surface. This includes buffers; callbacks; backend private data; input, damage, and opaque regions; and a few other bookkeeping bits. * The weston_view structure represents an entity in the scenegraph and storres all of the geometry information. This includes clip region, alpha, position, and the transformation list as well as all of the temporary information derived from the geometry state. Because a view, and not a surface, is a scenegraph element, the view is what is placed in layers and planes. There are a few things worth noting about the surface/view split: 1. This is *not* a modification to the protocol. It is, instead, a modification to Weston's internal scenegraph to allow a single surface to exist in multiple places at a time. Clients are completely unaware of how many views to a particular surface exist. 2. A view is considered a direct child of a surface and is destroyed when the surface is destroyed. Because of this, the view.surface pointer is always valid and non-null. 3. The compositor's surface_list is replaced with a view_list. Due to subsurfaces, building the view list is a little more complicated than it used to be and involves building a tree of views on the fly whenever subsurfaces are used. However, this means that backends can remain completely subsurface-agnostic. 4. Surfaces and views both keep track of which outputs they are on. 5. The weston_surface structure now has width and height fields. These are populated when a new buffer is attached before surface.configure is called. This is because there are many surface-based operations that really require the width and height and digging through the views didn't work well. Signed-off-by: Jason Ekstrand <jason@jlekstrand.net>
2013-10-13 07:38:11 +04:00
if (wl_list_empty(&sub->surface->subsurface_list)) {
wl_list_insert(compositor->view_list.prev, &view->link);
return;
}
wl_list_for_each(child, &sub->surface->subsurface_list, parent_link) {
if (child->surface == sub->surface)
wl_list_insert(compositor->view_list.prev, &view->link);
else
Split the geometry information from weston_surface out into weston_view The weston_surface structure is split into two structures: * The weston_surface structure storres everything required for a client-side or server-side surface. This includes buffers; callbacks; backend private data; input, damage, and opaque regions; and a few other bookkeeping bits. * The weston_view structure represents an entity in the scenegraph and storres all of the geometry information. This includes clip region, alpha, position, and the transformation list as well as all of the temporary information derived from the geometry state. Because a view, and not a surface, is a scenegraph element, the view is what is placed in layers and planes. There are a few things worth noting about the surface/view split: 1. This is *not* a modification to the protocol. It is, instead, a modification to Weston's internal scenegraph to allow a single surface to exist in multiple places at a time. Clients are completely unaware of how many views to a particular surface exist. 2. A view is considered a direct child of a surface and is destroyed when the surface is destroyed. Because of this, the view.surface pointer is always valid and non-null. 3. The compositor's surface_list is replaced with a view_list. Due to subsurfaces, building the view list is a little more complicated than it used to be and involves building a tree of views on the fly whenever subsurfaces are used. However, this means that backends can remain completely subsurface-agnostic. 4. Surfaces and views both keep track of which outputs they are on. 5. The weston_surface structure now has width and height fields. These are populated when a new buffer is attached before surface.configure is called. This is because there are many surface-based operations that really require the width and height and digging through the views didn't work well. Signed-off-by: Jason Ekstrand <jason@jlekstrand.net>
2013-10-13 07:38:11 +04:00
view_list_add_subsurface_view(compositor, child, view);
}
compositor: introduce sub-surfaces Implement the basic protocol for sub-surfaces: - expose wl_subcompositor global interface - error checking on protocol calls - associate a parent wl_surface to a sub-surface - introduce the sub-surface role, which is exclusive - an implementation of the wl_subsurface interface - allow nesting of sub-surfaces - proper surface transformation inheritance from parent to sub-surfaces - two different modes of wl_surface.commit for sub-surfaces - hook sub-surfaces up to repaint by modifying the repaint list code Struct weston_subsurface is dynamically allocated. For sub-surfaces, it is completely populated. For parent surfaces, weston_subsurface acts only as a link for stacking order purposes. The wl_resource is unused, parent_destroy_listener is not registered, the transform is not linked, etc. Sub-surfaces are not added directly into layers for display or input. Instead, they are hooked up via the sub-surface list present in parent weston_surface. This way sub-surfaces are inherently linked to the parent surface, and cannot be displayed unless the parent is mapped, too. This also eases restacking, as only the parent will be in a layer list. Also, only the main surface should be subject to shell actions. The surface list rebuilding in weston_output_repaint() is modified to process sub-surface lists, if they are non-empty. The sub-surface list always contains the parent, too, unless empty. The collection of frame_callback_list is moved to a later loop, to streamline the surface list rebuild functions. Features still lacking are: - full-surface alpha support for compound windows Changes in v2: - fix a bug in surface mapping: commit a sub-surface would cause the main surface to never be mapped. - remove debug printfs - detect attempt of making a surface its own parent - always zero-alloc weston_subsurface - apply wl_subsurface.set_position in commit, not immediately - add weston_surface_to_subsurface() - implement sub-surface commit modes parent-cached and independent - implement wl_subcompositor.destroy and wl_subsurface.destroy Changes in v3: - rebased, and use the new transform inheritance code - squashed the commit "add sub-surfaces to repaint list" - fixed a buffer reference leak in commit_from_cache() - Rewrite the sub-surface destructor code, and make it leave the wl_subsurface protocol object inert, if one destroys the corresponding wl_surface. - replaced set_commit_mode with set_sync and set_desync - allowed sub-surface nesting, and fixed repaint accordingly - implemented nested sub-surface commit modes - Made the sub-surface order changes from wl_subsurface.place_above and .place_below to be applied when the parent surface state is applied, instead of immediately. This conforms with the protocol specification now. Signed-off-by: Pekka Paalanen <ppaalanen@gmail.com>
2013-04-25 14:57:42 +04:00
}
/* This recursively adds the sub-surfaces for a view, relying on the
* sub-surface order. Thus, if a client restacks the sub-surfaces, that
* change first happens to the sub-surface list, and then automatically
* propagates here. See weston_surface_damage_subsurfaces() for how the
* sub-surfaces receive damage when the client changes the state.
*/
compositor: introduce sub-surfaces Implement the basic protocol for sub-surfaces: - expose wl_subcompositor global interface - error checking on protocol calls - associate a parent wl_surface to a sub-surface - introduce the sub-surface role, which is exclusive - an implementation of the wl_subsurface interface - allow nesting of sub-surfaces - proper surface transformation inheritance from parent to sub-surfaces - two different modes of wl_surface.commit for sub-surfaces - hook sub-surfaces up to repaint by modifying the repaint list code Struct weston_subsurface is dynamically allocated. For sub-surfaces, it is completely populated. For parent surfaces, weston_subsurface acts only as a link for stacking order purposes. The wl_resource is unused, parent_destroy_listener is not registered, the transform is not linked, etc. Sub-surfaces are not added directly into layers for display or input. Instead, they are hooked up via the sub-surface list present in parent weston_surface. This way sub-surfaces are inherently linked to the parent surface, and cannot be displayed unless the parent is mapped, too. This also eases restacking, as only the parent will be in a layer list. Also, only the main surface should be subject to shell actions. The surface list rebuilding in weston_output_repaint() is modified to process sub-surface lists, if they are non-empty. The sub-surface list always contains the parent, too, unless empty. The collection of frame_callback_list is moved to a later loop, to streamline the surface list rebuild functions. Features still lacking are: - full-surface alpha support for compound windows Changes in v2: - fix a bug in surface mapping: commit a sub-surface would cause the main surface to never be mapped. - remove debug printfs - detect attempt of making a surface its own parent - always zero-alloc weston_subsurface - apply wl_subsurface.set_position in commit, not immediately - add weston_surface_to_subsurface() - implement sub-surface commit modes parent-cached and independent - implement wl_subcompositor.destroy and wl_subsurface.destroy Changes in v3: - rebased, and use the new transform inheritance code - squashed the commit "add sub-surfaces to repaint list" - fixed a buffer reference leak in commit_from_cache() - Rewrite the sub-surface destructor code, and make it leave the wl_subsurface protocol object inert, if one destroys the corresponding wl_surface. - replaced set_commit_mode with set_sync and set_desync - allowed sub-surface nesting, and fixed repaint accordingly - implemented nested sub-surface commit modes - Made the sub-surface order changes from wl_subsurface.place_above and .place_below to be applied when the parent surface state is applied, instead of immediately. This conforms with the protocol specification now. Signed-off-by: Pekka Paalanen <ppaalanen@gmail.com>
2013-04-25 14:57:42 +04:00
static void
Split the geometry information from weston_surface out into weston_view The weston_surface structure is split into two structures: * The weston_surface structure storres everything required for a client-side or server-side surface. This includes buffers; callbacks; backend private data; input, damage, and opaque regions; and a few other bookkeeping bits. * The weston_view structure represents an entity in the scenegraph and storres all of the geometry information. This includes clip region, alpha, position, and the transformation list as well as all of the temporary information derived from the geometry state. Because a view, and not a surface, is a scenegraph element, the view is what is placed in layers and planes. There are a few things worth noting about the surface/view split: 1. This is *not* a modification to the protocol. It is, instead, a modification to Weston's internal scenegraph to allow a single surface to exist in multiple places at a time. Clients are completely unaware of how many views to a particular surface exist. 2. A view is considered a direct child of a surface and is destroyed when the surface is destroyed. Because of this, the view.surface pointer is always valid and non-null. 3. The compositor's surface_list is replaced with a view_list. Due to subsurfaces, building the view list is a little more complicated than it used to be and involves building a tree of views on the fly whenever subsurfaces are used. However, this means that backends can remain completely subsurface-agnostic. 4. Surfaces and views both keep track of which outputs they are on. 5. The weston_surface structure now has width and height fields. These are populated when a new buffer is attached before surface.configure is called. This is because there are many surface-based operations that really require the width and height and digging through the views didn't work well. Signed-off-by: Jason Ekstrand <jason@jlekstrand.net>
2013-10-13 07:38:11 +04:00
view_list_add(struct weston_compositor *compositor,
struct weston_view *view)
compositor: introduce sub-surfaces Implement the basic protocol for sub-surfaces: - expose wl_subcompositor global interface - error checking on protocol calls - associate a parent wl_surface to a sub-surface - introduce the sub-surface role, which is exclusive - an implementation of the wl_subsurface interface - allow nesting of sub-surfaces - proper surface transformation inheritance from parent to sub-surfaces - two different modes of wl_surface.commit for sub-surfaces - hook sub-surfaces up to repaint by modifying the repaint list code Struct weston_subsurface is dynamically allocated. For sub-surfaces, it is completely populated. For parent surfaces, weston_subsurface acts only as a link for stacking order purposes. The wl_resource is unused, parent_destroy_listener is not registered, the transform is not linked, etc. Sub-surfaces are not added directly into layers for display or input. Instead, they are hooked up via the sub-surface list present in parent weston_surface. This way sub-surfaces are inherently linked to the parent surface, and cannot be displayed unless the parent is mapped, too. This also eases restacking, as only the parent will be in a layer list. Also, only the main surface should be subject to shell actions. The surface list rebuilding in weston_output_repaint() is modified to process sub-surface lists, if they are non-empty. The sub-surface list always contains the parent, too, unless empty. The collection of frame_callback_list is moved to a later loop, to streamline the surface list rebuild functions. Features still lacking are: - full-surface alpha support for compound windows Changes in v2: - fix a bug in surface mapping: commit a sub-surface would cause the main surface to never be mapped. - remove debug printfs - detect attempt of making a surface its own parent - always zero-alloc weston_subsurface - apply wl_subsurface.set_position in commit, not immediately - add weston_surface_to_subsurface() - implement sub-surface commit modes parent-cached and independent - implement wl_subcompositor.destroy and wl_subsurface.destroy Changes in v3: - rebased, and use the new transform inheritance code - squashed the commit "add sub-surfaces to repaint list" - fixed a buffer reference leak in commit_from_cache() - Rewrite the sub-surface destructor code, and make it leave the wl_subsurface protocol object inert, if one destroys the corresponding wl_surface. - replaced set_commit_mode with set_sync and set_desync - allowed sub-surface nesting, and fixed repaint accordingly - implemented nested sub-surface commit modes - Made the sub-surface order changes from wl_subsurface.place_above and .place_below to be applied when the parent surface state is applied, instead of immediately. This conforms with the protocol specification now. Signed-off-by: Pekka Paalanen <ppaalanen@gmail.com>
2013-04-25 14:57:42 +04:00
{
Split the geometry information from weston_surface out into weston_view The weston_surface structure is split into two structures: * The weston_surface structure storres everything required for a client-side or server-side surface. This includes buffers; callbacks; backend private data; input, damage, and opaque regions; and a few other bookkeeping bits. * The weston_view structure represents an entity in the scenegraph and storres all of the geometry information. This includes clip region, alpha, position, and the transformation list as well as all of the temporary information derived from the geometry state. Because a view, and not a surface, is a scenegraph element, the view is what is placed in layers and planes. There are a few things worth noting about the surface/view split: 1. This is *not* a modification to the protocol. It is, instead, a modification to Weston's internal scenegraph to allow a single surface to exist in multiple places at a time. Clients are completely unaware of how many views to a particular surface exist. 2. A view is considered a direct child of a surface and is destroyed when the surface is destroyed. Because of this, the view.surface pointer is always valid and non-null. 3. The compositor's surface_list is replaced with a view_list. Due to subsurfaces, building the view list is a little more complicated than it used to be and involves building a tree of views on the fly whenever subsurfaces are used. However, this means that backends can remain completely subsurface-agnostic. 4. Surfaces and views both keep track of which outputs they are on. 5. The weston_surface structure now has width and height fields. These are populated when a new buffer is attached before surface.configure is called. This is because there are many surface-based operations that really require the width and height and digging through the views didn't work well. Signed-off-by: Jason Ekstrand <jason@jlekstrand.net>
2013-10-13 07:38:11 +04:00
struct weston_subsurface *sub;
weston_view_update_transform(view);
if (wl_list_empty(&view->surface->subsurface_list)) {
wl_list_insert(compositor->view_list.prev, &view->link);
return;
}
wl_list_for_each(sub, &view->surface->subsurface_list, parent_link) {
if (sub->surface == view->surface)
wl_list_insert(compositor->view_list.prev, &view->link);
else
Split the geometry information from weston_surface out into weston_view The weston_surface structure is split into two structures: * The weston_surface structure storres everything required for a client-side or server-side surface. This includes buffers; callbacks; backend private data; input, damage, and opaque regions; and a few other bookkeeping bits. * The weston_view structure represents an entity in the scenegraph and storres all of the geometry information. This includes clip region, alpha, position, and the transformation list as well as all of the temporary information derived from the geometry state. Because a view, and not a surface, is a scenegraph element, the view is what is placed in layers and planes. There are a few things worth noting about the surface/view split: 1. This is *not* a modification to the protocol. It is, instead, a modification to Weston's internal scenegraph to allow a single surface to exist in multiple places at a time. Clients are completely unaware of how many views to a particular surface exist. 2. A view is considered a direct child of a surface and is destroyed when the surface is destroyed. Because of this, the view.surface pointer is always valid and non-null. 3. The compositor's surface_list is replaced with a view_list. Due to subsurfaces, building the view list is a little more complicated than it used to be and involves building a tree of views on the fly whenever subsurfaces are used. However, this means that backends can remain completely subsurface-agnostic. 4. Surfaces and views both keep track of which outputs they are on. 5. The weston_surface structure now has width and height fields. These are populated when a new buffer is attached before surface.configure is called. This is because there are many surface-based operations that really require the width and height and digging through the views didn't work well. Signed-off-by: Jason Ekstrand <jason@jlekstrand.net>
2013-10-13 07:38:11 +04:00
view_list_add_subsurface_view(compositor, sub, view);
}
Split the geometry information from weston_surface out into weston_view The weston_surface structure is split into two structures: * The weston_surface structure storres everything required for a client-side or server-side surface. This includes buffers; callbacks; backend private data; input, damage, and opaque regions; and a few other bookkeeping bits. * The weston_view structure represents an entity in the scenegraph and storres all of the geometry information. This includes clip region, alpha, position, and the transformation list as well as all of the temporary information derived from the geometry state. Because a view, and not a surface, is a scenegraph element, the view is what is placed in layers and planes. There are a few things worth noting about the surface/view split: 1. This is *not* a modification to the protocol. It is, instead, a modification to Weston's internal scenegraph to allow a single surface to exist in multiple places at a time. Clients are completely unaware of how many views to a particular surface exist. 2. A view is considered a direct child of a surface and is destroyed when the surface is destroyed. Because of this, the view.surface pointer is always valid and non-null. 3. The compositor's surface_list is replaced with a view_list. Due to subsurfaces, building the view list is a little more complicated than it used to be and involves building a tree of views on the fly whenever subsurfaces are used. However, this means that backends can remain completely subsurface-agnostic. 4. Surfaces and views both keep track of which outputs they are on. 5. The weston_surface structure now has width and height fields. These are populated when a new buffer is attached before surface.configure is called. This is because there are many surface-based operations that really require the width and height and digging through the views didn't work well. Signed-off-by: Jason Ekstrand <jason@jlekstrand.net>
2013-10-13 07:38:11 +04:00
}
static void
weston_compositor_build_view_list(struct weston_compositor *compositor)
{
struct weston_view *view;
compositor: introduce sub-surfaces Implement the basic protocol for sub-surfaces: - expose wl_subcompositor global interface - error checking on protocol calls - associate a parent wl_surface to a sub-surface - introduce the sub-surface role, which is exclusive - an implementation of the wl_subsurface interface - allow nesting of sub-surfaces - proper surface transformation inheritance from parent to sub-surfaces - two different modes of wl_surface.commit for sub-surfaces - hook sub-surfaces up to repaint by modifying the repaint list code Struct weston_subsurface is dynamically allocated. For sub-surfaces, it is completely populated. For parent surfaces, weston_subsurface acts only as a link for stacking order purposes. The wl_resource is unused, parent_destroy_listener is not registered, the transform is not linked, etc. Sub-surfaces are not added directly into layers for display or input. Instead, they are hooked up via the sub-surface list present in parent weston_surface. This way sub-surfaces are inherently linked to the parent surface, and cannot be displayed unless the parent is mapped, too. This also eases restacking, as only the parent will be in a layer list. Also, only the main surface should be subject to shell actions. The surface list rebuilding in weston_output_repaint() is modified to process sub-surface lists, if they are non-empty. The sub-surface list always contains the parent, too, unless empty. The collection of frame_callback_list is moved to a later loop, to streamline the surface list rebuild functions. Features still lacking are: - full-surface alpha support for compound windows Changes in v2: - fix a bug in surface mapping: commit a sub-surface would cause the main surface to never be mapped. - remove debug printfs - detect attempt of making a surface its own parent - always zero-alloc weston_subsurface - apply wl_subsurface.set_position in commit, not immediately - add weston_surface_to_subsurface() - implement sub-surface commit modes parent-cached and independent - implement wl_subcompositor.destroy and wl_subsurface.destroy Changes in v3: - rebased, and use the new transform inheritance code - squashed the commit "add sub-surfaces to repaint list" - fixed a buffer reference leak in commit_from_cache() - Rewrite the sub-surface destructor code, and make it leave the wl_subsurface protocol object inert, if one destroys the corresponding wl_surface. - replaced set_commit_mode with set_sync and set_desync - allowed sub-surface nesting, and fixed repaint accordingly - implemented nested sub-surface commit modes - Made the sub-surface order changes from wl_subsurface.place_above and .place_below to be applied when the parent surface state is applied, instead of immediately. This conforms with the protocol specification now. Signed-off-by: Pekka Paalanen <ppaalanen@gmail.com>
2013-04-25 14:57:42 +04:00
struct weston_layer *layer;
Split the geometry information from weston_surface out into weston_view The weston_surface structure is split into two structures: * The weston_surface structure storres everything required for a client-side or server-side surface. This includes buffers; callbacks; backend private data; input, damage, and opaque regions; and a few other bookkeeping bits. * The weston_view structure represents an entity in the scenegraph and storres all of the geometry information. This includes clip region, alpha, position, and the transformation list as well as all of the temporary information derived from the geometry state. Because a view, and not a surface, is a scenegraph element, the view is what is placed in layers and planes. There are a few things worth noting about the surface/view split: 1. This is *not* a modification to the protocol. It is, instead, a modification to Weston's internal scenegraph to allow a single surface to exist in multiple places at a time. Clients are completely unaware of how many views to a particular surface exist. 2. A view is considered a direct child of a surface and is destroyed when the surface is destroyed. Because of this, the view.surface pointer is always valid and non-null. 3. The compositor's surface_list is replaced with a view_list. Due to subsurfaces, building the view list is a little more complicated than it used to be and involves building a tree of views on the fly whenever subsurfaces are used. However, this means that backends can remain completely subsurface-agnostic. 4. Surfaces and views both keep track of which outputs they are on. 5. The weston_surface structure now has width and height fields. These are populated when a new buffer is attached before surface.configure is called. This is because there are many surface-based operations that really require the width and height and digging through the views didn't work well. Signed-off-by: Jason Ekstrand <jason@jlekstrand.net>
2013-10-13 07:38:11 +04:00
wl_list_for_each(layer, &compositor->layer_list, link)
wl_list_for_each(view, &layer->view_list.link, layer_link.link)
Split the geometry information from weston_surface out into weston_view The weston_surface structure is split into two structures: * The weston_surface structure storres everything required for a client-side or server-side surface. This includes buffers; callbacks; backend private data; input, damage, and opaque regions; and a few other bookkeeping bits. * The weston_view structure represents an entity in the scenegraph and storres all of the geometry information. This includes clip region, alpha, position, and the transformation list as well as all of the temporary information derived from the geometry state. Because a view, and not a surface, is a scenegraph element, the view is what is placed in layers and planes. There are a few things worth noting about the surface/view split: 1. This is *not* a modification to the protocol. It is, instead, a modification to Weston's internal scenegraph to allow a single surface to exist in multiple places at a time. Clients are completely unaware of how many views to a particular surface exist. 2. A view is considered a direct child of a surface and is destroyed when the surface is destroyed. Because of this, the view.surface pointer is always valid and non-null. 3. The compositor's surface_list is replaced with a view_list. Due to subsurfaces, building the view list is a little more complicated than it used to be and involves building a tree of views on the fly whenever subsurfaces are used. However, this means that backends can remain completely subsurface-agnostic. 4. Surfaces and views both keep track of which outputs they are on. 5. The weston_surface structure now has width and height fields. These are populated when a new buffer is attached before surface.configure is called. This is because there are many surface-based operations that really require the width and height and digging through the views didn't work well. Signed-off-by: Jason Ekstrand <jason@jlekstrand.net>
2013-10-13 07:38:11 +04:00
surface_stash_subsurface_views(view->surface);
wl_list_init(&compositor->view_list);
compositor: introduce sub-surfaces Implement the basic protocol for sub-surfaces: - expose wl_subcompositor global interface - error checking on protocol calls - associate a parent wl_surface to a sub-surface - introduce the sub-surface role, which is exclusive - an implementation of the wl_subsurface interface - allow nesting of sub-surfaces - proper surface transformation inheritance from parent to sub-surfaces - two different modes of wl_surface.commit for sub-surfaces - hook sub-surfaces up to repaint by modifying the repaint list code Struct weston_subsurface is dynamically allocated. For sub-surfaces, it is completely populated. For parent surfaces, weston_subsurface acts only as a link for stacking order purposes. The wl_resource is unused, parent_destroy_listener is not registered, the transform is not linked, etc. Sub-surfaces are not added directly into layers for display or input. Instead, they are hooked up via the sub-surface list present in parent weston_surface. This way sub-surfaces are inherently linked to the parent surface, and cannot be displayed unless the parent is mapped, too. This also eases restacking, as only the parent will be in a layer list. Also, only the main surface should be subject to shell actions. The surface list rebuilding in weston_output_repaint() is modified to process sub-surface lists, if they are non-empty. The sub-surface list always contains the parent, too, unless empty. The collection of frame_callback_list is moved to a later loop, to streamline the surface list rebuild functions. Features still lacking are: - full-surface alpha support for compound windows Changes in v2: - fix a bug in surface mapping: commit a sub-surface would cause the main surface to never be mapped. - remove debug printfs - detect attempt of making a surface its own parent - always zero-alloc weston_subsurface - apply wl_subsurface.set_position in commit, not immediately - add weston_surface_to_subsurface() - implement sub-surface commit modes parent-cached and independent - implement wl_subcompositor.destroy and wl_subsurface.destroy Changes in v3: - rebased, and use the new transform inheritance code - squashed the commit "add sub-surfaces to repaint list" - fixed a buffer reference leak in commit_from_cache() - Rewrite the sub-surface destructor code, and make it leave the wl_subsurface protocol object inert, if one destroys the corresponding wl_surface. - replaced set_commit_mode with set_sync and set_desync - allowed sub-surface nesting, and fixed repaint accordingly - implemented nested sub-surface commit modes - Made the sub-surface order changes from wl_subsurface.place_above and .place_below to be applied when the parent surface state is applied, instead of immediately. This conforms with the protocol specification now. Signed-off-by: Pekka Paalanen <ppaalanen@gmail.com>
2013-04-25 14:57:42 +04:00
wl_list_for_each(layer, &compositor->layer_list, link) {
wl_list_for_each(view, &layer->view_list.link, layer_link.link) {
Split the geometry information from weston_surface out into weston_view The weston_surface structure is split into two structures: * The weston_surface structure storres everything required for a client-side or server-side surface. This includes buffers; callbacks; backend private data; input, damage, and opaque regions; and a few other bookkeeping bits. * The weston_view structure represents an entity in the scenegraph and storres all of the geometry information. This includes clip region, alpha, position, and the transformation list as well as all of the temporary information derived from the geometry state. Because a view, and not a surface, is a scenegraph element, the view is what is placed in layers and planes. There are a few things worth noting about the surface/view split: 1. This is *not* a modification to the protocol. It is, instead, a modification to Weston's internal scenegraph to allow a single surface to exist in multiple places at a time. Clients are completely unaware of how many views to a particular surface exist. 2. A view is considered a direct child of a surface and is destroyed when the surface is destroyed. Because of this, the view.surface pointer is always valid and non-null. 3. The compositor's surface_list is replaced with a view_list. Due to subsurfaces, building the view list is a little more complicated than it used to be and involves building a tree of views on the fly whenever subsurfaces are used. However, this means that backends can remain completely subsurface-agnostic. 4. Surfaces and views both keep track of which outputs they are on. 5. The weston_surface structure now has width and height fields. These are populated when a new buffer is attached before surface.configure is called. This is because there are many surface-based operations that really require the width and height and digging through the views didn't work well. Signed-off-by: Jason Ekstrand <jason@jlekstrand.net>
2013-10-13 07:38:11 +04:00
view_list_add(compositor, view);
compositor: introduce sub-surfaces Implement the basic protocol for sub-surfaces: - expose wl_subcompositor global interface - error checking on protocol calls - associate a parent wl_surface to a sub-surface - introduce the sub-surface role, which is exclusive - an implementation of the wl_subsurface interface - allow nesting of sub-surfaces - proper surface transformation inheritance from parent to sub-surfaces - two different modes of wl_surface.commit for sub-surfaces - hook sub-surfaces up to repaint by modifying the repaint list code Struct weston_subsurface is dynamically allocated. For sub-surfaces, it is completely populated. For parent surfaces, weston_subsurface acts only as a link for stacking order purposes. The wl_resource is unused, parent_destroy_listener is not registered, the transform is not linked, etc. Sub-surfaces are not added directly into layers for display or input. Instead, they are hooked up via the sub-surface list present in parent weston_surface. This way sub-surfaces are inherently linked to the parent surface, and cannot be displayed unless the parent is mapped, too. This also eases restacking, as only the parent will be in a layer list. Also, only the main surface should be subject to shell actions. The surface list rebuilding in weston_output_repaint() is modified to process sub-surface lists, if they are non-empty. The sub-surface list always contains the parent, too, unless empty. The collection of frame_callback_list is moved to a later loop, to streamline the surface list rebuild functions. Features still lacking are: - full-surface alpha support for compound windows Changes in v2: - fix a bug in surface mapping: commit a sub-surface would cause the main surface to never be mapped. - remove debug printfs - detect attempt of making a surface its own parent - always zero-alloc weston_subsurface - apply wl_subsurface.set_position in commit, not immediately - add weston_surface_to_subsurface() - implement sub-surface commit modes parent-cached and independent - implement wl_subcompositor.destroy and wl_subsurface.destroy Changes in v3: - rebased, and use the new transform inheritance code - squashed the commit "add sub-surfaces to repaint list" - fixed a buffer reference leak in commit_from_cache() - Rewrite the sub-surface destructor code, and make it leave the wl_subsurface protocol object inert, if one destroys the corresponding wl_surface. - replaced set_commit_mode with set_sync and set_desync - allowed sub-surface nesting, and fixed repaint accordingly - implemented nested sub-surface commit modes - Made the sub-surface order changes from wl_subsurface.place_above and .place_below to be applied when the parent surface state is applied, instead of immediately. This conforms with the protocol specification now. Signed-off-by: Pekka Paalanen <ppaalanen@gmail.com>
2013-04-25 14:57:42 +04:00
}
}
Split the geometry information from weston_surface out into weston_view The weston_surface structure is split into two structures: * The weston_surface structure storres everything required for a client-side or server-side surface. This includes buffers; callbacks; backend private data; input, damage, and opaque regions; and a few other bookkeeping bits. * The weston_view structure represents an entity in the scenegraph and storres all of the geometry information. This includes clip region, alpha, position, and the transformation list as well as all of the temporary information derived from the geometry state. Because a view, and not a surface, is a scenegraph element, the view is what is placed in layers and planes. There are a few things worth noting about the surface/view split: 1. This is *not* a modification to the protocol. It is, instead, a modification to Weston's internal scenegraph to allow a single surface to exist in multiple places at a time. Clients are completely unaware of how many views to a particular surface exist. 2. A view is considered a direct child of a surface and is destroyed when the surface is destroyed. Because of this, the view.surface pointer is always valid and non-null. 3. The compositor's surface_list is replaced with a view_list. Due to subsurfaces, building the view list is a little more complicated than it used to be and involves building a tree of views on the fly whenever subsurfaces are used. However, this means that backends can remain completely subsurface-agnostic. 4. Surfaces and views both keep track of which outputs they are on. 5. The weston_surface structure now has width and height fields. These are populated when a new buffer is attached before surface.configure is called. This is because there are many surface-based operations that really require the width and height and digging through the views didn't work well. Signed-off-by: Jason Ekstrand <jason@jlekstrand.net>
2013-10-13 07:38:11 +04:00
wl_list_for_each(layer, &compositor->layer_list, link)
wl_list_for_each(view, &layer->view_list.link, layer_link.link)
Split the geometry information from weston_surface out into weston_view The weston_surface structure is split into two structures: * The weston_surface structure storres everything required for a client-side or server-side surface. This includes buffers; callbacks; backend private data; input, damage, and opaque regions; and a few other bookkeeping bits. * The weston_view structure represents an entity in the scenegraph and storres all of the geometry information. This includes clip region, alpha, position, and the transformation list as well as all of the temporary information derived from the geometry state. Because a view, and not a surface, is a scenegraph element, the view is what is placed in layers and planes. There are a few things worth noting about the surface/view split: 1. This is *not* a modification to the protocol. It is, instead, a modification to Weston's internal scenegraph to allow a single surface to exist in multiple places at a time. Clients are completely unaware of how many views to a particular surface exist. 2. A view is considered a direct child of a surface and is destroyed when the surface is destroyed. Because of this, the view.surface pointer is always valid and non-null. 3. The compositor's surface_list is replaced with a view_list. Due to subsurfaces, building the view list is a little more complicated than it used to be and involves building a tree of views on the fly whenever subsurfaces are used. However, this means that backends can remain completely subsurface-agnostic. 4. Surfaces and views both keep track of which outputs they are on. 5. The weston_surface structure now has width and height fields. These are populated when a new buffer is attached before surface.configure is called. This is because there are many surface-based operations that really require the width and height and digging through the views didn't work well. Signed-off-by: Jason Ekstrand <jason@jlekstrand.net>
2013-10-13 07:38:11 +04:00
surface_free_unused_subsurface_views(view->surface);
compositor: introduce sub-surfaces Implement the basic protocol for sub-surfaces: - expose wl_subcompositor global interface - error checking on protocol calls - associate a parent wl_surface to a sub-surface - introduce the sub-surface role, which is exclusive - an implementation of the wl_subsurface interface - allow nesting of sub-surfaces - proper surface transformation inheritance from parent to sub-surfaces - two different modes of wl_surface.commit for sub-surfaces - hook sub-surfaces up to repaint by modifying the repaint list code Struct weston_subsurface is dynamically allocated. For sub-surfaces, it is completely populated. For parent surfaces, weston_subsurface acts only as a link for stacking order purposes. The wl_resource is unused, parent_destroy_listener is not registered, the transform is not linked, etc. Sub-surfaces are not added directly into layers for display or input. Instead, they are hooked up via the sub-surface list present in parent weston_surface. This way sub-surfaces are inherently linked to the parent surface, and cannot be displayed unless the parent is mapped, too. This also eases restacking, as only the parent will be in a layer list. Also, only the main surface should be subject to shell actions. The surface list rebuilding in weston_output_repaint() is modified to process sub-surface lists, if they are non-empty. The sub-surface list always contains the parent, too, unless empty. The collection of frame_callback_list is moved to a later loop, to streamline the surface list rebuild functions. Features still lacking are: - full-surface alpha support for compound windows Changes in v2: - fix a bug in surface mapping: commit a sub-surface would cause the main surface to never be mapped. - remove debug printfs - detect attempt of making a surface its own parent - always zero-alloc weston_subsurface - apply wl_subsurface.set_position in commit, not immediately - add weston_surface_to_subsurface() - implement sub-surface commit modes parent-cached and independent - implement wl_subcompositor.destroy and wl_subsurface.destroy Changes in v3: - rebased, and use the new transform inheritance code - squashed the commit "add sub-surfaces to repaint list" - fixed a buffer reference leak in commit_from_cache() - Rewrite the sub-surface destructor code, and make it leave the wl_subsurface protocol object inert, if one destroys the corresponding wl_surface. - replaced set_commit_mode with set_sync and set_desync - allowed sub-surface nesting, and fixed repaint accordingly - implemented nested sub-surface commit modes - Made the sub-surface order changes from wl_subsurface.place_above and .place_below to be applied when the parent surface state is applied, instead of immediately. This conforms with the protocol specification now. Signed-off-by: Pekka Paalanen <ppaalanen@gmail.com>
2013-04-25 14:57:42 +04:00
}
static void
weston_output_take_feedback_list(struct weston_output *output,
struct weston_surface *surface)
{
struct weston_view *view;
struct weston_presentation_feedback *feedback;
uint32_t flags = 0xffffffff;
if (wl_list_empty(&surface->feedback_list))
return;
/* All views must have the flag for the flag to survive. */
wl_list_for_each(view, &surface->views, surface_link) {
/* ignore views that are not on this output at all */
if (view->output_mask & (1u << output->id))
flags &= view->psf_flags;
}
wl_list_for_each(feedback, &surface->feedback_list, link)
feedback->psf_flags = flags;
wl_list_insert_list(&output->feedback_list, &surface->feedback_list);
wl_list_init(&surface->feedback_list);
}
static int
weston_output_repaint(struct weston_output *output)
{
struct weston_compositor *ec = output->compositor;
Split the geometry information from weston_surface out into weston_view The weston_surface structure is split into two structures: * The weston_surface structure storres everything required for a client-side or server-side surface. This includes buffers; callbacks; backend private data; input, damage, and opaque regions; and a few other bookkeeping bits. * The weston_view structure represents an entity in the scenegraph and storres all of the geometry information. This includes clip region, alpha, position, and the transformation list as well as all of the temporary information derived from the geometry state. Because a view, and not a surface, is a scenegraph element, the view is what is placed in layers and planes. There are a few things worth noting about the surface/view split: 1. This is *not* a modification to the protocol. It is, instead, a modification to Weston's internal scenegraph to allow a single surface to exist in multiple places at a time. Clients are completely unaware of how many views to a particular surface exist. 2. A view is considered a direct child of a surface and is destroyed when the surface is destroyed. Because of this, the view.surface pointer is always valid and non-null. 3. The compositor's surface_list is replaced with a view_list. Due to subsurfaces, building the view list is a little more complicated than it used to be and involves building a tree of views on the fly whenever subsurfaces are used. However, this means that backends can remain completely subsurface-agnostic. 4. Surfaces and views both keep track of which outputs they are on. 5. The weston_surface structure now has width and height fields. These are populated when a new buffer is attached before surface.configure is called. This is because there are many surface-based operations that really require the width and height and digging through the views didn't work well. Signed-off-by: Jason Ekstrand <jason@jlekstrand.net>
2013-10-13 07:38:11 +04:00
struct weston_view *ev;
struct weston_animation *animation, *next;
struct weston_frame_callback *cb, *cnext;
struct wl_list frame_callback_list;
pixman_region32_t output_damage;
int r;
if (output->destroying)
return 0;
compositor: Implement JSON-timeline logging Logging is activated and deactivated with the debug key binding 't'. When activated, it creates a new log file, where it records the events. The log file contains events and detailed object information entries in JSON format, and is meant to be parsed in sequence from beginning to the end. The emitted events are mostly related to the output repaint cycle, like when repaint begins, is submitted to GPU, and when it completes on a vblank. This is recorded per-output. Also some per-surface events are recorded, including when surface damage is flushed. To reduce the log size, events refer to objects like outputs and surfaces by id numbers. Detailed object information is emitted only as needed: on the first object occurrence, and afterwards only if weston_timeline_object::force_refresh asks for it. The detailed information for surfaces includes the string returned by weston_surface::get_label. Therefore it is important to set weston_timeline_object::force_refresh = 1 whenever the string would change, so that the new details get recorded. A rudimentary parser and SVG generator can be found at: https://github.com/ppaalanen/wesgr The timeline logs can answer questions including: - How does the compositor repaint cycle work timing-wise? - When was the vblank deadline missed? - What is the latency from surface commit to showing the new content on screen? - How long does it take to process the scenegraph? v2: weston_surface::get_description renamed to get_label. v3: reafctor a bit into fprint_quoted_string(). Signed-off-by: Pekka Paalanen <pekka.paalanen@collabora.co.uk>
2014-11-12 16:09:24 +03:00
TL_POINT("core_repaint_begin", TLP_OUTPUT(output), TLP_END);
/* Rebuild the surface list and update surface transforms up front. */
Split the geometry information from weston_surface out into weston_view The weston_surface structure is split into two structures: * The weston_surface structure storres everything required for a client-side or server-side surface. This includes buffers; callbacks; backend private data; input, damage, and opaque regions; and a few other bookkeeping bits. * The weston_view structure represents an entity in the scenegraph and storres all of the geometry information. This includes clip region, alpha, position, and the transformation list as well as all of the temporary information derived from the geometry state. Because a view, and not a surface, is a scenegraph element, the view is what is placed in layers and planes. There are a few things worth noting about the surface/view split: 1. This is *not* a modification to the protocol. It is, instead, a modification to Weston's internal scenegraph to allow a single surface to exist in multiple places at a time. Clients are completely unaware of how many views to a particular surface exist. 2. A view is considered a direct child of a surface and is destroyed when the surface is destroyed. Because of this, the view.surface pointer is always valid and non-null. 3. The compositor's surface_list is replaced with a view_list. Due to subsurfaces, building the view list is a little more complicated than it used to be and involves building a tree of views on the fly whenever subsurfaces are used. However, this means that backends can remain completely subsurface-agnostic. 4. Surfaces and views both keep track of which outputs they are on. 5. The weston_surface structure now has width and height fields. These are populated when a new buffer is attached before surface.configure is called. This is because there are many surface-based operations that really require the width and height and digging through the views didn't work well. Signed-off-by: Jason Ekstrand <jason@jlekstrand.net>
2013-10-13 07:38:11 +04:00
weston_compositor_build_view_list(ec);
if (output->assign_planes && !output->disable_planes) {
output->assign_planes(output);
} else {
wl_list_for_each(ev, &ec->view_list, link) {
Split the geometry information from weston_surface out into weston_view The weston_surface structure is split into two structures: * The weston_surface structure storres everything required for a client-side or server-side surface. This includes buffers; callbacks; backend private data; input, damage, and opaque regions; and a few other bookkeeping bits. * The weston_view structure represents an entity in the scenegraph and storres all of the geometry information. This includes clip region, alpha, position, and the transformation list as well as all of the temporary information derived from the geometry state. Because a view, and not a surface, is a scenegraph element, the view is what is placed in layers and planes. There are a few things worth noting about the surface/view split: 1. This is *not* a modification to the protocol. It is, instead, a modification to Weston's internal scenegraph to allow a single surface to exist in multiple places at a time. Clients are completely unaware of how many views to a particular surface exist. 2. A view is considered a direct child of a surface and is destroyed when the surface is destroyed. Because of this, the view.surface pointer is always valid and non-null. 3. The compositor's surface_list is replaced with a view_list. Due to subsurfaces, building the view list is a little more complicated than it used to be and involves building a tree of views on the fly whenever subsurfaces are used. However, this means that backends can remain completely subsurface-agnostic. 4. Surfaces and views both keep track of which outputs they are on. 5. The weston_surface structure now has width and height fields. These are populated when a new buffer is attached before surface.configure is called. This is because there are many surface-based operations that really require the width and height and digging through the views didn't work well. Signed-off-by: Jason Ekstrand <jason@jlekstrand.net>
2013-10-13 07:38:11 +04:00
weston_view_move_to_plane(ev, &ec->primary_plane);
ev->psf_flags = 0;
}
}
compositor: introduce sub-surfaces Implement the basic protocol for sub-surfaces: - expose wl_subcompositor global interface - error checking on protocol calls - associate a parent wl_surface to a sub-surface - introduce the sub-surface role, which is exclusive - an implementation of the wl_subsurface interface - allow nesting of sub-surfaces - proper surface transformation inheritance from parent to sub-surfaces - two different modes of wl_surface.commit for sub-surfaces - hook sub-surfaces up to repaint by modifying the repaint list code Struct weston_subsurface is dynamically allocated. For sub-surfaces, it is completely populated. For parent surfaces, weston_subsurface acts only as a link for stacking order purposes. The wl_resource is unused, parent_destroy_listener is not registered, the transform is not linked, etc. Sub-surfaces are not added directly into layers for display or input. Instead, they are hooked up via the sub-surface list present in parent weston_surface. This way sub-surfaces are inherently linked to the parent surface, and cannot be displayed unless the parent is mapped, too. This also eases restacking, as only the parent will be in a layer list. Also, only the main surface should be subject to shell actions. The surface list rebuilding in weston_output_repaint() is modified to process sub-surface lists, if they are non-empty. The sub-surface list always contains the parent, too, unless empty. The collection of frame_callback_list is moved to a later loop, to streamline the surface list rebuild functions. Features still lacking are: - full-surface alpha support for compound windows Changes in v2: - fix a bug in surface mapping: commit a sub-surface would cause the main surface to never be mapped. - remove debug printfs - detect attempt of making a surface its own parent - always zero-alloc weston_subsurface - apply wl_subsurface.set_position in commit, not immediately - add weston_surface_to_subsurface() - implement sub-surface commit modes parent-cached and independent - implement wl_subcompositor.destroy and wl_subsurface.destroy Changes in v3: - rebased, and use the new transform inheritance code - squashed the commit "add sub-surfaces to repaint list" - fixed a buffer reference leak in commit_from_cache() - Rewrite the sub-surface destructor code, and make it leave the wl_subsurface protocol object inert, if one destroys the corresponding wl_surface. - replaced set_commit_mode with set_sync and set_desync - allowed sub-surface nesting, and fixed repaint accordingly - implemented nested sub-surface commit modes - Made the sub-surface order changes from wl_subsurface.place_above and .place_below to be applied when the parent surface state is applied, instead of immediately. This conforms with the protocol specification now. Signed-off-by: Pekka Paalanen <ppaalanen@gmail.com>
2013-04-25 14:57:42 +04:00
wl_list_init(&frame_callback_list);
Split the geometry information from weston_surface out into weston_view The weston_surface structure is split into two structures: * The weston_surface structure storres everything required for a client-side or server-side surface. This includes buffers; callbacks; backend private data; input, damage, and opaque regions; and a few other bookkeeping bits. * The weston_view structure represents an entity in the scenegraph and storres all of the geometry information. This includes clip region, alpha, position, and the transformation list as well as all of the temporary information derived from the geometry state. Because a view, and not a surface, is a scenegraph element, the view is what is placed in layers and planes. There are a few things worth noting about the surface/view split: 1. This is *not* a modification to the protocol. It is, instead, a modification to Weston's internal scenegraph to allow a single surface to exist in multiple places at a time. Clients are completely unaware of how many views to a particular surface exist. 2. A view is considered a direct child of a surface and is destroyed when the surface is destroyed. Because of this, the view.surface pointer is always valid and non-null. 3. The compositor's surface_list is replaced with a view_list. Due to subsurfaces, building the view list is a little more complicated than it used to be and involves building a tree of views on the fly whenever subsurfaces are used. However, this means that backends can remain completely subsurface-agnostic. 4. Surfaces and views both keep track of which outputs they are on. 5. The weston_surface structure now has width and height fields. These are populated when a new buffer is attached before surface.configure is called. This is because there are many surface-based operations that really require the width and height and digging through the views didn't work well. Signed-off-by: Jason Ekstrand <jason@jlekstrand.net>
2013-10-13 07:38:11 +04:00
wl_list_for_each(ev, &ec->view_list, link) {
/* Note: This operation is safe to do multiple times on the
* same surface.
*/
if (ev->surface->output == output) {
compositor: introduce sub-surfaces Implement the basic protocol for sub-surfaces: - expose wl_subcompositor global interface - error checking on protocol calls - associate a parent wl_surface to a sub-surface - introduce the sub-surface role, which is exclusive - an implementation of the wl_subsurface interface - allow nesting of sub-surfaces - proper surface transformation inheritance from parent to sub-surfaces - two different modes of wl_surface.commit for sub-surfaces - hook sub-surfaces up to repaint by modifying the repaint list code Struct weston_subsurface is dynamically allocated. For sub-surfaces, it is completely populated. For parent surfaces, weston_subsurface acts only as a link for stacking order purposes. The wl_resource is unused, parent_destroy_listener is not registered, the transform is not linked, etc. Sub-surfaces are not added directly into layers for display or input. Instead, they are hooked up via the sub-surface list present in parent weston_surface. This way sub-surfaces are inherently linked to the parent surface, and cannot be displayed unless the parent is mapped, too. This also eases restacking, as only the parent will be in a layer list. Also, only the main surface should be subject to shell actions. The surface list rebuilding in weston_output_repaint() is modified to process sub-surface lists, if they are non-empty. The sub-surface list always contains the parent, too, unless empty. The collection of frame_callback_list is moved to a later loop, to streamline the surface list rebuild functions. Features still lacking are: - full-surface alpha support for compound windows Changes in v2: - fix a bug in surface mapping: commit a sub-surface would cause the main surface to never be mapped. - remove debug printfs - detect attempt of making a surface its own parent - always zero-alloc weston_subsurface - apply wl_subsurface.set_position in commit, not immediately - add weston_surface_to_subsurface() - implement sub-surface commit modes parent-cached and independent - implement wl_subcompositor.destroy and wl_subsurface.destroy Changes in v3: - rebased, and use the new transform inheritance code - squashed the commit "add sub-surfaces to repaint list" - fixed a buffer reference leak in commit_from_cache() - Rewrite the sub-surface destructor code, and make it leave the wl_subsurface protocol object inert, if one destroys the corresponding wl_surface. - replaced set_commit_mode with set_sync and set_desync - allowed sub-surface nesting, and fixed repaint accordingly - implemented nested sub-surface commit modes - Made the sub-surface order changes from wl_subsurface.place_above and .place_below to be applied when the parent surface state is applied, instead of immediately. This conforms with the protocol specification now. Signed-off-by: Pekka Paalanen <ppaalanen@gmail.com>
2013-04-25 14:57:42 +04:00
wl_list_insert_list(&frame_callback_list,
Split the geometry information from weston_surface out into weston_view The weston_surface structure is split into two structures: * The weston_surface structure storres everything required for a client-side or server-side surface. This includes buffers; callbacks; backend private data; input, damage, and opaque regions; and a few other bookkeeping bits. * The weston_view structure represents an entity in the scenegraph and storres all of the geometry information. This includes clip region, alpha, position, and the transformation list as well as all of the temporary information derived from the geometry state. Because a view, and not a surface, is a scenegraph element, the view is what is placed in layers and planes. There are a few things worth noting about the surface/view split: 1. This is *not* a modification to the protocol. It is, instead, a modification to Weston's internal scenegraph to allow a single surface to exist in multiple places at a time. Clients are completely unaware of how many views to a particular surface exist. 2. A view is considered a direct child of a surface and is destroyed when the surface is destroyed. Because of this, the view.surface pointer is always valid and non-null. 3. The compositor's surface_list is replaced with a view_list. Due to subsurfaces, building the view list is a little more complicated than it used to be and involves building a tree of views on the fly whenever subsurfaces are used. However, this means that backends can remain completely subsurface-agnostic. 4. Surfaces and views both keep track of which outputs they are on. 5. The weston_surface structure now has width and height fields. These are populated when a new buffer is attached before surface.configure is called. This is because there are many surface-based operations that really require the width and height and digging through the views didn't work well. Signed-off-by: Jason Ekstrand <jason@jlekstrand.net>
2013-10-13 07:38:11 +04:00
&ev->surface->frame_callback_list);
wl_list_init(&ev->surface->frame_callback_list);
weston_output_take_feedback_list(output, ev->surface);
compositor: introduce sub-surfaces Implement the basic protocol for sub-surfaces: - expose wl_subcompositor global interface - error checking on protocol calls - associate a parent wl_surface to a sub-surface - introduce the sub-surface role, which is exclusive - an implementation of the wl_subsurface interface - allow nesting of sub-surfaces - proper surface transformation inheritance from parent to sub-surfaces - two different modes of wl_surface.commit for sub-surfaces - hook sub-surfaces up to repaint by modifying the repaint list code Struct weston_subsurface is dynamically allocated. For sub-surfaces, it is completely populated. For parent surfaces, weston_subsurface acts only as a link for stacking order purposes. The wl_resource is unused, parent_destroy_listener is not registered, the transform is not linked, etc. Sub-surfaces are not added directly into layers for display or input. Instead, they are hooked up via the sub-surface list present in parent weston_surface. This way sub-surfaces are inherently linked to the parent surface, and cannot be displayed unless the parent is mapped, too. This also eases restacking, as only the parent will be in a layer list. Also, only the main surface should be subject to shell actions. The surface list rebuilding in weston_output_repaint() is modified to process sub-surface lists, if they are non-empty. The sub-surface list always contains the parent, too, unless empty. The collection of frame_callback_list is moved to a later loop, to streamline the surface list rebuild functions. Features still lacking are: - full-surface alpha support for compound windows Changes in v2: - fix a bug in surface mapping: commit a sub-surface would cause the main surface to never be mapped. - remove debug printfs - detect attempt of making a surface its own parent - always zero-alloc weston_subsurface - apply wl_subsurface.set_position in commit, not immediately - add weston_surface_to_subsurface() - implement sub-surface commit modes parent-cached and independent - implement wl_subcompositor.destroy and wl_subsurface.destroy Changes in v3: - rebased, and use the new transform inheritance code - squashed the commit "add sub-surfaces to repaint list" - fixed a buffer reference leak in commit_from_cache() - Rewrite the sub-surface destructor code, and make it leave the wl_subsurface protocol object inert, if one destroys the corresponding wl_surface. - replaced set_commit_mode with set_sync and set_desync - allowed sub-surface nesting, and fixed repaint accordingly - implemented nested sub-surface commit modes - Made the sub-surface order changes from wl_subsurface.place_above and .place_below to be applied when the parent surface state is applied, instead of immediately. This conforms with the protocol specification now. Signed-off-by: Pekka Paalanen <ppaalanen@gmail.com>
2013-04-25 14:57:42 +04:00
}
}
compositor_accumulate_damage(ec);
pixman_region32_init(&output_damage);
pixman_region32_intersect(&output_damage,
&ec->primary_plane.damage, &output->region);
pixman_region32_subtract(&output_damage,
&output_damage, &ec->primary_plane.clip);
if (output->dirty)
weston_output_update_matrix(output);
r = output->repaint(output, &output_damage);
pixman_region32_fini(&output_damage);
output->repaint_needed = 0;
weston_compositor_repick(ec);
wl_list_for_each_safe(cb, cnext, &frame_callback_list, link) {
wl_callback_send_done(cb->resource, output->frame_time);
wl_resource_destroy(cb->resource);
}
2012-06-09 05:40:54 +04:00
wl_list_for_each_safe(animation, next, &output->animation_list, link) {
animation->frame_counter++;
animation->frame(animation, output, output->frame_time);
2012-06-09 05:40:54 +04:00
}
compositor: Implement JSON-timeline logging Logging is activated and deactivated with the debug key binding 't'. When activated, it creates a new log file, where it records the events. The log file contains events and detailed object information entries in JSON format, and is meant to be parsed in sequence from beginning to the end. The emitted events are mostly related to the output repaint cycle, like when repaint begins, is submitted to GPU, and when it completes on a vblank. This is recorded per-output. Also some per-surface events are recorded, including when surface damage is flushed. To reduce the log size, events refer to objects like outputs and surfaces by id numbers. Detailed object information is emitted only as needed: on the first object occurrence, and afterwards only if weston_timeline_object::force_refresh asks for it. The detailed information for surfaces includes the string returned by weston_surface::get_label. Therefore it is important to set weston_timeline_object::force_refresh = 1 whenever the string would change, so that the new details get recorded. A rudimentary parser and SVG generator can be found at: https://github.com/ppaalanen/wesgr The timeline logs can answer questions including: - How does the compositor repaint cycle work timing-wise? - When was the vblank deadline missed? - What is the latency from surface commit to showing the new content on screen? - How long does it take to process the scenegraph? v2: weston_surface::get_description renamed to get_label. v3: reafctor a bit into fprint_quoted_string(). Signed-off-by: Pekka Paalanen <pekka.paalanen@collabora.co.uk>
2014-11-12 16:09:24 +03:00
TL_POINT("core_repaint_posted", TLP_OUTPUT(output), TLP_END);
return r;
}
static void
weston_output_schedule_repaint_reset(struct weston_output *output)
{
output->repaint_scheduled = 0;
TL_POINT("core_repaint_exit_loop", TLP_OUTPUT(output), TLP_END);
}
compositor: add repaint delay timer This timer delays the output_repaint towards the end of the refresh period, reducing the time from repaint to present. The length of the repaint window can be set in weston.ini. The call to weston_output_schedule_repaint_reset() is delayed by one more period. If we exit the continuous repaint loop (set output->repaint_scheduled to false) in finish_frame, we may call start_repaint_loop() unnecessarily. The problem case was actually observed with two outputs on the DRM backend at 60 Hz, and 7 ms repaint-window. During a window move, one output was constantly falling off the continuous repaint loop and introducing additional one frame latency, leading to jerky window motion. This code now avoids the problem. Changes in v2: - Rename repaint_delay_timer to repaint_timer and output_repaint_delay_handler to output_repaint_timer_handler. - When computing the delay, take the current time into account. The timer uses a relative timeout, so we have to subtract any time already gone. Note, that 'gone' may also be negative. DRM has a habit of predicting the page flip timestamp so it may be still in the future when we get the completion event. - Do also a sanity check 'msec > 1000'. In the unlikely case that something fails to provide a good timestamp, never delay for more than one second. Signed-off-by: Pekka Paalanen <pekka.paalanen@collabora.co.uk> Reviewed-By: Derek Foreman <derekf@osg.samsung.com> Reviewed-by: Bryce Harrington <bryce@osg.samsung.com>
2014-05-21 17:17:27 +04:00
static int
output_repaint_timer_handler(void *data)
{
struct weston_output *output = data;
struct weston_compositor *compositor = output->compositor;
int ret;
compositor: add repaint delay timer This timer delays the output_repaint towards the end of the refresh period, reducing the time from repaint to present. The length of the repaint window can be set in weston.ini. The call to weston_output_schedule_repaint_reset() is delayed by one more period. If we exit the continuous repaint loop (set output->repaint_scheduled to false) in finish_frame, we may call start_repaint_loop() unnecessarily. The problem case was actually observed with two outputs on the DRM backend at 60 Hz, and 7 ms repaint-window. During a window move, one output was constantly falling off the continuous repaint loop and introducing additional one frame latency, leading to jerky window motion. This code now avoids the problem. Changes in v2: - Rename repaint_delay_timer to repaint_timer and output_repaint_delay_handler to output_repaint_timer_handler. - When computing the delay, take the current time into account. The timer uses a relative timeout, so we have to subtract any time already gone. Note, that 'gone' may also be negative. DRM has a habit of predicting the page flip timestamp so it may be still in the future when we get the completion event. - Do also a sanity check 'msec > 1000'. In the unlikely case that something fails to provide a good timestamp, never delay for more than one second. Signed-off-by: Pekka Paalanen <pekka.paalanen@collabora.co.uk> Reviewed-By: Derek Foreman <derekf@osg.samsung.com> Reviewed-by: Bryce Harrington <bryce@osg.samsung.com>
2014-05-21 17:17:27 +04:00
/* If we're sleeping, drop the repaint machinery entirely; we will
* explicitly repaint all outputs when we come back. */
if (compositor->state == WESTON_COMPOSITOR_SLEEPING ||
compositor->state == WESTON_COMPOSITOR_OFFSCREEN)
goto err;
compositor: add repaint delay timer This timer delays the output_repaint towards the end of the refresh period, reducing the time from repaint to present. The length of the repaint window can be set in weston.ini. The call to weston_output_schedule_repaint_reset() is delayed by one more period. If we exit the continuous repaint loop (set output->repaint_scheduled to false) in finish_frame, we may call start_repaint_loop() unnecessarily. The problem case was actually observed with two outputs on the DRM backend at 60 Hz, and 7 ms repaint-window. During a window move, one output was constantly falling off the continuous repaint loop and introducing additional one frame latency, leading to jerky window motion. This code now avoids the problem. Changes in v2: - Rename repaint_delay_timer to repaint_timer and output_repaint_delay_handler to output_repaint_timer_handler. - When computing the delay, take the current time into account. The timer uses a relative timeout, so we have to subtract any time already gone. Note, that 'gone' may also be negative. DRM has a habit of predicting the page flip timestamp so it may be still in the future when we get the completion event. - Do also a sanity check 'msec > 1000'. In the unlikely case that something fails to provide a good timestamp, never delay for more than one second. Signed-off-by: Pekka Paalanen <pekka.paalanen@collabora.co.uk> Reviewed-By: Derek Foreman <derekf@osg.samsung.com> Reviewed-by: Bryce Harrington <bryce@osg.samsung.com>
2014-05-21 17:17:27 +04:00
/* We don't actually need to repaint this output; drop it from
* repaint until something causes damage. */
if (!output->repaint_needed)
goto err;
/* If repaint fails, we aren't going to get weston_output_finish_frame
* to trigger a new repaint, so drop it from repaint and hope
* something schedules a successful repaint later. */
ret = weston_output_repaint(output);
if (ret != 0)
goto err;
compositor: add repaint delay timer This timer delays the output_repaint towards the end of the refresh period, reducing the time from repaint to present. The length of the repaint window can be set in weston.ini. The call to weston_output_schedule_repaint_reset() is delayed by one more period. If we exit the continuous repaint loop (set output->repaint_scheduled to false) in finish_frame, we may call start_repaint_loop() unnecessarily. The problem case was actually observed with two outputs on the DRM backend at 60 Hz, and 7 ms repaint-window. During a window move, one output was constantly falling off the continuous repaint loop and introducing additional one frame latency, leading to jerky window motion. This code now avoids the problem. Changes in v2: - Rename repaint_delay_timer to repaint_timer and output_repaint_delay_handler to output_repaint_timer_handler. - When computing the delay, take the current time into account. The timer uses a relative timeout, so we have to subtract any time already gone. Note, that 'gone' may also be negative. DRM has a habit of predicting the page flip timestamp so it may be still in the future when we get the completion event. - Do also a sanity check 'msec > 1000'. In the unlikely case that something fails to provide a good timestamp, never delay for more than one second. Signed-off-by: Pekka Paalanen <pekka.paalanen@collabora.co.uk> Reviewed-By: Derek Foreman <derekf@osg.samsung.com> Reviewed-by: Bryce Harrington <bryce@osg.samsung.com>
2014-05-21 17:17:27 +04:00
return 0;
err:
weston_output_schedule_repaint_reset(output);
return 0;
compositor: add repaint delay timer This timer delays the output_repaint towards the end of the refresh period, reducing the time from repaint to present. The length of the repaint window can be set in weston.ini. The call to weston_output_schedule_repaint_reset() is delayed by one more period. If we exit the continuous repaint loop (set output->repaint_scheduled to false) in finish_frame, we may call start_repaint_loop() unnecessarily. The problem case was actually observed with two outputs on the DRM backend at 60 Hz, and 7 ms repaint-window. During a window move, one output was constantly falling off the continuous repaint loop and introducing additional one frame latency, leading to jerky window motion. This code now avoids the problem. Changes in v2: - Rename repaint_delay_timer to repaint_timer and output_repaint_delay_handler to output_repaint_timer_handler. - When computing the delay, take the current time into account. The timer uses a relative timeout, so we have to subtract any time already gone. Note, that 'gone' may also be negative. DRM has a habit of predicting the page flip timestamp so it may be still in the future when we get the completion event. - Do also a sanity check 'msec > 1000'. In the unlikely case that something fails to provide a good timestamp, never delay for more than one second. Signed-off-by: Pekka Paalanen <pekka.paalanen@collabora.co.uk> Reviewed-By: Derek Foreman <derekf@osg.samsung.com> Reviewed-by: Bryce Harrington <bryce@osg.samsung.com>
2014-05-21 17:17:27 +04:00
}
WL_EXPORT void
weston_output_finish_frame(struct weston_output *output,
compositor: set presentation.presented flags Change weston_output_finish_frame() signature so that backends are required to set the flags, that will be reported on the Presentation 'presented' event. This is meant for output-wide feedback flags. Flags that vary per wl_surface are subject for the following patch. All start_repaint_loop functions use the special private flag PRESENTATION_FEEDBACK_INVALID to mark, that this call of weston_output_finish_frame() cannot trigger the 'presented' event. If it does, we now hit an assert, and should then investigate why a fake update triggered Presentation feedback. DRM: Page flip is always vsync'd, and always gets the completion timestamp from the kernel which should correspond well to hardware. Completion is triggered by the kernel/hardware. Vblank handler is only used with the broken planes path, therefore do not report VSYNC, because we cannot guarantee all the planes updated at the same time. We cannot set the INVALID, because it would abort the compositor if the broken planes path was ever used. This is a hack that will get fixed with nuclear pageflip support in the future. fbdev: No vsync, update done by copy, no completion event from hardware, and completion time is totally fake. headless: No real output to update. RDP: Guessing that maybe no vsync, fake time, and copy make sense (pixels sent over network). Also no event that the pixels have been shown? RPI: Presumably Dispmanx updates are vsync'd. We get a completion event from the driver, but need to read the clock ourselves, so the completion time is somewhat unreliable. Zero-copy flag not implemented though it would be theoretically possible with EGL clients (zero-copy is a per-surface flag anyway, so in this patch). Wayland: No information how the host compositor is doing updates, so make a safe guess without assuming vsync or hardware completion event. While we do get some timestamp from the host compositor, it is not the completion time. Would need to hook to the Presentation extension of the host compositor to get more accurate flags. X11: No idea about vsync, completion event, or copying. Also the timestamp is a fake. Signed-off-by: Pekka Paalanen <pekka.paalanen@collabora.co.uk> Reviewed-by: Mario Kleiner <mario.kleiner.de@gmail.com> Tested-by: Mario Kleiner <mario.kleiner.de@gmail.com> Acked-by: Mario Kleiner <mario.kleiner.de@gmail.com>
2014-12-17 17:20:40 +03:00
const struct timespec *stamp,
uint32_t presented_flags)
{
struct weston_compositor *compositor = output->compositor;
compositor: add repaint delay timer This timer delays the output_repaint towards the end of the refresh period, reducing the time from repaint to present. The length of the repaint window can be set in weston.ini. The call to weston_output_schedule_repaint_reset() is delayed by one more period. If we exit the continuous repaint loop (set output->repaint_scheduled to false) in finish_frame, we may call start_repaint_loop() unnecessarily. The problem case was actually observed with two outputs on the DRM backend at 60 Hz, and 7 ms repaint-window. During a window move, one output was constantly falling off the continuous repaint loop and introducing additional one frame latency, leading to jerky window motion. This code now avoids the problem. Changes in v2: - Rename repaint_delay_timer to repaint_timer and output_repaint_delay_handler to output_repaint_timer_handler. - When computing the delay, take the current time into account. The timer uses a relative timeout, so we have to subtract any time already gone. Note, that 'gone' may also be negative. DRM has a habit of predicting the page flip timestamp so it may be still in the future when we get the completion event. - Do also a sanity check 'msec > 1000'. In the unlikely case that something fails to provide a good timestamp, never delay for more than one second. Signed-off-by: Pekka Paalanen <pekka.paalanen@collabora.co.uk> Reviewed-By: Derek Foreman <derekf@osg.samsung.com> Reviewed-by: Bryce Harrington <bryce@osg.samsung.com>
2014-05-21 17:17:27 +04:00
int32_t refresh_nsec;
struct timespec now;
struct timespec gone;
int msec;
compositor: Implement JSON-timeline logging Logging is activated and deactivated with the debug key binding 't'. When activated, it creates a new log file, where it records the events. The log file contains events and detailed object information entries in JSON format, and is meant to be parsed in sequence from beginning to the end. The emitted events are mostly related to the output repaint cycle, like when repaint begins, is submitted to GPU, and when it completes on a vblank. This is recorded per-output. Also some per-surface events are recorded, including when surface damage is flushed. To reduce the log size, events refer to objects like outputs and surfaces by id numbers. Detailed object information is emitted only as needed: on the first object occurrence, and afterwards only if weston_timeline_object::force_refresh asks for it. The detailed information for surfaces includes the string returned by weston_surface::get_label. Therefore it is important to set weston_timeline_object::force_refresh = 1 whenever the string would change, so that the new details get recorded. A rudimentary parser and SVG generator can be found at: https://github.com/ppaalanen/wesgr The timeline logs can answer questions including: - How does the compositor repaint cycle work timing-wise? - When was the vblank deadline missed? - What is the latency from surface commit to showing the new content on screen? - How long does it take to process the scenegraph? v2: weston_surface::get_description renamed to get_label. v3: reafctor a bit into fprint_quoted_string(). Signed-off-by: Pekka Paalanen <pekka.paalanen@collabora.co.uk>
2014-11-12 16:09:24 +03:00
TL_POINT("core_repaint_finished", TLP_OUTPUT(output),
TLP_VBLANK(stamp), TLP_END);
refresh_nsec = millihz_to_nsec(output->current_mode->refresh);
weston_presentation_feedback_present_list(&output->feedback_list,
output, refresh_nsec, stamp,
compositor: set presentation.presented flags Change weston_output_finish_frame() signature so that backends are required to set the flags, that will be reported on the Presentation 'presented' event. This is meant for output-wide feedback flags. Flags that vary per wl_surface are subject for the following patch. All start_repaint_loop functions use the special private flag PRESENTATION_FEEDBACK_INVALID to mark, that this call of weston_output_finish_frame() cannot trigger the 'presented' event. If it does, we now hit an assert, and should then investigate why a fake update triggered Presentation feedback. DRM: Page flip is always vsync'd, and always gets the completion timestamp from the kernel which should correspond well to hardware. Completion is triggered by the kernel/hardware. Vblank handler is only used with the broken planes path, therefore do not report VSYNC, because we cannot guarantee all the planes updated at the same time. We cannot set the INVALID, because it would abort the compositor if the broken planes path was ever used. This is a hack that will get fixed with nuclear pageflip support in the future. fbdev: No vsync, update done by copy, no completion event from hardware, and completion time is totally fake. headless: No real output to update. RDP: Guessing that maybe no vsync, fake time, and copy make sense (pixels sent over network). Also no event that the pixels have been shown? RPI: Presumably Dispmanx updates are vsync'd. We get a completion event from the driver, but need to read the clock ourselves, so the completion time is somewhat unreliable. Zero-copy flag not implemented though it would be theoretically possible with EGL clients (zero-copy is a per-surface flag anyway, so in this patch). Wayland: No information how the host compositor is doing updates, so make a safe guess without assuming vsync or hardware completion event. While we do get some timestamp from the host compositor, it is not the completion time. Would need to hook to the Presentation extension of the host compositor to get more accurate flags. X11: No idea about vsync, completion event, or copying. Also the timestamp is a fake. Signed-off-by: Pekka Paalanen <pekka.paalanen@collabora.co.uk> Reviewed-by: Mario Kleiner <mario.kleiner.de@gmail.com> Tested-by: Mario Kleiner <mario.kleiner.de@gmail.com> Acked-by: Mario Kleiner <mario.kleiner.de@gmail.com>
2014-12-17 17:20:40 +03:00
output->msc,
presented_flags);
output->frame_time = stamp->tv_sec * 1000 + stamp->tv_nsec / 1000000;
compositor: add repaint delay timer This timer delays the output_repaint towards the end of the refresh period, reducing the time from repaint to present. The length of the repaint window can be set in weston.ini. The call to weston_output_schedule_repaint_reset() is delayed by one more period. If we exit the continuous repaint loop (set output->repaint_scheduled to false) in finish_frame, we may call start_repaint_loop() unnecessarily. The problem case was actually observed with two outputs on the DRM backend at 60 Hz, and 7 ms repaint-window. During a window move, one output was constantly falling off the continuous repaint loop and introducing additional one frame latency, leading to jerky window motion. This code now avoids the problem. Changes in v2: - Rename repaint_delay_timer to repaint_timer and output_repaint_delay_handler to output_repaint_timer_handler. - When computing the delay, take the current time into account. The timer uses a relative timeout, so we have to subtract any time already gone. Note, that 'gone' may also be negative. DRM has a habit of predicting the page flip timestamp so it may be still in the future when we get the completion event. - Do also a sanity check 'msec > 1000'. In the unlikely case that something fails to provide a good timestamp, never delay for more than one second. Signed-off-by: Pekka Paalanen <pekka.paalanen@collabora.co.uk> Reviewed-By: Derek Foreman <derekf@osg.samsung.com> Reviewed-by: Bryce Harrington <bryce@osg.samsung.com>
2014-05-21 17:17:27 +04:00
weston_compositor_read_presentation_clock(compositor, &now);
timespec_sub(&gone, &now, stamp);
msec = (refresh_nsec - timespec_to_nsec(&gone)) / 1000000; /* floor */
msec -= compositor->repaint_msec;
if (msec < -1000 || msec > 1000) {
static bool warned;
if (!warned)
weston_log("Warning: computed repaint delay is "
"insane: %d msec\n", msec);
warned = true;
msec = 0;
}
compositor: Respect repaint-window when restarting repaint loop. If a stopped repaint loop gets restarted due to posting of new damage, and this restart of the repaint loop happens late in the video refresh cycle, ie. already inside the repaint-window and thereby after the composition deadline for the current frame, then defer the actual output repaint to the composition deadline of the next video refresh cycle by setting the repaint timer accordingly. This tries to make sure that: a) Client(s) posting damage timely before the composition deadline (video refresh duration - "repaint-window" duration) of the current refresh cycle will trigger a repaint within the current refresh cycle, thereby avoiding one extra frame of compositor lag due to the needed restart of the repaint loop if the loop was stopped. This allows them to benefit from the earlier "instant repaint restart" commit to keep latency low. b) Late clients which post damage close to the end of a refresh cycle can't race other clients if the repaint loop is restarted. Instead they will get deferred to the next compositor cycle, just as if the repaint loop would have been already running - the semantic of the "repaint-window" parameter is preserved. This is especially important to prevent a very late client from triggering a repaint very close to the vblank, which would cause the compositor to certainly miss the vblank and skip one frame and then cause a delay of another frame for other clients which posted their damage in time for the following frame. Iow. this provides clients with a more predictable compositor timing and makes it easier for them to latch onto the compositors repaint cycle. Signed-off-by: Mario Kleiner <mario.kleiner.de@gmail.com> Reviewed-by: Pekka Paalanen <pekka.paalanen@collabora.co.uk>
2015-06-21 22:25:15 +03:00
/* Called from restart_repaint_loop and restart happens already after
* the deadline given by repaint_msec? In that case we delay until
* the deadline of the next frame, to give clients a more predictable
* timing of the repaint cycle to lock on. */
if (presented_flags == WP_PRESENTATION_FEEDBACK_INVALID && msec < 0)
compositor: Respect repaint-window when restarting repaint loop. If a stopped repaint loop gets restarted due to posting of new damage, and this restart of the repaint loop happens late in the video refresh cycle, ie. already inside the repaint-window and thereby after the composition deadline for the current frame, then defer the actual output repaint to the composition deadline of the next video refresh cycle by setting the repaint timer accordingly. This tries to make sure that: a) Client(s) posting damage timely before the composition deadline (video refresh duration - "repaint-window" duration) of the current refresh cycle will trigger a repaint within the current refresh cycle, thereby avoiding one extra frame of compositor lag due to the needed restart of the repaint loop if the loop was stopped. This allows them to benefit from the earlier "instant repaint restart" commit to keep latency low. b) Late clients which post damage close to the end of a refresh cycle can't race other clients if the repaint loop is restarted. Instead they will get deferred to the next compositor cycle, just as if the repaint loop would have been already running - the semantic of the "repaint-window" parameter is preserved. This is especially important to prevent a very late client from triggering a repaint very close to the vblank, which would cause the compositor to certainly miss the vblank and skip one frame and then cause a delay of another frame for other clients which posted their damage in time for the following frame. Iow. this provides clients with a more predictable compositor timing and makes it easier for them to latch onto the compositors repaint cycle. Signed-off-by: Mario Kleiner <mario.kleiner.de@gmail.com> Reviewed-by: Pekka Paalanen <pekka.paalanen@collabora.co.uk>
2015-06-21 22:25:15 +03:00
msec += refresh_nsec / 1000000;
if (msec < 1)
compositor: add repaint delay timer This timer delays the output_repaint towards the end of the refresh period, reducing the time from repaint to present. The length of the repaint window can be set in weston.ini. The call to weston_output_schedule_repaint_reset() is delayed by one more period. If we exit the continuous repaint loop (set output->repaint_scheduled to false) in finish_frame, we may call start_repaint_loop() unnecessarily. The problem case was actually observed with two outputs on the DRM backend at 60 Hz, and 7 ms repaint-window. During a window move, one output was constantly falling off the continuous repaint loop and introducing additional one frame latency, leading to jerky window motion. This code now avoids the problem. Changes in v2: - Rename repaint_delay_timer to repaint_timer and output_repaint_delay_handler to output_repaint_timer_handler. - When computing the delay, take the current time into account. The timer uses a relative timeout, so we have to subtract any time already gone. Note, that 'gone' may also be negative. DRM has a habit of predicting the page flip timestamp so it may be still in the future when we get the completion event. - Do also a sanity check 'msec > 1000'. In the unlikely case that something fails to provide a good timestamp, never delay for more than one second. Signed-off-by: Pekka Paalanen <pekka.paalanen@collabora.co.uk> Reviewed-By: Derek Foreman <derekf@osg.samsung.com> Reviewed-by: Bryce Harrington <bryce@osg.samsung.com>
2014-05-21 17:17:27 +04:00
output_repaint_timer_handler(output);
else
wl_event_source_timer_update(output->repaint_timer, msec);
}
static void
idle_repaint(void *data)
{
struct weston_output *output = data;
output->start_repaint_loop(output);
}
WL_EXPORT void
weston_layer_entry_insert(struct weston_layer_entry *list,
struct weston_layer_entry *entry)
{
wl_list_insert(&list->link, &entry->link);
entry->layer = list->layer;
}
WL_EXPORT void
weston_layer_entry_remove(struct weston_layer_entry *entry)
{
wl_list_remove(&entry->link);
wl_list_init(&entry->link);
entry->layer = NULL;
}
/** Initialize the weston_layer struct.
*
* \param compositor The compositor instance
* \param layer The layer to initialize
*/
WL_EXPORT void
weston_layer_init(struct weston_layer *layer,
struct weston_compositor *compositor)
{
layer->compositor = compositor;
wl_list_init(&layer->link);
wl_list_init(&layer->view_list.link);
layer->view_list.layer = layer;
weston_layer_set_mask_infinite(layer);
}
/** Sets the position of the layer in the layer list. The layer will be placed
* below any layer with the same position value, if any.
* This function is safe to call if the layer is already on the list, but the
* layer may be moved below other layers at the same position, if any.
*
* \param layer The layer to modify
* \param position The position the layer will be placed at
*/
WL_EXPORT void
weston_layer_set_position(struct weston_layer *layer,
enum weston_layer_position position)
{
struct weston_layer *below;
wl_list_remove(&layer->link);
/* layer_list is ordered from top to bottom, the last layer being the
* background with the smallest position value */
layer->position = position;
wl_list_for_each_reverse(below, &layer->compositor->layer_list, link) {
if (below->position >= layer->position) {
wl_list_insert(&below->link, &layer->link);
return;
}
}
wl_list_insert(&layer->compositor->layer_list, &layer->link);
}
/** Hide a layer by taking it off the layer list.
* This function is safe to call if the layer is not on the list.
*
* \param layer The layer to hide
*/
WL_EXPORT void
weston_layer_unset_position(struct weston_layer *layer)
{
wl_list_remove(&layer->link);
wl_list_init(&layer->link);
}
WL_EXPORT void
weston_layer_set_mask(struct weston_layer *layer,
int x, int y, int width, int height)
{
struct weston_view *view;
layer->mask.x1 = x;
layer->mask.x2 = x + width;
layer->mask.y1 = y;
layer->mask.y2 = y + height;
wl_list_for_each(view, &layer->view_list.link, layer_link.link) {
weston_view_geometry_dirty(view);
}
}
WL_EXPORT void
weston_layer_set_mask_infinite(struct weston_layer *layer)
{
weston_layer_set_mask(layer, INT32_MIN, INT32_MIN,
UINT32_MAX, UINT32_MAX);
}
WL_EXPORT void
weston_output_schedule_repaint(struct weston_output *output)
{
struct weston_compositor *compositor = output->compositor;
struct wl_event_loop *loop;
if (compositor->state == WESTON_COMPOSITOR_SLEEPING ||
compositor->state == WESTON_COMPOSITOR_OFFSCREEN)
return;
compositor: Implement JSON-timeline logging Logging is activated and deactivated with the debug key binding 't'. When activated, it creates a new log file, where it records the events. The log file contains events and detailed object information entries in JSON format, and is meant to be parsed in sequence from beginning to the end. The emitted events are mostly related to the output repaint cycle, like when repaint begins, is submitted to GPU, and when it completes on a vblank. This is recorded per-output. Also some per-surface events are recorded, including when surface damage is flushed. To reduce the log size, events refer to objects like outputs and surfaces by id numbers. Detailed object information is emitted only as needed: on the first object occurrence, and afterwards only if weston_timeline_object::force_refresh asks for it. The detailed information for surfaces includes the string returned by weston_surface::get_label. Therefore it is important to set weston_timeline_object::force_refresh = 1 whenever the string would change, so that the new details get recorded. A rudimentary parser and SVG generator can be found at: https://github.com/ppaalanen/wesgr The timeline logs can answer questions including: - How does the compositor repaint cycle work timing-wise? - When was the vblank deadline missed? - What is the latency from surface commit to showing the new content on screen? - How long does it take to process the scenegraph? v2: weston_surface::get_description renamed to get_label. v3: reafctor a bit into fprint_quoted_string(). Signed-off-by: Pekka Paalanen <pekka.paalanen@collabora.co.uk>
2014-11-12 16:09:24 +03:00
if (!output->repaint_needed)
TL_POINT("core_repaint_req", TLP_OUTPUT(output), TLP_END);
loop = wl_display_get_event_loop(compositor->wl_display);
output->repaint_needed = 1;
if (output->repaint_scheduled)
return;
wl_event_loop_add_idle(loop, idle_repaint, output);
output->repaint_scheduled = 1;
compositor: Implement JSON-timeline logging Logging is activated and deactivated with the debug key binding 't'. When activated, it creates a new log file, where it records the events. The log file contains events and detailed object information entries in JSON format, and is meant to be parsed in sequence from beginning to the end. The emitted events are mostly related to the output repaint cycle, like when repaint begins, is submitted to GPU, and when it completes on a vblank. This is recorded per-output. Also some per-surface events are recorded, including when surface damage is flushed. To reduce the log size, events refer to objects like outputs and surfaces by id numbers. Detailed object information is emitted only as needed: on the first object occurrence, and afterwards only if weston_timeline_object::force_refresh asks for it. The detailed information for surfaces includes the string returned by weston_surface::get_label. Therefore it is important to set weston_timeline_object::force_refresh = 1 whenever the string would change, so that the new details get recorded. A rudimentary parser and SVG generator can be found at: https://github.com/ppaalanen/wesgr The timeline logs can answer questions including: - How does the compositor repaint cycle work timing-wise? - When was the vblank deadline missed? - What is the latency from surface commit to showing the new content on screen? - How long does it take to process the scenegraph? v2: weston_surface::get_description renamed to get_label. v3: reafctor a bit into fprint_quoted_string(). Signed-off-by: Pekka Paalanen <pekka.paalanen@collabora.co.uk>
2014-11-12 16:09:24 +03:00
TL_POINT("core_repaint_enter_loop", TLP_OUTPUT(output), TLP_END);
}
2009-02-09 23:17:46 +03:00
WL_EXPORT void
weston_compositor_schedule_repaint(struct weston_compositor *compositor)
{
struct weston_output *output;
wl_list_for_each(output, &compositor->output_list, link)
weston_output_schedule_repaint(output);
}
static void
surface_destroy(struct wl_client *client, struct wl_resource *resource)
{
wl_resource_destroy(resource);
}
static void
surface_attach(struct wl_client *client,
struct wl_resource *resource,
struct wl_resource *buffer_resource, int32_t sx, int32_t sy)
{
struct weston_surface *surface = wl_resource_get_user_data(resource);
struct weston_buffer *buffer = NULL;
if (buffer_resource) {
buffer = weston_buffer_from_resource(buffer_resource);
if (buffer == NULL) {
wl_client_post_no_memory(client);
return;
}
}
/* Attach, attach, without commit in between does not send
* wl_buffer.release. */
weston_surface_state_set_buffer(&surface->pending, buffer);
surface->pending.sx = sx;
surface->pending.sy = sy;
surface->pending.newly_attached = 1;
}
static void
surface_damage(struct wl_client *client,
struct wl_resource *resource,
int32_t x, int32_t y, int32_t width, int32_t height)
{
struct weston_surface *surface = wl_resource_get_user_data(resource);
if (width <= 0 || height <= 0)
return;
pixman_region32_union_rect(&surface->pending.damage_surface,
&surface->pending.damage_surface,
x, y, width, height);
}
static void
surface_damage_buffer(struct wl_client *client,
struct wl_resource *resource,
int32_t x, int32_t y, int32_t width, int32_t height)
{
struct weston_surface *surface = wl_resource_get_user_data(resource);
if (width <= 0 || height <= 0)
return;
pixman_region32_union_rect(&surface->pending.damage_buffer,
&surface->pending.damage_buffer,
x, y, width, height);
}
2011-08-17 07:01:28 +04:00
static void
destroy_frame_callback(struct wl_resource *resource)
2011-08-17 07:01:28 +04:00
{
struct weston_frame_callback *cb = wl_resource_get_user_data(resource);
wl_list_remove(&cb->link);
free(cb);
2011-08-17 07:01:28 +04:00
}
static void
surface_frame(struct wl_client *client,
struct wl_resource *resource, uint32_t callback)
2011-08-17 07:01:28 +04:00
{
struct weston_frame_callback *cb;
struct weston_surface *surface = wl_resource_get_user_data(resource);
2011-08-17 07:01:28 +04:00
cb = malloc(sizeof *cb);
if (cb == NULL) {
2011-09-01 17:54:57 +04:00
wl_resource_post_no_memory(resource);
2011-08-17 07:01:28 +04:00
return;
}
cb->resource = wl_resource_create(client, &wl_callback_interface, 1,
callback);
if (cb->resource == NULL) {
free(cb);
wl_resource_post_no_memory(resource);
return;
}
wl_resource_set_implementation(cb->resource, NULL, cb,
destroy_frame_callback);
2011-08-17 07:01:28 +04:00
wl_list_insert(surface->pending.frame_callback_list.prev, &cb->link);
2011-08-17 07:01:28 +04:00
}
static void
surface_set_opaque_region(struct wl_client *client,
struct wl_resource *resource,
struct wl_resource *region_resource)
{
struct weston_surface *surface = wl_resource_get_user_data(resource);
struct weston_region *region;
if (region_resource) {
region = wl_resource_get_user_data(region_resource);
pixman_region32_copy(&surface->pending.opaque,
&region->region);
} else {
pixman_region32_clear(&surface->pending.opaque);
}
}
static void
surface_set_input_region(struct wl_client *client,
struct wl_resource *resource,
struct wl_resource *region_resource)
{
struct weston_surface *surface = wl_resource_get_user_data(resource);
struct weston_region *region;
if (region_resource) {
region = wl_resource_get_user_data(region_resource);
pixman_region32_copy(&surface->pending.input,
&region->region);
} else {
pixman_region32_fini(&surface->pending.input);
region_init_infinite(&surface->pending.input);
}
}
/* Cause damage to this sub-surface and all its children.
*
* This is useful when there are state changes that need an implicit
* damage, e.g. a z-order change.
*/
static void
weston_surface_damage_subsurfaces(struct weston_subsurface *sub)
{
struct weston_subsurface *child;
weston_surface_damage(sub->surface);
sub->reordered = false;
wl_list_for_each(child, &sub->surface->subsurface_list, parent_link)
if (child != sub)
weston_surface_damage_subsurfaces(child);
}
static void
compositor: introduce sub-surfaces Implement the basic protocol for sub-surfaces: - expose wl_subcompositor global interface - error checking on protocol calls - associate a parent wl_surface to a sub-surface - introduce the sub-surface role, which is exclusive - an implementation of the wl_subsurface interface - allow nesting of sub-surfaces - proper surface transformation inheritance from parent to sub-surfaces - two different modes of wl_surface.commit for sub-surfaces - hook sub-surfaces up to repaint by modifying the repaint list code Struct weston_subsurface is dynamically allocated. For sub-surfaces, it is completely populated. For parent surfaces, weston_subsurface acts only as a link for stacking order purposes. The wl_resource is unused, parent_destroy_listener is not registered, the transform is not linked, etc. Sub-surfaces are not added directly into layers for display or input. Instead, they are hooked up via the sub-surface list present in parent weston_surface. This way sub-surfaces are inherently linked to the parent surface, and cannot be displayed unless the parent is mapped, too. This also eases restacking, as only the parent will be in a layer list. Also, only the main surface should be subject to shell actions. The surface list rebuilding in weston_output_repaint() is modified to process sub-surface lists, if they are non-empty. The sub-surface list always contains the parent, too, unless empty. The collection of frame_callback_list is moved to a later loop, to streamline the surface list rebuild functions. Features still lacking are: - full-surface alpha support for compound windows Changes in v2: - fix a bug in surface mapping: commit a sub-surface would cause the main surface to never be mapped. - remove debug printfs - detect attempt of making a surface its own parent - always zero-alloc weston_subsurface - apply wl_subsurface.set_position in commit, not immediately - add weston_surface_to_subsurface() - implement sub-surface commit modes parent-cached and independent - implement wl_subcompositor.destroy and wl_subsurface.destroy Changes in v3: - rebased, and use the new transform inheritance code - squashed the commit "add sub-surfaces to repaint list" - fixed a buffer reference leak in commit_from_cache() - Rewrite the sub-surface destructor code, and make it leave the wl_subsurface protocol object inert, if one destroys the corresponding wl_surface. - replaced set_commit_mode with set_sync and set_desync - allowed sub-surface nesting, and fixed repaint accordingly - implemented nested sub-surface commit modes - Made the sub-surface order changes from wl_subsurface.place_above and .place_below to be applied when the parent surface state is applied, instead of immediately. This conforms with the protocol specification now. Signed-off-by: Pekka Paalanen <ppaalanen@gmail.com>
2013-04-25 14:57:42 +04:00
weston_surface_commit_subsurface_order(struct weston_surface *surface)
{
struct weston_subsurface *sub;
wl_list_for_each_reverse(sub, &surface->subsurface_list_pending,
parent_link_pending) {
wl_list_remove(&sub->parent_link);
wl_list_insert(&surface->subsurface_list, &sub->parent_link);
if (sub->reordered)
weston_surface_damage_subsurfaces(sub);
compositor: introduce sub-surfaces Implement the basic protocol for sub-surfaces: - expose wl_subcompositor global interface - error checking on protocol calls - associate a parent wl_surface to a sub-surface - introduce the sub-surface role, which is exclusive - an implementation of the wl_subsurface interface - allow nesting of sub-surfaces - proper surface transformation inheritance from parent to sub-surfaces - two different modes of wl_surface.commit for sub-surfaces - hook sub-surfaces up to repaint by modifying the repaint list code Struct weston_subsurface is dynamically allocated. For sub-surfaces, it is completely populated. For parent surfaces, weston_subsurface acts only as a link for stacking order purposes. The wl_resource is unused, parent_destroy_listener is not registered, the transform is not linked, etc. Sub-surfaces are not added directly into layers for display or input. Instead, they are hooked up via the sub-surface list present in parent weston_surface. This way sub-surfaces are inherently linked to the parent surface, and cannot be displayed unless the parent is mapped, too. This also eases restacking, as only the parent will be in a layer list. Also, only the main surface should be subject to shell actions. The surface list rebuilding in weston_output_repaint() is modified to process sub-surface lists, if they are non-empty. The sub-surface list always contains the parent, too, unless empty. The collection of frame_callback_list is moved to a later loop, to streamline the surface list rebuild functions. Features still lacking are: - full-surface alpha support for compound windows Changes in v2: - fix a bug in surface mapping: commit a sub-surface would cause the main surface to never be mapped. - remove debug printfs - detect attempt of making a surface its own parent - always zero-alloc weston_subsurface - apply wl_subsurface.set_position in commit, not immediately - add weston_surface_to_subsurface() - implement sub-surface commit modes parent-cached and independent - implement wl_subcompositor.destroy and wl_subsurface.destroy Changes in v3: - rebased, and use the new transform inheritance code - squashed the commit "add sub-surfaces to repaint list" - fixed a buffer reference leak in commit_from_cache() - Rewrite the sub-surface destructor code, and make it leave the wl_subsurface protocol object inert, if one destroys the corresponding wl_surface. - replaced set_commit_mode with set_sync and set_desync - allowed sub-surface nesting, and fixed repaint accordingly - implemented nested sub-surface commit modes - Made the sub-surface order changes from wl_subsurface.place_above and .place_below to be applied when the parent surface state is applied, instead of immediately. This conforms with the protocol specification now. Signed-off-by: Pekka Paalanen <ppaalanen@gmail.com>
2013-04-25 14:57:42 +04:00
}
}
static void
weston_surface_build_buffer_matrix(const struct weston_surface *surface,
struct weston_matrix *matrix)
{
const struct weston_buffer_viewport *vp = &surface->buffer_viewport;
double src_width, src_height, dest_width, dest_height;
weston_matrix_init(matrix);
if (vp->buffer.src_width == wl_fixed_from_int(-1)) {
src_width = surface->width_from_buffer;
src_height = surface->height_from_buffer;
} else {
src_width = wl_fixed_to_double(vp->buffer.src_width);
src_height = wl_fixed_to_double(vp->buffer.src_height);
}
if (vp->surface.width == -1) {
dest_width = src_width;
dest_height = src_height;
} else {
dest_width = vp->surface.width;
dest_height = vp->surface.height;
}
if (src_width != dest_width || src_height != dest_height)
weston_matrix_scale(matrix,
src_width / dest_width,
src_height / dest_height, 1);
if (vp->buffer.src_width != wl_fixed_from_int(-1))
weston_matrix_translate(matrix,
wl_fixed_to_double(vp->buffer.src_x),
wl_fixed_to_double(vp->buffer.src_y),
0);
switch (vp->buffer.transform) {
case WL_OUTPUT_TRANSFORM_FLIPPED:
case WL_OUTPUT_TRANSFORM_FLIPPED_90:
case WL_OUTPUT_TRANSFORM_FLIPPED_180:
case WL_OUTPUT_TRANSFORM_FLIPPED_270:
weston_matrix_scale(matrix, -1, 1, 1);
weston_matrix_translate(matrix,
surface->width_from_buffer, 0, 0);
break;
}
switch (vp->buffer.transform) {
default:
case WL_OUTPUT_TRANSFORM_NORMAL:
case WL_OUTPUT_TRANSFORM_FLIPPED:
break;
case WL_OUTPUT_TRANSFORM_90:
case WL_OUTPUT_TRANSFORM_FLIPPED_90:
weston_matrix_rotate_xy(matrix, 0, 1);
weston_matrix_translate(matrix,
surface->height_from_buffer, 0, 0);
break;
case WL_OUTPUT_TRANSFORM_180:
case WL_OUTPUT_TRANSFORM_FLIPPED_180:
weston_matrix_rotate_xy(matrix, -1, 0);
weston_matrix_translate(matrix,
surface->width_from_buffer,
surface->height_from_buffer, 0);
break;
case WL_OUTPUT_TRANSFORM_270:
case WL_OUTPUT_TRANSFORM_FLIPPED_270:
weston_matrix_rotate_xy(matrix, 0, -1);
weston_matrix_translate(matrix,
0, surface->width_from_buffer, 0);
break;
}
weston_matrix_scale(matrix, vp->buffer.scale, vp->buffer.scale, 1);
}
/**
* Compute a + b > c while being safe to overflows.
*/
static bool
fixed_sum_gt(wl_fixed_t a, wl_fixed_t b, wl_fixed_t c)
{
return (int64_t)a + (int64_t)b > (int64_t)c;
}
static bool
weston_surface_is_pending_viewport_source_valid(
const struct weston_surface *surface)
{
const struct weston_surface_state *pend = &surface->pending;
const struct weston_buffer_viewport *vp = &pend->buffer_viewport;
int width_from_buffer = 0;
int height_from_buffer = 0;
wl_fixed_t w;
wl_fixed_t h;
/* If viewport source rect is not set, it is always ok. */
if (vp->buffer.src_width == wl_fixed_from_int(-1))
return true;
if (pend->newly_attached) {
if (pend->buffer) {
convert_size_by_transform_scale(&width_from_buffer,
&height_from_buffer,
pend->buffer->width,
pend->buffer->height,
vp->buffer.transform,
vp->buffer.scale);
} else {
/* No buffer: viewport is irrelevant. */
return true;
}
} else {
width_from_buffer = surface->width_from_buffer;
height_from_buffer = surface->height_from_buffer;
}
assert((width_from_buffer == 0) == (height_from_buffer == 0));
assert(width_from_buffer >= 0 && height_from_buffer >= 0);
/* No buffer: viewport is irrelevant. */
if (width_from_buffer == 0 || height_from_buffer == 0)
return true;
/* overflow checks for wl_fixed_from_int() */
if (width_from_buffer > wl_fixed_to_int(INT32_MAX))
return false;
if (height_from_buffer > wl_fixed_to_int(INT32_MAX))
return false;
w = wl_fixed_from_int(width_from_buffer);
h = wl_fixed_from_int(height_from_buffer);
if (fixed_sum_gt(vp->buffer.src_x, vp->buffer.src_width, w))
return false;
if (fixed_sum_gt(vp->buffer.src_y, vp->buffer.src_height, h))
return false;
return true;
}
static bool
fixed_is_integer(wl_fixed_t v)
{
return (v & 0xff) == 0;
}
static bool
weston_surface_is_pending_viewport_dst_size_int(
const struct weston_surface *surface)
{
const struct weston_buffer_viewport *vp =
&surface->pending.buffer_viewport;
if (vp->surface.width != -1) {
assert(vp->surface.width > 0 && vp->surface.height > 0);
return true;
}
return fixed_is_integer(vp->buffer.src_width) &&
fixed_is_integer(vp->buffer.src_height);
}
/* Translate pending damage in buffer co-ordinates to surface
* co-ordinates and union it with a pixman_region32_t.
* This should only be called after the buffer is attached.
*/
static void
apply_damage_buffer(pixman_region32_t *dest,
struct weston_surface *surface,
struct weston_surface_state *state)
{
struct weston_buffer *buffer = surface->buffer_ref.buffer;
/* wl_surface.damage_buffer needs to be clipped to the buffer,
* translated into surface co-ordinates and unioned with
* any other surface damage.
* None of this makes sense if there is no buffer though.
*/
if (buffer && pixman_region32_not_empty(&state->damage_buffer)) {
pixman_region32_t buffer_damage;
pixman_region32_intersect_rect(&state->damage_buffer,
&state->damage_buffer,
0, 0, buffer->width,
buffer->height);
pixman_region32_init(&buffer_damage);
weston_matrix_transform_region(&buffer_damage,
&surface->buffer_to_surface_matrix,
&state->damage_buffer);
pixman_region32_union(dest, dest, &buffer_damage);
pixman_region32_fini(&buffer_damage);
}
/* We should clear this on commit even if there was no buffer */
pixman_region32_clear(&state->damage_buffer);
}
compositor: introduce sub-surfaces Implement the basic protocol for sub-surfaces: - expose wl_subcompositor global interface - error checking on protocol calls - associate a parent wl_surface to a sub-surface - introduce the sub-surface role, which is exclusive - an implementation of the wl_subsurface interface - allow nesting of sub-surfaces - proper surface transformation inheritance from parent to sub-surfaces - two different modes of wl_surface.commit for sub-surfaces - hook sub-surfaces up to repaint by modifying the repaint list code Struct weston_subsurface is dynamically allocated. For sub-surfaces, it is completely populated. For parent surfaces, weston_subsurface acts only as a link for stacking order purposes. The wl_resource is unused, parent_destroy_listener is not registered, the transform is not linked, etc. Sub-surfaces are not added directly into layers for display or input. Instead, they are hooked up via the sub-surface list present in parent weston_surface. This way sub-surfaces are inherently linked to the parent surface, and cannot be displayed unless the parent is mapped, too. This also eases restacking, as only the parent will be in a layer list. Also, only the main surface should be subject to shell actions. The surface list rebuilding in weston_output_repaint() is modified to process sub-surface lists, if they are non-empty. The sub-surface list always contains the parent, too, unless empty. The collection of frame_callback_list is moved to a later loop, to streamline the surface list rebuild functions. Features still lacking are: - full-surface alpha support for compound windows Changes in v2: - fix a bug in surface mapping: commit a sub-surface would cause the main surface to never be mapped. - remove debug printfs - detect attempt of making a surface its own parent - always zero-alloc weston_subsurface - apply wl_subsurface.set_position in commit, not immediately - add weston_surface_to_subsurface() - implement sub-surface commit modes parent-cached and independent - implement wl_subcompositor.destroy and wl_subsurface.destroy Changes in v3: - rebased, and use the new transform inheritance code - squashed the commit "add sub-surfaces to repaint list" - fixed a buffer reference leak in commit_from_cache() - Rewrite the sub-surface destructor code, and make it leave the wl_subsurface protocol object inert, if one destroys the corresponding wl_surface. - replaced set_commit_mode with set_sync and set_desync - allowed sub-surface nesting, and fixed repaint accordingly - implemented nested sub-surface commit modes - Made the sub-surface order changes from wl_subsurface.place_above and .place_below to be applied when the parent surface state is applied, instead of immediately. This conforms with the protocol specification now. Signed-off-by: Pekka Paalanen <ppaalanen@gmail.com>
2013-04-25 14:57:42 +04:00
static void
weston_surface_commit_state(struct weston_surface *surface,
struct weston_surface_state *state)
{
Split the geometry information from weston_surface out into weston_view The weston_surface structure is split into two structures: * The weston_surface structure storres everything required for a client-side or server-side surface. This includes buffers; callbacks; backend private data; input, damage, and opaque regions; and a few other bookkeeping bits. * The weston_view structure represents an entity in the scenegraph and storres all of the geometry information. This includes clip region, alpha, position, and the transformation list as well as all of the temporary information derived from the geometry state. Because a view, and not a surface, is a scenegraph element, the view is what is placed in layers and planes. There are a few things worth noting about the surface/view split: 1. This is *not* a modification to the protocol. It is, instead, a modification to Weston's internal scenegraph to allow a single surface to exist in multiple places at a time. Clients are completely unaware of how many views to a particular surface exist. 2. A view is considered a direct child of a surface and is destroyed when the surface is destroyed. Because of this, the view.surface pointer is always valid and non-null. 3. The compositor's surface_list is replaced with a view_list. Due to subsurfaces, building the view list is a little more complicated than it used to be and involves building a tree of views on the fly whenever subsurfaces are used. However, this means that backends can remain completely subsurface-agnostic. 4. Surfaces and views both keep track of which outputs they are on. 5. The weston_surface structure now has width and height fields. These are populated when a new buffer is attached before surface.configure is called. This is because there are many surface-based operations that really require the width and height and digging through the views didn't work well. Signed-off-by: Jason Ekstrand <jason@jlekstrand.net>
2013-10-13 07:38:11 +04:00
struct weston_view *view;
pixman_region32_t opaque;
/* wl_surface.set_buffer_transform */
/* wl_surface.set_buffer_scale */
/* wp_viewport.set_source */
/* wp_viewport.set_destination */
surface->buffer_viewport = state->buffer_viewport;
/* wl_surface.attach */
if (state->newly_attached)
weston_surface_attach(surface, state->buffer);
weston_surface_state_set_buffer(state, NULL);
weston_surface_build_buffer_matrix(surface,
&surface->surface_to_buffer_matrix);
weston_matrix_invert(&surface->buffer_to_surface_matrix,
&surface->surface_to_buffer_matrix);
if (state->newly_attached || state->buffer_viewport.changed) {
weston_surface_update_size(surface);
if (surface->committed)
surface->committed(surface, state->sx, state->sy);
}
state->sx = 0;
state->sy = 0;
state->newly_attached = 0;
state->buffer_viewport.changed = 0;
/* wl_surface.damage and wl_surface.damage_buffer */
compositor: Implement JSON-timeline logging Logging is activated and deactivated with the debug key binding 't'. When activated, it creates a new log file, where it records the events. The log file contains events and detailed object information entries in JSON format, and is meant to be parsed in sequence from beginning to the end. The emitted events are mostly related to the output repaint cycle, like when repaint begins, is submitted to GPU, and when it completes on a vblank. This is recorded per-output. Also some per-surface events are recorded, including when surface damage is flushed. To reduce the log size, events refer to objects like outputs and surfaces by id numbers. Detailed object information is emitted only as needed: on the first object occurrence, and afterwards only if weston_timeline_object::force_refresh asks for it. The detailed information for surfaces includes the string returned by weston_surface::get_label. Therefore it is important to set weston_timeline_object::force_refresh = 1 whenever the string would change, so that the new details get recorded. A rudimentary parser and SVG generator can be found at: https://github.com/ppaalanen/wesgr The timeline logs can answer questions including: - How does the compositor repaint cycle work timing-wise? - When was the vblank deadline missed? - What is the latency from surface commit to showing the new content on screen? - How long does it take to process the scenegraph? v2: weston_surface::get_description renamed to get_label. v3: reafctor a bit into fprint_quoted_string(). Signed-off-by: Pekka Paalanen <pekka.paalanen@collabora.co.uk>
2014-11-12 16:09:24 +03:00
if (weston_timeline_enabled_ &&
(pixman_region32_not_empty(&state->damage_surface) ||
pixman_region32_not_empty(&state->damage_buffer)))
compositor: Implement JSON-timeline logging Logging is activated and deactivated with the debug key binding 't'. When activated, it creates a new log file, where it records the events. The log file contains events and detailed object information entries in JSON format, and is meant to be parsed in sequence from beginning to the end. The emitted events are mostly related to the output repaint cycle, like when repaint begins, is submitted to GPU, and when it completes on a vblank. This is recorded per-output. Also some per-surface events are recorded, including when surface damage is flushed. To reduce the log size, events refer to objects like outputs and surfaces by id numbers. Detailed object information is emitted only as needed: on the first object occurrence, and afterwards only if weston_timeline_object::force_refresh asks for it. The detailed information for surfaces includes the string returned by weston_surface::get_label. Therefore it is important to set weston_timeline_object::force_refresh = 1 whenever the string would change, so that the new details get recorded. A rudimentary parser and SVG generator can be found at: https://github.com/ppaalanen/wesgr The timeline logs can answer questions including: - How does the compositor repaint cycle work timing-wise? - When was the vblank deadline missed? - What is the latency from surface commit to showing the new content on screen? - How long does it take to process the scenegraph? v2: weston_surface::get_description renamed to get_label. v3: reafctor a bit into fprint_quoted_string(). Signed-off-by: Pekka Paalanen <pekka.paalanen@collabora.co.uk>
2014-11-12 16:09:24 +03:00
TL_POINT("core_commit_damage", TLP_SURFACE(surface), TLP_END);
pixman_region32_union(&surface->damage, &surface->damage,
&state->damage_surface);
apply_damage_buffer(&surface->damage, surface, state);
pixman_region32_intersect_rect(&surface->damage, &surface->damage,
0, 0, surface->width, surface->height);
pixman_region32_clear(&state->damage_surface);
/* wl_surface.set_opaque_region */
pixman_region32_init(&opaque);
pixman_region32_intersect_rect(&opaque, &state->opaque,
0, 0, surface->width, surface->height);
if (!pixman_region32_equal(&opaque, &surface->opaque)) {
pixman_region32_copy(&surface->opaque, &opaque);
Split the geometry information from weston_surface out into weston_view The weston_surface structure is split into two structures: * The weston_surface structure storres everything required for a client-side or server-side surface. This includes buffers; callbacks; backend private data; input, damage, and opaque regions; and a few other bookkeeping bits. * The weston_view structure represents an entity in the scenegraph and storres all of the geometry information. This includes clip region, alpha, position, and the transformation list as well as all of the temporary information derived from the geometry state. Because a view, and not a surface, is a scenegraph element, the view is what is placed in layers and planes. There are a few things worth noting about the surface/view split: 1. This is *not* a modification to the protocol. It is, instead, a modification to Weston's internal scenegraph to allow a single surface to exist in multiple places at a time. Clients are completely unaware of how many views to a particular surface exist. 2. A view is considered a direct child of a surface and is destroyed when the surface is destroyed. Because of this, the view.surface pointer is always valid and non-null. 3. The compositor's surface_list is replaced with a view_list. Due to subsurfaces, building the view list is a little more complicated than it used to be and involves building a tree of views on the fly whenever subsurfaces are used. However, this means that backends can remain completely subsurface-agnostic. 4. Surfaces and views both keep track of which outputs they are on. 5. The weston_surface structure now has width and height fields. These are populated when a new buffer is attached before surface.configure is called. This is because there are many surface-based operations that really require the width and height and digging through the views didn't work well. Signed-off-by: Jason Ekstrand <jason@jlekstrand.net>
2013-10-13 07:38:11 +04:00
wl_list_for_each(view, &surface->views, surface_link)
weston_view_geometry_dirty(view);
}
pixman_region32_fini(&opaque);
/* wl_surface.set_input_region */
pixman_region32_intersect_rect(&surface->input, &state->input,
0, 0, surface->width, surface->height);
/* wl_surface.frame */
wl_list_insert_list(&surface->frame_callback_list,
&state->frame_callback_list);
wl_list_init(&state->frame_callback_list);
/* XXX:
* What should happen with a feedback request, if there
* is no wl_buffer attached for this commit?
*/
/* presentation.feedback */
wl_list_insert_list(&surface->feedback_list,
&state->feedback_list);
wl_list_init(&state->feedback_list);
wl_signal_emit(&surface->commit_signal, surface);
}
static void
weston_surface_commit(struct weston_surface *surface)
{
weston_surface_commit_state(surface, &surface->pending);
compositor: introduce sub-surfaces Implement the basic protocol for sub-surfaces: - expose wl_subcompositor global interface - error checking on protocol calls - associate a parent wl_surface to a sub-surface - introduce the sub-surface role, which is exclusive - an implementation of the wl_subsurface interface - allow nesting of sub-surfaces - proper surface transformation inheritance from parent to sub-surfaces - two different modes of wl_surface.commit for sub-surfaces - hook sub-surfaces up to repaint by modifying the repaint list code Struct weston_subsurface is dynamically allocated. For sub-surfaces, it is completely populated. For parent surfaces, weston_subsurface acts only as a link for stacking order purposes. The wl_resource is unused, parent_destroy_listener is not registered, the transform is not linked, etc. Sub-surfaces are not added directly into layers for display or input. Instead, they are hooked up via the sub-surface list present in parent weston_surface. This way sub-surfaces are inherently linked to the parent surface, and cannot be displayed unless the parent is mapped, too. This also eases restacking, as only the parent will be in a layer list. Also, only the main surface should be subject to shell actions. The surface list rebuilding in weston_output_repaint() is modified to process sub-surface lists, if they are non-empty. The sub-surface list always contains the parent, too, unless empty. The collection of frame_callback_list is moved to a later loop, to streamline the surface list rebuild functions. Features still lacking are: - full-surface alpha support for compound windows Changes in v2: - fix a bug in surface mapping: commit a sub-surface would cause the main surface to never be mapped. - remove debug printfs - detect attempt of making a surface its own parent - always zero-alloc weston_subsurface - apply wl_subsurface.set_position in commit, not immediately - add weston_surface_to_subsurface() - implement sub-surface commit modes parent-cached and independent - implement wl_subcompositor.destroy and wl_subsurface.destroy Changes in v3: - rebased, and use the new transform inheritance code - squashed the commit "add sub-surfaces to repaint list" - fixed a buffer reference leak in commit_from_cache() - Rewrite the sub-surface destructor code, and make it leave the wl_subsurface protocol object inert, if one destroys the corresponding wl_surface. - replaced set_commit_mode with set_sync and set_desync - allowed sub-surface nesting, and fixed repaint accordingly - implemented nested sub-surface commit modes - Made the sub-surface order changes from wl_subsurface.place_above and .place_below to be applied when the parent surface state is applied, instead of immediately. This conforms with the protocol specification now. Signed-off-by: Pekka Paalanen <ppaalanen@gmail.com>
2013-04-25 14:57:42 +04:00
weston_surface_commit_subsurface_order(surface);
weston_surface_schedule_repaint(surface);
}
compositor: introduce sub-surfaces Implement the basic protocol for sub-surfaces: - expose wl_subcompositor global interface - error checking on protocol calls - associate a parent wl_surface to a sub-surface - introduce the sub-surface role, which is exclusive - an implementation of the wl_subsurface interface - allow nesting of sub-surfaces - proper surface transformation inheritance from parent to sub-surfaces - two different modes of wl_surface.commit for sub-surfaces - hook sub-surfaces up to repaint by modifying the repaint list code Struct weston_subsurface is dynamically allocated. For sub-surfaces, it is completely populated. For parent surfaces, weston_subsurface acts only as a link for stacking order purposes. The wl_resource is unused, parent_destroy_listener is not registered, the transform is not linked, etc. Sub-surfaces are not added directly into layers for display or input. Instead, they are hooked up via the sub-surface list present in parent weston_surface. This way sub-surfaces are inherently linked to the parent surface, and cannot be displayed unless the parent is mapped, too. This also eases restacking, as only the parent will be in a layer list. Also, only the main surface should be subject to shell actions. The surface list rebuilding in weston_output_repaint() is modified to process sub-surface lists, if they are non-empty. The sub-surface list always contains the parent, too, unless empty. The collection of frame_callback_list is moved to a later loop, to streamline the surface list rebuild functions. Features still lacking are: - full-surface alpha support for compound windows Changes in v2: - fix a bug in surface mapping: commit a sub-surface would cause the main surface to never be mapped. - remove debug printfs - detect attempt of making a surface its own parent - always zero-alloc weston_subsurface - apply wl_subsurface.set_position in commit, not immediately - add weston_surface_to_subsurface() - implement sub-surface commit modes parent-cached and independent - implement wl_subcompositor.destroy and wl_subsurface.destroy Changes in v3: - rebased, and use the new transform inheritance code - squashed the commit "add sub-surfaces to repaint list" - fixed a buffer reference leak in commit_from_cache() - Rewrite the sub-surface destructor code, and make it leave the wl_subsurface protocol object inert, if one destroys the corresponding wl_surface. - replaced set_commit_mode with set_sync and set_desync - allowed sub-surface nesting, and fixed repaint accordingly - implemented nested sub-surface commit modes - Made the sub-surface order changes from wl_subsurface.place_above and .place_below to be applied when the parent surface state is applied, instead of immediately. This conforms with the protocol specification now. Signed-off-by: Pekka Paalanen <ppaalanen@gmail.com>
2013-04-25 14:57:42 +04:00
static void
weston_subsurface_commit(struct weston_subsurface *sub);
static void
weston_subsurface_parent_commit(struct weston_subsurface *sub,
int parent_is_synchronized);
static void
surface_commit(struct wl_client *client, struct wl_resource *resource)
{
struct weston_surface *surface = wl_resource_get_user_data(resource);
compositor: introduce sub-surfaces Implement the basic protocol for sub-surfaces: - expose wl_subcompositor global interface - error checking on protocol calls - associate a parent wl_surface to a sub-surface - introduce the sub-surface role, which is exclusive - an implementation of the wl_subsurface interface - allow nesting of sub-surfaces - proper surface transformation inheritance from parent to sub-surfaces - two different modes of wl_surface.commit for sub-surfaces - hook sub-surfaces up to repaint by modifying the repaint list code Struct weston_subsurface is dynamically allocated. For sub-surfaces, it is completely populated. For parent surfaces, weston_subsurface acts only as a link for stacking order purposes. The wl_resource is unused, parent_destroy_listener is not registered, the transform is not linked, etc. Sub-surfaces are not added directly into layers for display or input. Instead, they are hooked up via the sub-surface list present in parent weston_surface. This way sub-surfaces are inherently linked to the parent surface, and cannot be displayed unless the parent is mapped, too. This also eases restacking, as only the parent will be in a layer list. Also, only the main surface should be subject to shell actions. The surface list rebuilding in weston_output_repaint() is modified to process sub-surface lists, if they are non-empty. The sub-surface list always contains the parent, too, unless empty. The collection of frame_callback_list is moved to a later loop, to streamline the surface list rebuild functions. Features still lacking are: - full-surface alpha support for compound windows Changes in v2: - fix a bug in surface mapping: commit a sub-surface would cause the main surface to never be mapped. - remove debug printfs - detect attempt of making a surface its own parent - always zero-alloc weston_subsurface - apply wl_subsurface.set_position in commit, not immediately - add weston_surface_to_subsurface() - implement sub-surface commit modes parent-cached and independent - implement wl_subcompositor.destroy and wl_subsurface.destroy Changes in v3: - rebased, and use the new transform inheritance code - squashed the commit "add sub-surfaces to repaint list" - fixed a buffer reference leak in commit_from_cache() - Rewrite the sub-surface destructor code, and make it leave the wl_subsurface protocol object inert, if one destroys the corresponding wl_surface. - replaced set_commit_mode with set_sync and set_desync - allowed sub-surface nesting, and fixed repaint accordingly - implemented nested sub-surface commit modes - Made the sub-surface order changes from wl_subsurface.place_above and .place_below to be applied when the parent surface state is applied, instead of immediately. This conforms with the protocol specification now. Signed-off-by: Pekka Paalanen <ppaalanen@gmail.com>
2013-04-25 14:57:42 +04:00
struct weston_subsurface *sub = weston_surface_to_subsurface(surface);
if (!weston_surface_is_pending_viewport_source_valid(surface)) {
assert(surface->viewport_resource);
wl_resource_post_error(surface->viewport_resource,
WP_VIEWPORT_ERROR_OUT_OF_BUFFER,
"wl_surface@%d has viewport source outside buffer",
wl_resource_get_id(resource));
return;
}
if (!weston_surface_is_pending_viewport_dst_size_int(surface)) {
assert(surface->viewport_resource);
wl_resource_post_error(surface->viewport_resource,
WP_VIEWPORT_ERROR_BAD_SIZE,
"wl_surface@%d viewport dst size not integer",
wl_resource_get_id(resource));
return;
}
compositor: introduce sub-surfaces Implement the basic protocol for sub-surfaces: - expose wl_subcompositor global interface - error checking on protocol calls - associate a parent wl_surface to a sub-surface - introduce the sub-surface role, which is exclusive - an implementation of the wl_subsurface interface - allow nesting of sub-surfaces - proper surface transformation inheritance from parent to sub-surfaces - two different modes of wl_surface.commit for sub-surfaces - hook sub-surfaces up to repaint by modifying the repaint list code Struct weston_subsurface is dynamically allocated. For sub-surfaces, it is completely populated. For parent surfaces, weston_subsurface acts only as a link for stacking order purposes. The wl_resource is unused, parent_destroy_listener is not registered, the transform is not linked, etc. Sub-surfaces are not added directly into layers for display or input. Instead, they are hooked up via the sub-surface list present in parent weston_surface. This way sub-surfaces are inherently linked to the parent surface, and cannot be displayed unless the parent is mapped, too. This also eases restacking, as only the parent will be in a layer list. Also, only the main surface should be subject to shell actions. The surface list rebuilding in weston_output_repaint() is modified to process sub-surface lists, if they are non-empty. The sub-surface list always contains the parent, too, unless empty. The collection of frame_callback_list is moved to a later loop, to streamline the surface list rebuild functions. Features still lacking are: - full-surface alpha support for compound windows Changes in v2: - fix a bug in surface mapping: commit a sub-surface would cause the main surface to never be mapped. - remove debug printfs - detect attempt of making a surface its own parent - always zero-alloc weston_subsurface - apply wl_subsurface.set_position in commit, not immediately - add weston_surface_to_subsurface() - implement sub-surface commit modes parent-cached and independent - implement wl_subcompositor.destroy and wl_subsurface.destroy Changes in v3: - rebased, and use the new transform inheritance code - squashed the commit "add sub-surfaces to repaint list" - fixed a buffer reference leak in commit_from_cache() - Rewrite the sub-surface destructor code, and make it leave the wl_subsurface protocol object inert, if one destroys the corresponding wl_surface. - replaced set_commit_mode with set_sync and set_desync - allowed sub-surface nesting, and fixed repaint accordingly - implemented nested sub-surface commit modes - Made the sub-surface order changes from wl_subsurface.place_above and .place_below to be applied when the parent surface state is applied, instead of immediately. This conforms with the protocol specification now. Signed-off-by: Pekka Paalanen <ppaalanen@gmail.com>
2013-04-25 14:57:42 +04:00
if (sub) {
weston_subsurface_commit(sub);
return;
}
weston_surface_commit(surface);
wl_list_for_each(sub, &surface->subsurface_list, parent_link) {
if (sub->surface != surface)
weston_subsurface_parent_commit(sub, 0);
}
}
static void
surface_set_buffer_transform(struct wl_client *client,
struct wl_resource *resource, int transform)
{
struct weston_surface *surface = wl_resource_get_user_data(resource);
/* if wl_output.transform grows more members this will need to be updated. */
if (transform < 0 ||
transform > WL_OUTPUT_TRANSFORM_FLIPPED_270) {
wl_resource_post_error(resource,
WL_SURFACE_ERROR_INVALID_TRANSFORM,
"buffer transform must be a valid transform "
"('%d' specified)", transform);
return;
}
surface->pending.buffer_viewport.buffer.transform = transform;
surface->pending.buffer_viewport.changed = 1;
}
static void
surface_set_buffer_scale(struct wl_client *client,
struct wl_resource *resource,
int32_t scale)
{
struct weston_surface *surface = wl_resource_get_user_data(resource);
if (scale < 1) {
wl_resource_post_error(resource,
WL_SURFACE_ERROR_INVALID_SCALE,
"buffer scale must be at least one "
"('%d' specified)", scale);
return;
}
surface->pending.buffer_viewport.buffer.scale = scale;
surface->pending.buffer_viewport.changed = 1;
}
static const struct wl_surface_interface surface_interface = {
surface_destroy,
surface_attach,
2011-08-17 07:01:28 +04:00
surface_damage,
surface_frame,
surface_set_opaque_region,
surface_set_input_region,
surface_commit,
surface_set_buffer_transform,
surface_set_buffer_scale,
surface_damage_buffer
};
static void
compositor_create_surface(struct wl_client *client,
struct wl_resource *resource, uint32_t id)
2008-11-07 16:39:37 +03:00
{
struct weston_compositor *ec = wl_resource_get_user_data(resource);
struct weston_surface *surface;
2008-11-07 16:39:37 +03:00
surface = weston_surface_create(ec);
if (surface == NULL) {
2011-09-01 17:54:57 +04:00
wl_resource_post_no_memory(resource);
return;
}
surface->resource =
wl_resource_create(client, &wl_surface_interface,
wl_resource_get_version(resource), id);
if (surface->resource == NULL) {
weston_surface_destroy(surface);
wl_resource_post_no_memory(resource);
return;
}
wl_resource_set_implementation(surface->resource, &surface_interface,
surface, destroy_surface);
wl_signal_emit(&ec->create_surface_signal, surface);
}
static void
destroy_region(struct wl_resource *resource)
{
struct weston_region *region = wl_resource_get_user_data(resource);
pixman_region32_fini(&region->region);
free(region);
}
static void
region_destroy(struct wl_client *client, struct wl_resource *resource)
{
wl_resource_destroy(resource);
}
static void
region_add(struct wl_client *client, struct wl_resource *resource,
int32_t x, int32_t y, int32_t width, int32_t height)
{
struct weston_region *region = wl_resource_get_user_data(resource);
pixman_region32_union_rect(&region->region, &region->region,
x, y, width, height);
}
static void
region_subtract(struct wl_client *client, struct wl_resource *resource,
int32_t x, int32_t y, int32_t width, int32_t height)
{
struct weston_region *region = wl_resource_get_user_data(resource);
pixman_region32_t rect;
pixman_region32_init_rect(&rect, x, y, width, height);
pixman_region32_subtract(&region->region, &region->region, &rect);
pixman_region32_fini(&rect);
}
static const struct wl_region_interface region_interface = {
region_destroy,
region_add,
region_subtract
};
static void
compositor_create_region(struct wl_client *client,
struct wl_resource *resource, uint32_t id)
{
struct weston_region *region;
region = malloc(sizeof *region);
if (region == NULL) {
wl_resource_post_no_memory(resource);
return;
}
pixman_region32_init(&region->region);
region->resource =
wl_resource_create(client, &wl_region_interface, 1, id);
if (region->resource == NULL) {
free(region);
wl_resource_post_no_memory(resource);
return;
}
wl_resource_set_implementation(region->resource, &region_interface,
region, destroy_region);
}
static const struct wl_compositor_interface compositor_interface = {
compositor_create_surface,
compositor_create_region
};
compositor: introduce sub-surfaces Implement the basic protocol for sub-surfaces: - expose wl_subcompositor global interface - error checking on protocol calls - associate a parent wl_surface to a sub-surface - introduce the sub-surface role, which is exclusive - an implementation of the wl_subsurface interface - allow nesting of sub-surfaces - proper surface transformation inheritance from parent to sub-surfaces - two different modes of wl_surface.commit for sub-surfaces - hook sub-surfaces up to repaint by modifying the repaint list code Struct weston_subsurface is dynamically allocated. For sub-surfaces, it is completely populated. For parent surfaces, weston_subsurface acts only as a link for stacking order purposes. The wl_resource is unused, parent_destroy_listener is not registered, the transform is not linked, etc. Sub-surfaces are not added directly into layers for display or input. Instead, they are hooked up via the sub-surface list present in parent weston_surface. This way sub-surfaces are inherently linked to the parent surface, and cannot be displayed unless the parent is mapped, too. This also eases restacking, as only the parent will be in a layer list. Also, only the main surface should be subject to shell actions. The surface list rebuilding in weston_output_repaint() is modified to process sub-surface lists, if they are non-empty. The sub-surface list always contains the parent, too, unless empty. The collection of frame_callback_list is moved to a later loop, to streamline the surface list rebuild functions. Features still lacking are: - full-surface alpha support for compound windows Changes in v2: - fix a bug in surface mapping: commit a sub-surface would cause the main surface to never be mapped. - remove debug printfs - detect attempt of making a surface its own parent - always zero-alloc weston_subsurface - apply wl_subsurface.set_position in commit, not immediately - add weston_surface_to_subsurface() - implement sub-surface commit modes parent-cached and independent - implement wl_subcompositor.destroy and wl_subsurface.destroy Changes in v3: - rebased, and use the new transform inheritance code - squashed the commit "add sub-surfaces to repaint list" - fixed a buffer reference leak in commit_from_cache() - Rewrite the sub-surface destructor code, and make it leave the wl_subsurface protocol object inert, if one destroys the corresponding wl_surface. - replaced set_commit_mode with set_sync and set_desync - allowed sub-surface nesting, and fixed repaint accordingly - implemented nested sub-surface commit modes - Made the sub-surface order changes from wl_subsurface.place_above and .place_below to be applied when the parent surface state is applied, instead of immediately. This conforms with the protocol specification now. Signed-off-by: Pekka Paalanen <ppaalanen@gmail.com>
2013-04-25 14:57:42 +04:00
static void
weston_subsurface_commit_from_cache(struct weston_subsurface *sub)
{
struct weston_surface *surface = sub->surface;
weston_surface_commit_state(surface, &sub->cached);
weston_buffer_reference(&sub->cached_buffer_ref, NULL);
compositor: introduce sub-surfaces Implement the basic protocol for sub-surfaces: - expose wl_subcompositor global interface - error checking on protocol calls - associate a parent wl_surface to a sub-surface - introduce the sub-surface role, which is exclusive - an implementation of the wl_subsurface interface - allow nesting of sub-surfaces - proper surface transformation inheritance from parent to sub-surfaces - two different modes of wl_surface.commit for sub-surfaces - hook sub-surfaces up to repaint by modifying the repaint list code Struct weston_subsurface is dynamically allocated. For sub-surfaces, it is completely populated. For parent surfaces, weston_subsurface acts only as a link for stacking order purposes. The wl_resource is unused, parent_destroy_listener is not registered, the transform is not linked, etc. Sub-surfaces are not added directly into layers for display or input. Instead, they are hooked up via the sub-surface list present in parent weston_surface. This way sub-surfaces are inherently linked to the parent surface, and cannot be displayed unless the parent is mapped, too. This also eases restacking, as only the parent will be in a layer list. Also, only the main surface should be subject to shell actions. The surface list rebuilding in weston_output_repaint() is modified to process sub-surface lists, if they are non-empty. The sub-surface list always contains the parent, too, unless empty. The collection of frame_callback_list is moved to a later loop, to streamline the surface list rebuild functions. Features still lacking are: - full-surface alpha support for compound windows Changes in v2: - fix a bug in surface mapping: commit a sub-surface would cause the main surface to never be mapped. - remove debug printfs - detect attempt of making a surface its own parent - always zero-alloc weston_subsurface - apply wl_subsurface.set_position in commit, not immediately - add weston_surface_to_subsurface() - implement sub-surface commit modes parent-cached and independent - implement wl_subcompositor.destroy and wl_subsurface.destroy Changes in v3: - rebased, and use the new transform inheritance code - squashed the commit "add sub-surfaces to repaint list" - fixed a buffer reference leak in commit_from_cache() - Rewrite the sub-surface destructor code, and make it leave the wl_subsurface protocol object inert, if one destroys the corresponding wl_surface. - replaced set_commit_mode with set_sync and set_desync - allowed sub-surface nesting, and fixed repaint accordingly - implemented nested sub-surface commit modes - Made the sub-surface order changes from wl_subsurface.place_above and .place_below to be applied when the parent surface state is applied, instead of immediately. This conforms with the protocol specification now. Signed-off-by: Pekka Paalanen <ppaalanen@gmail.com>
2013-04-25 14:57:42 +04:00
weston_surface_commit_subsurface_order(surface);
weston_surface_schedule_repaint(surface);
sub->has_cached_data = 0;
compositor: introduce sub-surfaces Implement the basic protocol for sub-surfaces: - expose wl_subcompositor global interface - error checking on protocol calls - associate a parent wl_surface to a sub-surface - introduce the sub-surface role, which is exclusive - an implementation of the wl_subsurface interface - allow nesting of sub-surfaces - proper surface transformation inheritance from parent to sub-surfaces - two different modes of wl_surface.commit for sub-surfaces - hook sub-surfaces up to repaint by modifying the repaint list code Struct weston_subsurface is dynamically allocated. For sub-surfaces, it is completely populated. For parent surfaces, weston_subsurface acts only as a link for stacking order purposes. The wl_resource is unused, parent_destroy_listener is not registered, the transform is not linked, etc. Sub-surfaces are not added directly into layers for display or input. Instead, they are hooked up via the sub-surface list present in parent weston_surface. This way sub-surfaces are inherently linked to the parent surface, and cannot be displayed unless the parent is mapped, too. This also eases restacking, as only the parent will be in a layer list. Also, only the main surface should be subject to shell actions. The surface list rebuilding in weston_output_repaint() is modified to process sub-surface lists, if they are non-empty. The sub-surface list always contains the parent, too, unless empty. The collection of frame_callback_list is moved to a later loop, to streamline the surface list rebuild functions. Features still lacking are: - full-surface alpha support for compound windows Changes in v2: - fix a bug in surface mapping: commit a sub-surface would cause the main surface to never be mapped. - remove debug printfs - detect attempt of making a surface its own parent - always zero-alloc weston_subsurface - apply wl_subsurface.set_position in commit, not immediately - add weston_surface_to_subsurface() - implement sub-surface commit modes parent-cached and independent - implement wl_subcompositor.destroy and wl_subsurface.destroy Changes in v3: - rebased, and use the new transform inheritance code - squashed the commit "add sub-surfaces to repaint list" - fixed a buffer reference leak in commit_from_cache() - Rewrite the sub-surface destructor code, and make it leave the wl_subsurface protocol object inert, if one destroys the corresponding wl_surface. - replaced set_commit_mode with set_sync and set_desync - allowed sub-surface nesting, and fixed repaint accordingly - implemented nested sub-surface commit modes - Made the sub-surface order changes from wl_subsurface.place_above and .place_below to be applied when the parent surface state is applied, instead of immediately. This conforms with the protocol specification now. Signed-off-by: Pekka Paalanen <ppaalanen@gmail.com>
2013-04-25 14:57:42 +04:00
}
static void
weston_subsurface_commit_to_cache(struct weston_subsurface *sub)
{
struct weston_surface *surface = sub->surface;
/*
* If this commit would cause the surface to move by the
* attach(dx, dy) parameters, the old damage region must be
* translated to correspond to the new surface coordinate system
* origin.
compositor: introduce sub-surfaces Implement the basic protocol for sub-surfaces: - expose wl_subcompositor global interface - error checking on protocol calls - associate a parent wl_surface to a sub-surface - introduce the sub-surface role, which is exclusive - an implementation of the wl_subsurface interface - allow nesting of sub-surfaces - proper surface transformation inheritance from parent to sub-surfaces - two different modes of wl_surface.commit for sub-surfaces - hook sub-surfaces up to repaint by modifying the repaint list code Struct weston_subsurface is dynamically allocated. For sub-surfaces, it is completely populated. For parent surfaces, weston_subsurface acts only as a link for stacking order purposes. The wl_resource is unused, parent_destroy_listener is not registered, the transform is not linked, etc. Sub-surfaces are not added directly into layers for display or input. Instead, they are hooked up via the sub-surface list present in parent weston_surface. This way sub-surfaces are inherently linked to the parent surface, and cannot be displayed unless the parent is mapped, too. This also eases restacking, as only the parent will be in a layer list. Also, only the main surface should be subject to shell actions. The surface list rebuilding in weston_output_repaint() is modified to process sub-surface lists, if they are non-empty. The sub-surface list always contains the parent, too, unless empty. The collection of frame_callback_list is moved to a later loop, to streamline the surface list rebuild functions. Features still lacking are: - full-surface alpha support for compound windows Changes in v2: - fix a bug in surface mapping: commit a sub-surface would cause the main surface to never be mapped. - remove debug printfs - detect attempt of making a surface its own parent - always zero-alloc weston_subsurface - apply wl_subsurface.set_position in commit, not immediately - add weston_surface_to_subsurface() - implement sub-surface commit modes parent-cached and independent - implement wl_subcompositor.destroy and wl_subsurface.destroy Changes in v3: - rebased, and use the new transform inheritance code - squashed the commit "add sub-surfaces to repaint list" - fixed a buffer reference leak in commit_from_cache() - Rewrite the sub-surface destructor code, and make it leave the wl_subsurface protocol object inert, if one destroys the corresponding wl_surface. - replaced set_commit_mode with set_sync and set_desync - allowed sub-surface nesting, and fixed repaint accordingly - implemented nested sub-surface commit modes - Made the sub-surface order changes from wl_subsurface.place_above and .place_below to be applied when the parent surface state is applied, instead of immediately. This conforms with the protocol specification now. Signed-off-by: Pekka Paalanen <ppaalanen@gmail.com>
2013-04-25 14:57:42 +04:00
*/
pixman_region32_translate(&sub->cached.damage_surface,
compositor: introduce sub-surfaces Implement the basic protocol for sub-surfaces: - expose wl_subcompositor global interface - error checking on protocol calls - associate a parent wl_surface to a sub-surface - introduce the sub-surface role, which is exclusive - an implementation of the wl_subsurface interface - allow nesting of sub-surfaces - proper surface transformation inheritance from parent to sub-surfaces - two different modes of wl_surface.commit for sub-surfaces - hook sub-surfaces up to repaint by modifying the repaint list code Struct weston_subsurface is dynamically allocated. For sub-surfaces, it is completely populated. For parent surfaces, weston_subsurface acts only as a link for stacking order purposes. The wl_resource is unused, parent_destroy_listener is not registered, the transform is not linked, etc. Sub-surfaces are not added directly into layers for display or input. Instead, they are hooked up via the sub-surface list present in parent weston_surface. This way sub-surfaces are inherently linked to the parent surface, and cannot be displayed unless the parent is mapped, too. This also eases restacking, as only the parent will be in a layer list. Also, only the main surface should be subject to shell actions. The surface list rebuilding in weston_output_repaint() is modified to process sub-surface lists, if they are non-empty. The sub-surface list always contains the parent, too, unless empty. The collection of frame_callback_list is moved to a later loop, to streamline the surface list rebuild functions. Features still lacking are: - full-surface alpha support for compound windows Changes in v2: - fix a bug in surface mapping: commit a sub-surface would cause the main surface to never be mapped. - remove debug printfs - detect attempt of making a surface its own parent - always zero-alloc weston_subsurface - apply wl_subsurface.set_position in commit, not immediately - add weston_surface_to_subsurface() - implement sub-surface commit modes parent-cached and independent - implement wl_subcompositor.destroy and wl_subsurface.destroy Changes in v3: - rebased, and use the new transform inheritance code - squashed the commit "add sub-surfaces to repaint list" - fixed a buffer reference leak in commit_from_cache() - Rewrite the sub-surface destructor code, and make it leave the wl_subsurface protocol object inert, if one destroys the corresponding wl_surface. - replaced set_commit_mode with set_sync and set_desync - allowed sub-surface nesting, and fixed repaint accordingly - implemented nested sub-surface commit modes - Made the sub-surface order changes from wl_subsurface.place_above and .place_below to be applied when the parent surface state is applied, instead of immediately. This conforms with the protocol specification now. Signed-off-by: Pekka Paalanen <ppaalanen@gmail.com>
2013-04-25 14:57:42 +04:00
-surface->pending.sx, -surface->pending.sy);
pixman_region32_union(&sub->cached.damage_surface,
&sub->cached.damage_surface,
&surface->pending.damage_surface);
pixman_region32_clear(&surface->pending.damage_surface);
compositor: introduce sub-surfaces Implement the basic protocol for sub-surfaces: - expose wl_subcompositor global interface - error checking on protocol calls - associate a parent wl_surface to a sub-surface - introduce the sub-surface role, which is exclusive - an implementation of the wl_subsurface interface - allow nesting of sub-surfaces - proper surface transformation inheritance from parent to sub-surfaces - two different modes of wl_surface.commit for sub-surfaces - hook sub-surfaces up to repaint by modifying the repaint list code Struct weston_subsurface is dynamically allocated. For sub-surfaces, it is completely populated. For parent surfaces, weston_subsurface acts only as a link for stacking order purposes. The wl_resource is unused, parent_destroy_listener is not registered, the transform is not linked, etc. Sub-surfaces are not added directly into layers for display or input. Instead, they are hooked up via the sub-surface list present in parent weston_surface. This way sub-surfaces are inherently linked to the parent surface, and cannot be displayed unless the parent is mapped, too. This also eases restacking, as only the parent will be in a layer list. Also, only the main surface should be subject to shell actions. The surface list rebuilding in weston_output_repaint() is modified to process sub-surface lists, if they are non-empty. The sub-surface list always contains the parent, too, unless empty. The collection of frame_callback_list is moved to a later loop, to streamline the surface list rebuild functions. Features still lacking are: - full-surface alpha support for compound windows Changes in v2: - fix a bug in surface mapping: commit a sub-surface would cause the main surface to never be mapped. - remove debug printfs - detect attempt of making a surface its own parent - always zero-alloc weston_subsurface - apply wl_subsurface.set_position in commit, not immediately - add weston_surface_to_subsurface() - implement sub-surface commit modes parent-cached and independent - implement wl_subcompositor.destroy and wl_subsurface.destroy Changes in v3: - rebased, and use the new transform inheritance code - squashed the commit "add sub-surfaces to repaint list" - fixed a buffer reference leak in commit_from_cache() - Rewrite the sub-surface destructor code, and make it leave the wl_subsurface protocol object inert, if one destroys the corresponding wl_surface. - replaced set_commit_mode with set_sync and set_desync - allowed sub-surface nesting, and fixed repaint accordingly - implemented nested sub-surface commit modes - Made the sub-surface order changes from wl_subsurface.place_above and .place_below to be applied when the parent surface state is applied, instead of immediately. This conforms with the protocol specification now. Signed-off-by: Pekka Paalanen <ppaalanen@gmail.com>
2013-04-25 14:57:42 +04:00
if (surface->pending.newly_attached) {
sub->cached.newly_attached = 1;
weston_surface_state_set_buffer(&sub->cached,
surface->pending.buffer);
weston_buffer_reference(&sub->cached_buffer_ref,
compositor: introduce sub-surfaces Implement the basic protocol for sub-surfaces: - expose wl_subcompositor global interface - error checking on protocol calls - associate a parent wl_surface to a sub-surface - introduce the sub-surface role, which is exclusive - an implementation of the wl_subsurface interface - allow nesting of sub-surfaces - proper surface transformation inheritance from parent to sub-surfaces - two different modes of wl_surface.commit for sub-surfaces - hook sub-surfaces up to repaint by modifying the repaint list code Struct weston_subsurface is dynamically allocated. For sub-surfaces, it is completely populated. For parent surfaces, weston_subsurface acts only as a link for stacking order purposes. The wl_resource is unused, parent_destroy_listener is not registered, the transform is not linked, etc. Sub-surfaces are not added directly into layers for display or input. Instead, they are hooked up via the sub-surface list present in parent weston_surface. This way sub-surfaces are inherently linked to the parent surface, and cannot be displayed unless the parent is mapped, too. This also eases restacking, as only the parent will be in a layer list. Also, only the main surface should be subject to shell actions. The surface list rebuilding in weston_output_repaint() is modified to process sub-surface lists, if they are non-empty. The sub-surface list always contains the parent, too, unless empty. The collection of frame_callback_list is moved to a later loop, to streamline the surface list rebuild functions. Features still lacking are: - full-surface alpha support for compound windows Changes in v2: - fix a bug in surface mapping: commit a sub-surface would cause the main surface to never be mapped. - remove debug printfs - detect attempt of making a surface its own parent - always zero-alloc weston_subsurface - apply wl_subsurface.set_position in commit, not immediately - add weston_surface_to_subsurface() - implement sub-surface commit modes parent-cached and independent - implement wl_subcompositor.destroy and wl_subsurface.destroy Changes in v3: - rebased, and use the new transform inheritance code - squashed the commit "add sub-surfaces to repaint list" - fixed a buffer reference leak in commit_from_cache() - Rewrite the sub-surface destructor code, and make it leave the wl_subsurface protocol object inert, if one destroys the corresponding wl_surface. - replaced set_commit_mode with set_sync and set_desync - allowed sub-surface nesting, and fixed repaint accordingly - implemented nested sub-surface commit modes - Made the sub-surface order changes from wl_subsurface.place_above and .place_below to be applied when the parent surface state is applied, instead of immediately. This conforms with the protocol specification now. Signed-off-by: Pekka Paalanen <ppaalanen@gmail.com>
2013-04-25 14:57:42 +04:00
surface->pending.buffer);
weston_presentation_feedback_discard_list(
&sub->cached.feedback_list);
compositor: introduce sub-surfaces Implement the basic protocol for sub-surfaces: - expose wl_subcompositor global interface - error checking on protocol calls - associate a parent wl_surface to a sub-surface - introduce the sub-surface role, which is exclusive - an implementation of the wl_subsurface interface - allow nesting of sub-surfaces - proper surface transformation inheritance from parent to sub-surfaces - two different modes of wl_surface.commit for sub-surfaces - hook sub-surfaces up to repaint by modifying the repaint list code Struct weston_subsurface is dynamically allocated. For sub-surfaces, it is completely populated. For parent surfaces, weston_subsurface acts only as a link for stacking order purposes. The wl_resource is unused, parent_destroy_listener is not registered, the transform is not linked, etc. Sub-surfaces are not added directly into layers for display or input. Instead, they are hooked up via the sub-surface list present in parent weston_surface. This way sub-surfaces are inherently linked to the parent surface, and cannot be displayed unless the parent is mapped, too. This also eases restacking, as only the parent will be in a layer list. Also, only the main surface should be subject to shell actions. The surface list rebuilding in weston_output_repaint() is modified to process sub-surface lists, if they are non-empty. The sub-surface list always contains the parent, too, unless empty. The collection of frame_callback_list is moved to a later loop, to streamline the surface list rebuild functions. Features still lacking are: - full-surface alpha support for compound windows Changes in v2: - fix a bug in surface mapping: commit a sub-surface would cause the main surface to never be mapped. - remove debug printfs - detect attempt of making a surface its own parent - always zero-alloc weston_subsurface - apply wl_subsurface.set_position in commit, not immediately - add weston_surface_to_subsurface() - implement sub-surface commit modes parent-cached and independent - implement wl_subcompositor.destroy and wl_subsurface.destroy Changes in v3: - rebased, and use the new transform inheritance code - squashed the commit "add sub-surfaces to repaint list" - fixed a buffer reference leak in commit_from_cache() - Rewrite the sub-surface destructor code, and make it leave the wl_subsurface protocol object inert, if one destroys the corresponding wl_surface. - replaced set_commit_mode with set_sync and set_desync - allowed sub-surface nesting, and fixed repaint accordingly - implemented nested sub-surface commit modes - Made the sub-surface order changes from wl_subsurface.place_above and .place_below to be applied when the parent surface state is applied, instead of immediately. This conforms with the protocol specification now. Signed-off-by: Pekka Paalanen <ppaalanen@gmail.com>
2013-04-25 14:57:42 +04:00
}
sub->cached.sx += surface->pending.sx;
sub->cached.sy += surface->pending.sy;
apply_damage_buffer(&sub->cached.damage_surface, surface, &surface->pending);
sub->cached.buffer_viewport.changed |=
surface->pending.buffer_viewport.changed;
sub->cached.buffer_viewport.buffer =
surface->pending.buffer_viewport.buffer;
sub->cached.buffer_viewport.surface =
surface->pending.buffer_viewport.surface;
compositor: introduce sub-surfaces Implement the basic protocol for sub-surfaces: - expose wl_subcompositor global interface - error checking on protocol calls - associate a parent wl_surface to a sub-surface - introduce the sub-surface role, which is exclusive - an implementation of the wl_subsurface interface - allow nesting of sub-surfaces - proper surface transformation inheritance from parent to sub-surfaces - two different modes of wl_surface.commit for sub-surfaces - hook sub-surfaces up to repaint by modifying the repaint list code Struct weston_subsurface is dynamically allocated. For sub-surfaces, it is completely populated. For parent surfaces, weston_subsurface acts only as a link for stacking order purposes. The wl_resource is unused, parent_destroy_listener is not registered, the transform is not linked, etc. Sub-surfaces are not added directly into layers for display or input. Instead, they are hooked up via the sub-surface list present in parent weston_surface. This way sub-surfaces are inherently linked to the parent surface, and cannot be displayed unless the parent is mapped, too. This also eases restacking, as only the parent will be in a layer list. Also, only the main surface should be subject to shell actions. The surface list rebuilding in weston_output_repaint() is modified to process sub-surface lists, if they are non-empty. The sub-surface list always contains the parent, too, unless empty. The collection of frame_callback_list is moved to a later loop, to streamline the surface list rebuild functions. Features still lacking are: - full-surface alpha support for compound windows Changes in v2: - fix a bug in surface mapping: commit a sub-surface would cause the main surface to never be mapped. - remove debug printfs - detect attempt of making a surface its own parent - always zero-alloc weston_subsurface - apply wl_subsurface.set_position in commit, not immediately - add weston_surface_to_subsurface() - implement sub-surface commit modes parent-cached and independent - implement wl_subcompositor.destroy and wl_subsurface.destroy Changes in v3: - rebased, and use the new transform inheritance code - squashed the commit "add sub-surfaces to repaint list" - fixed a buffer reference leak in commit_from_cache() - Rewrite the sub-surface destructor code, and make it leave the wl_subsurface protocol object inert, if one destroys the corresponding wl_surface. - replaced set_commit_mode with set_sync and set_desync - allowed sub-surface nesting, and fixed repaint accordingly - implemented nested sub-surface commit modes - Made the sub-surface order changes from wl_subsurface.place_above and .place_below to be applied when the parent surface state is applied, instead of immediately. This conforms with the protocol specification now. Signed-off-by: Pekka Paalanen <ppaalanen@gmail.com>
2013-04-25 14:57:42 +04:00
weston_surface_reset_pending_buffer(surface);
compositor: introduce sub-surfaces Implement the basic protocol for sub-surfaces: - expose wl_subcompositor global interface - error checking on protocol calls - associate a parent wl_surface to a sub-surface - introduce the sub-surface role, which is exclusive - an implementation of the wl_subsurface interface - allow nesting of sub-surfaces - proper surface transformation inheritance from parent to sub-surfaces - two different modes of wl_surface.commit for sub-surfaces - hook sub-surfaces up to repaint by modifying the repaint list code Struct weston_subsurface is dynamically allocated. For sub-surfaces, it is completely populated. For parent surfaces, weston_subsurface acts only as a link for stacking order purposes. The wl_resource is unused, parent_destroy_listener is not registered, the transform is not linked, etc. Sub-surfaces are not added directly into layers for display or input. Instead, they are hooked up via the sub-surface list present in parent weston_surface. This way sub-surfaces are inherently linked to the parent surface, and cannot be displayed unless the parent is mapped, too. This also eases restacking, as only the parent will be in a layer list. Also, only the main surface should be subject to shell actions. The surface list rebuilding in weston_output_repaint() is modified to process sub-surface lists, if they are non-empty. The sub-surface list always contains the parent, too, unless empty. The collection of frame_callback_list is moved to a later loop, to streamline the surface list rebuild functions. Features still lacking are: - full-surface alpha support for compound windows Changes in v2: - fix a bug in surface mapping: commit a sub-surface would cause the main surface to never be mapped. - remove debug printfs - detect attempt of making a surface its own parent - always zero-alloc weston_subsurface - apply wl_subsurface.set_position in commit, not immediately - add weston_surface_to_subsurface() - implement sub-surface commit modes parent-cached and independent - implement wl_subcompositor.destroy and wl_subsurface.destroy Changes in v3: - rebased, and use the new transform inheritance code - squashed the commit "add sub-surfaces to repaint list" - fixed a buffer reference leak in commit_from_cache() - Rewrite the sub-surface destructor code, and make it leave the wl_subsurface protocol object inert, if one destroys the corresponding wl_surface. - replaced set_commit_mode with set_sync and set_desync - allowed sub-surface nesting, and fixed repaint accordingly - implemented nested sub-surface commit modes - Made the sub-surface order changes from wl_subsurface.place_above and .place_below to be applied when the parent surface state is applied, instead of immediately. This conforms with the protocol specification now. Signed-off-by: Pekka Paalanen <ppaalanen@gmail.com>
2013-04-25 14:57:42 +04:00
pixman_region32_copy(&sub->cached.opaque, &surface->pending.opaque);
pixman_region32_copy(&sub->cached.input, &surface->pending.input);
wl_list_insert_list(&sub->cached.frame_callback_list,
&surface->pending.frame_callback_list);
wl_list_init(&surface->pending.frame_callback_list);
wl_list_insert_list(&sub->cached.feedback_list,
&surface->pending.feedback_list);
wl_list_init(&surface->pending.feedback_list);
sub->has_cached_data = 1;
compositor: introduce sub-surfaces Implement the basic protocol for sub-surfaces: - expose wl_subcompositor global interface - error checking on protocol calls - associate a parent wl_surface to a sub-surface - introduce the sub-surface role, which is exclusive - an implementation of the wl_subsurface interface - allow nesting of sub-surfaces - proper surface transformation inheritance from parent to sub-surfaces - two different modes of wl_surface.commit for sub-surfaces - hook sub-surfaces up to repaint by modifying the repaint list code Struct weston_subsurface is dynamically allocated. For sub-surfaces, it is completely populated. For parent surfaces, weston_subsurface acts only as a link for stacking order purposes. The wl_resource is unused, parent_destroy_listener is not registered, the transform is not linked, etc. Sub-surfaces are not added directly into layers for display or input. Instead, they are hooked up via the sub-surface list present in parent weston_surface. This way sub-surfaces are inherently linked to the parent surface, and cannot be displayed unless the parent is mapped, too. This also eases restacking, as only the parent will be in a layer list. Also, only the main surface should be subject to shell actions. The surface list rebuilding in weston_output_repaint() is modified to process sub-surface lists, if they are non-empty. The sub-surface list always contains the parent, too, unless empty. The collection of frame_callback_list is moved to a later loop, to streamline the surface list rebuild functions. Features still lacking are: - full-surface alpha support for compound windows Changes in v2: - fix a bug in surface mapping: commit a sub-surface would cause the main surface to never be mapped. - remove debug printfs - detect attempt of making a surface its own parent - always zero-alloc weston_subsurface - apply wl_subsurface.set_position in commit, not immediately - add weston_surface_to_subsurface() - implement sub-surface commit modes parent-cached and independent - implement wl_subcompositor.destroy and wl_subsurface.destroy Changes in v3: - rebased, and use the new transform inheritance code - squashed the commit "add sub-surfaces to repaint list" - fixed a buffer reference leak in commit_from_cache() - Rewrite the sub-surface destructor code, and make it leave the wl_subsurface protocol object inert, if one destroys the corresponding wl_surface. - replaced set_commit_mode with set_sync and set_desync - allowed sub-surface nesting, and fixed repaint accordingly - implemented nested sub-surface commit modes - Made the sub-surface order changes from wl_subsurface.place_above and .place_below to be applied when the parent surface state is applied, instead of immediately. This conforms with the protocol specification now. Signed-off-by: Pekka Paalanen <ppaalanen@gmail.com>
2013-04-25 14:57:42 +04:00
}
static bool
compositor: introduce sub-surfaces Implement the basic protocol for sub-surfaces: - expose wl_subcompositor global interface - error checking on protocol calls - associate a parent wl_surface to a sub-surface - introduce the sub-surface role, which is exclusive - an implementation of the wl_subsurface interface - allow nesting of sub-surfaces - proper surface transformation inheritance from parent to sub-surfaces - two different modes of wl_surface.commit for sub-surfaces - hook sub-surfaces up to repaint by modifying the repaint list code Struct weston_subsurface is dynamically allocated. For sub-surfaces, it is completely populated. For parent surfaces, weston_subsurface acts only as a link for stacking order purposes. The wl_resource is unused, parent_destroy_listener is not registered, the transform is not linked, etc. Sub-surfaces are not added directly into layers for display or input. Instead, they are hooked up via the sub-surface list present in parent weston_surface. This way sub-surfaces are inherently linked to the parent surface, and cannot be displayed unless the parent is mapped, too. This also eases restacking, as only the parent will be in a layer list. Also, only the main surface should be subject to shell actions. The surface list rebuilding in weston_output_repaint() is modified to process sub-surface lists, if they are non-empty. The sub-surface list always contains the parent, too, unless empty. The collection of frame_callback_list is moved to a later loop, to streamline the surface list rebuild functions. Features still lacking are: - full-surface alpha support for compound windows Changes in v2: - fix a bug in surface mapping: commit a sub-surface would cause the main surface to never be mapped. - remove debug printfs - detect attempt of making a surface its own parent - always zero-alloc weston_subsurface - apply wl_subsurface.set_position in commit, not immediately - add weston_surface_to_subsurface() - implement sub-surface commit modes parent-cached and independent - implement wl_subcompositor.destroy and wl_subsurface.destroy Changes in v3: - rebased, and use the new transform inheritance code - squashed the commit "add sub-surfaces to repaint list" - fixed a buffer reference leak in commit_from_cache() - Rewrite the sub-surface destructor code, and make it leave the wl_subsurface protocol object inert, if one destroys the corresponding wl_surface. - replaced set_commit_mode with set_sync and set_desync - allowed sub-surface nesting, and fixed repaint accordingly - implemented nested sub-surface commit modes - Made the sub-surface order changes from wl_subsurface.place_above and .place_below to be applied when the parent surface state is applied, instead of immediately. This conforms with the protocol specification now. Signed-off-by: Pekka Paalanen <ppaalanen@gmail.com>
2013-04-25 14:57:42 +04:00
weston_subsurface_is_synchronized(struct weston_subsurface *sub)
{
while (sub) {
if (sub->synchronized)
return true;
compositor: introduce sub-surfaces Implement the basic protocol for sub-surfaces: - expose wl_subcompositor global interface - error checking on protocol calls - associate a parent wl_surface to a sub-surface - introduce the sub-surface role, which is exclusive - an implementation of the wl_subsurface interface - allow nesting of sub-surfaces - proper surface transformation inheritance from parent to sub-surfaces - two different modes of wl_surface.commit for sub-surfaces - hook sub-surfaces up to repaint by modifying the repaint list code Struct weston_subsurface is dynamically allocated. For sub-surfaces, it is completely populated. For parent surfaces, weston_subsurface acts only as a link for stacking order purposes. The wl_resource is unused, parent_destroy_listener is not registered, the transform is not linked, etc. Sub-surfaces are not added directly into layers for display or input. Instead, they are hooked up via the sub-surface list present in parent weston_surface. This way sub-surfaces are inherently linked to the parent surface, and cannot be displayed unless the parent is mapped, too. This also eases restacking, as only the parent will be in a layer list. Also, only the main surface should be subject to shell actions. The surface list rebuilding in weston_output_repaint() is modified to process sub-surface lists, if they are non-empty. The sub-surface list always contains the parent, too, unless empty. The collection of frame_callback_list is moved to a later loop, to streamline the surface list rebuild functions. Features still lacking are: - full-surface alpha support for compound windows Changes in v2: - fix a bug in surface mapping: commit a sub-surface would cause the main surface to never be mapped. - remove debug printfs - detect attempt of making a surface its own parent - always zero-alloc weston_subsurface - apply wl_subsurface.set_position in commit, not immediately - add weston_surface_to_subsurface() - implement sub-surface commit modes parent-cached and independent - implement wl_subcompositor.destroy and wl_subsurface.destroy Changes in v3: - rebased, and use the new transform inheritance code - squashed the commit "add sub-surfaces to repaint list" - fixed a buffer reference leak in commit_from_cache() - Rewrite the sub-surface destructor code, and make it leave the wl_subsurface protocol object inert, if one destroys the corresponding wl_surface. - replaced set_commit_mode with set_sync and set_desync - allowed sub-surface nesting, and fixed repaint accordingly - implemented nested sub-surface commit modes - Made the sub-surface order changes from wl_subsurface.place_above and .place_below to be applied when the parent surface state is applied, instead of immediately. This conforms with the protocol specification now. Signed-off-by: Pekka Paalanen <ppaalanen@gmail.com>
2013-04-25 14:57:42 +04:00
if (!sub->parent)
return false;
compositor: introduce sub-surfaces Implement the basic protocol for sub-surfaces: - expose wl_subcompositor global interface - error checking on protocol calls - associate a parent wl_surface to a sub-surface - introduce the sub-surface role, which is exclusive - an implementation of the wl_subsurface interface - allow nesting of sub-surfaces - proper surface transformation inheritance from parent to sub-surfaces - two different modes of wl_surface.commit for sub-surfaces - hook sub-surfaces up to repaint by modifying the repaint list code Struct weston_subsurface is dynamically allocated. For sub-surfaces, it is completely populated. For parent surfaces, weston_subsurface acts only as a link for stacking order purposes. The wl_resource is unused, parent_destroy_listener is not registered, the transform is not linked, etc. Sub-surfaces are not added directly into layers for display or input. Instead, they are hooked up via the sub-surface list present in parent weston_surface. This way sub-surfaces are inherently linked to the parent surface, and cannot be displayed unless the parent is mapped, too. This also eases restacking, as only the parent will be in a layer list. Also, only the main surface should be subject to shell actions. The surface list rebuilding in weston_output_repaint() is modified to process sub-surface lists, if they are non-empty. The sub-surface list always contains the parent, too, unless empty. The collection of frame_callback_list is moved to a later loop, to streamline the surface list rebuild functions. Features still lacking are: - full-surface alpha support for compound windows Changes in v2: - fix a bug in surface mapping: commit a sub-surface would cause the main surface to never be mapped. - remove debug printfs - detect attempt of making a surface its own parent - always zero-alloc weston_subsurface - apply wl_subsurface.set_position in commit, not immediately - add weston_surface_to_subsurface() - implement sub-surface commit modes parent-cached and independent - implement wl_subcompositor.destroy and wl_subsurface.destroy Changes in v3: - rebased, and use the new transform inheritance code - squashed the commit "add sub-surfaces to repaint list" - fixed a buffer reference leak in commit_from_cache() - Rewrite the sub-surface destructor code, and make it leave the wl_subsurface protocol object inert, if one destroys the corresponding wl_surface. - replaced set_commit_mode with set_sync and set_desync - allowed sub-surface nesting, and fixed repaint accordingly - implemented nested sub-surface commit modes - Made the sub-surface order changes from wl_subsurface.place_above and .place_below to be applied when the parent surface state is applied, instead of immediately. This conforms with the protocol specification now. Signed-off-by: Pekka Paalanen <ppaalanen@gmail.com>
2013-04-25 14:57:42 +04:00
sub = weston_surface_to_subsurface(sub->parent);
}
return false;
compositor: introduce sub-surfaces Implement the basic protocol for sub-surfaces: - expose wl_subcompositor global interface - error checking on protocol calls - associate a parent wl_surface to a sub-surface - introduce the sub-surface role, which is exclusive - an implementation of the wl_subsurface interface - allow nesting of sub-surfaces - proper surface transformation inheritance from parent to sub-surfaces - two different modes of wl_surface.commit for sub-surfaces - hook sub-surfaces up to repaint by modifying the repaint list code Struct weston_subsurface is dynamically allocated. For sub-surfaces, it is completely populated. For parent surfaces, weston_subsurface acts only as a link for stacking order purposes. The wl_resource is unused, parent_destroy_listener is not registered, the transform is not linked, etc. Sub-surfaces are not added directly into layers for display or input. Instead, they are hooked up via the sub-surface list present in parent weston_surface. This way sub-surfaces are inherently linked to the parent surface, and cannot be displayed unless the parent is mapped, too. This also eases restacking, as only the parent will be in a layer list. Also, only the main surface should be subject to shell actions. The surface list rebuilding in weston_output_repaint() is modified to process sub-surface lists, if they are non-empty. The sub-surface list always contains the parent, too, unless empty. The collection of frame_callback_list is moved to a later loop, to streamline the surface list rebuild functions. Features still lacking are: - full-surface alpha support for compound windows Changes in v2: - fix a bug in surface mapping: commit a sub-surface would cause the main surface to never be mapped. - remove debug printfs - detect attempt of making a surface its own parent - always zero-alloc weston_subsurface - apply wl_subsurface.set_position in commit, not immediately - add weston_surface_to_subsurface() - implement sub-surface commit modes parent-cached and independent - implement wl_subcompositor.destroy and wl_subsurface.destroy Changes in v3: - rebased, and use the new transform inheritance code - squashed the commit "add sub-surfaces to repaint list" - fixed a buffer reference leak in commit_from_cache() - Rewrite the sub-surface destructor code, and make it leave the wl_subsurface protocol object inert, if one destroys the corresponding wl_surface. - replaced set_commit_mode with set_sync and set_desync - allowed sub-surface nesting, and fixed repaint accordingly - implemented nested sub-surface commit modes - Made the sub-surface order changes from wl_subsurface.place_above and .place_below to be applied when the parent surface state is applied, instead of immediately. This conforms with the protocol specification now. Signed-off-by: Pekka Paalanen <ppaalanen@gmail.com>
2013-04-25 14:57:42 +04:00
}
static void
weston_subsurface_commit(struct weston_subsurface *sub)
{
struct weston_surface *surface = sub->surface;
struct weston_subsurface *tmp;
/* Recursive check for effectively synchronized. */
if (weston_subsurface_is_synchronized(sub)) {
weston_subsurface_commit_to_cache(sub);
} else {
if (sub->has_cached_data) {
compositor: introduce sub-surfaces Implement the basic protocol for sub-surfaces: - expose wl_subcompositor global interface - error checking on protocol calls - associate a parent wl_surface to a sub-surface - introduce the sub-surface role, which is exclusive - an implementation of the wl_subsurface interface - allow nesting of sub-surfaces - proper surface transformation inheritance from parent to sub-surfaces - two different modes of wl_surface.commit for sub-surfaces - hook sub-surfaces up to repaint by modifying the repaint list code Struct weston_subsurface is dynamically allocated. For sub-surfaces, it is completely populated. For parent surfaces, weston_subsurface acts only as a link for stacking order purposes. The wl_resource is unused, parent_destroy_listener is not registered, the transform is not linked, etc. Sub-surfaces are not added directly into layers for display or input. Instead, they are hooked up via the sub-surface list present in parent weston_surface. This way sub-surfaces are inherently linked to the parent surface, and cannot be displayed unless the parent is mapped, too. This also eases restacking, as only the parent will be in a layer list. Also, only the main surface should be subject to shell actions. The surface list rebuilding in weston_output_repaint() is modified to process sub-surface lists, if they are non-empty. The sub-surface list always contains the parent, too, unless empty. The collection of frame_callback_list is moved to a later loop, to streamline the surface list rebuild functions. Features still lacking are: - full-surface alpha support for compound windows Changes in v2: - fix a bug in surface mapping: commit a sub-surface would cause the main surface to never be mapped. - remove debug printfs - detect attempt of making a surface its own parent - always zero-alloc weston_subsurface - apply wl_subsurface.set_position in commit, not immediately - add weston_surface_to_subsurface() - implement sub-surface commit modes parent-cached and independent - implement wl_subcompositor.destroy and wl_subsurface.destroy Changes in v3: - rebased, and use the new transform inheritance code - squashed the commit "add sub-surfaces to repaint list" - fixed a buffer reference leak in commit_from_cache() - Rewrite the sub-surface destructor code, and make it leave the wl_subsurface protocol object inert, if one destroys the corresponding wl_surface. - replaced set_commit_mode with set_sync and set_desync - allowed sub-surface nesting, and fixed repaint accordingly - implemented nested sub-surface commit modes - Made the sub-surface order changes from wl_subsurface.place_above and .place_below to be applied when the parent surface state is applied, instead of immediately. This conforms with the protocol specification now. Signed-off-by: Pekka Paalanen <ppaalanen@gmail.com>
2013-04-25 14:57:42 +04:00
/* flush accumulated state from cache */
weston_subsurface_commit_to_cache(sub);
weston_subsurface_commit_from_cache(sub);
} else {
weston_surface_commit(surface);
}
wl_list_for_each(tmp, &surface->subsurface_list, parent_link) {
if (tmp->surface != surface)
weston_subsurface_parent_commit(tmp, 0);
}
}
}
static void
weston_subsurface_synchronized_commit(struct weston_subsurface *sub)
compositor: introduce sub-surfaces Implement the basic protocol for sub-surfaces: - expose wl_subcompositor global interface - error checking on protocol calls - associate a parent wl_surface to a sub-surface - introduce the sub-surface role, which is exclusive - an implementation of the wl_subsurface interface - allow nesting of sub-surfaces - proper surface transformation inheritance from parent to sub-surfaces - two different modes of wl_surface.commit for sub-surfaces - hook sub-surfaces up to repaint by modifying the repaint list code Struct weston_subsurface is dynamically allocated. For sub-surfaces, it is completely populated. For parent surfaces, weston_subsurface acts only as a link for stacking order purposes. The wl_resource is unused, parent_destroy_listener is not registered, the transform is not linked, etc. Sub-surfaces are not added directly into layers for display or input. Instead, they are hooked up via the sub-surface list present in parent weston_surface. This way sub-surfaces are inherently linked to the parent surface, and cannot be displayed unless the parent is mapped, too. This also eases restacking, as only the parent will be in a layer list. Also, only the main surface should be subject to shell actions. The surface list rebuilding in weston_output_repaint() is modified to process sub-surface lists, if they are non-empty. The sub-surface list always contains the parent, too, unless empty. The collection of frame_callback_list is moved to a later loop, to streamline the surface list rebuild functions. Features still lacking are: - full-surface alpha support for compound windows Changes in v2: - fix a bug in surface mapping: commit a sub-surface would cause the main surface to never be mapped. - remove debug printfs - detect attempt of making a surface its own parent - always zero-alloc weston_subsurface - apply wl_subsurface.set_position in commit, not immediately - add weston_surface_to_subsurface() - implement sub-surface commit modes parent-cached and independent - implement wl_subcompositor.destroy and wl_subsurface.destroy Changes in v3: - rebased, and use the new transform inheritance code - squashed the commit "add sub-surfaces to repaint list" - fixed a buffer reference leak in commit_from_cache() - Rewrite the sub-surface destructor code, and make it leave the wl_subsurface protocol object inert, if one destroys the corresponding wl_surface. - replaced set_commit_mode with set_sync and set_desync - allowed sub-surface nesting, and fixed repaint accordingly - implemented nested sub-surface commit modes - Made the sub-surface order changes from wl_subsurface.place_above and .place_below to be applied when the parent surface state is applied, instead of immediately. This conforms with the protocol specification now. Signed-off-by: Pekka Paalanen <ppaalanen@gmail.com>
2013-04-25 14:57:42 +04:00
{
struct weston_surface *surface = sub->surface;
struct weston_subsurface *tmp;
/* From now on, commit_from_cache the whole sub-tree, regardless of
* the synchronized mode of each child. This sub-surface or some
* of its ancestors were synchronized, so we are synchronized
* all the way down.
*/
if (sub->has_cached_data)
compositor: introduce sub-surfaces Implement the basic protocol for sub-surfaces: - expose wl_subcompositor global interface - error checking on protocol calls - associate a parent wl_surface to a sub-surface - introduce the sub-surface role, which is exclusive - an implementation of the wl_subsurface interface - allow nesting of sub-surfaces - proper surface transformation inheritance from parent to sub-surfaces - two different modes of wl_surface.commit for sub-surfaces - hook sub-surfaces up to repaint by modifying the repaint list code Struct weston_subsurface is dynamically allocated. For sub-surfaces, it is completely populated. For parent surfaces, weston_subsurface acts only as a link for stacking order purposes. The wl_resource is unused, parent_destroy_listener is not registered, the transform is not linked, etc. Sub-surfaces are not added directly into layers for display or input. Instead, they are hooked up via the sub-surface list present in parent weston_surface. This way sub-surfaces are inherently linked to the parent surface, and cannot be displayed unless the parent is mapped, too. This also eases restacking, as only the parent will be in a layer list. Also, only the main surface should be subject to shell actions. The surface list rebuilding in weston_output_repaint() is modified to process sub-surface lists, if they are non-empty. The sub-surface list always contains the parent, too, unless empty. The collection of frame_callback_list is moved to a later loop, to streamline the surface list rebuild functions. Features still lacking are: - full-surface alpha support for compound windows Changes in v2: - fix a bug in surface mapping: commit a sub-surface would cause the main surface to never be mapped. - remove debug printfs - detect attempt of making a surface its own parent - always zero-alloc weston_subsurface - apply wl_subsurface.set_position in commit, not immediately - add weston_surface_to_subsurface() - implement sub-surface commit modes parent-cached and independent - implement wl_subcompositor.destroy and wl_subsurface.destroy Changes in v3: - rebased, and use the new transform inheritance code - squashed the commit "add sub-surfaces to repaint list" - fixed a buffer reference leak in commit_from_cache() - Rewrite the sub-surface destructor code, and make it leave the wl_subsurface protocol object inert, if one destroys the corresponding wl_surface. - replaced set_commit_mode with set_sync and set_desync - allowed sub-surface nesting, and fixed repaint accordingly - implemented nested sub-surface commit modes - Made the sub-surface order changes from wl_subsurface.place_above and .place_below to be applied when the parent surface state is applied, instead of immediately. This conforms with the protocol specification now. Signed-off-by: Pekka Paalanen <ppaalanen@gmail.com>
2013-04-25 14:57:42 +04:00
weston_subsurface_commit_from_cache(sub);
wl_list_for_each(tmp, &surface->subsurface_list, parent_link) {
if (tmp->surface != surface)
weston_subsurface_parent_commit(tmp, 1);
}
}
static void
weston_subsurface_parent_commit(struct weston_subsurface *sub,
int parent_is_synchronized)
{
Split the geometry information from weston_surface out into weston_view The weston_surface structure is split into two structures: * The weston_surface structure storres everything required for a client-side or server-side surface. This includes buffers; callbacks; backend private data; input, damage, and opaque regions; and a few other bookkeeping bits. * The weston_view structure represents an entity in the scenegraph and storres all of the geometry information. This includes clip region, alpha, position, and the transformation list as well as all of the temporary information derived from the geometry state. Because a view, and not a surface, is a scenegraph element, the view is what is placed in layers and planes. There are a few things worth noting about the surface/view split: 1. This is *not* a modification to the protocol. It is, instead, a modification to Weston's internal scenegraph to allow a single surface to exist in multiple places at a time. Clients are completely unaware of how many views to a particular surface exist. 2. A view is considered a direct child of a surface and is destroyed when the surface is destroyed. Because of this, the view.surface pointer is always valid and non-null. 3. The compositor's surface_list is replaced with a view_list. Due to subsurfaces, building the view list is a little more complicated than it used to be and involves building a tree of views on the fly whenever subsurfaces are used. However, this means that backends can remain completely subsurface-agnostic. 4. Surfaces and views both keep track of which outputs they are on. 5. The weston_surface structure now has width and height fields. These are populated when a new buffer is attached before surface.configure is called. This is because there are many surface-based operations that really require the width and height and digging through the views didn't work well. Signed-off-by: Jason Ekstrand <jason@jlekstrand.net>
2013-10-13 07:38:11 +04:00
struct weston_view *view;
if (sub->position.set) {
Split the geometry information from weston_surface out into weston_view The weston_surface structure is split into two structures: * The weston_surface structure storres everything required for a client-side or server-side surface. This includes buffers; callbacks; backend private data; input, damage, and opaque regions; and a few other bookkeeping bits. * The weston_view structure represents an entity in the scenegraph and storres all of the geometry information. This includes clip region, alpha, position, and the transformation list as well as all of the temporary information derived from the geometry state. Because a view, and not a surface, is a scenegraph element, the view is what is placed in layers and planes. There are a few things worth noting about the surface/view split: 1. This is *not* a modification to the protocol. It is, instead, a modification to Weston's internal scenegraph to allow a single surface to exist in multiple places at a time. Clients are completely unaware of how many views to a particular surface exist. 2. A view is considered a direct child of a surface and is destroyed when the surface is destroyed. Because of this, the view.surface pointer is always valid and non-null. 3. The compositor's surface_list is replaced with a view_list. Due to subsurfaces, building the view list is a little more complicated than it used to be and involves building a tree of views on the fly whenever subsurfaces are used. However, this means that backends can remain completely subsurface-agnostic. 4. Surfaces and views both keep track of which outputs they are on. 5. The weston_surface structure now has width and height fields. These are populated when a new buffer is attached before surface.configure is called. This is because there are many surface-based operations that really require the width and height and digging through the views didn't work well. Signed-off-by: Jason Ekstrand <jason@jlekstrand.net>
2013-10-13 07:38:11 +04:00
wl_list_for_each(view, &sub->surface->views, surface_link)
weston_view_set_position(view,
sub->position.x,
sub->position.y);
sub->position.set = 0;
}
if (parent_is_synchronized || sub->synchronized)
weston_subsurface_synchronized_commit(sub);
}
static int
subsurface_get_label(struct weston_surface *surface, char *buf, size_t len)
{
return snprintf(buf, len, "sub-surface");
}
compositor: introduce sub-surfaces Implement the basic protocol for sub-surfaces: - expose wl_subcompositor global interface - error checking on protocol calls - associate a parent wl_surface to a sub-surface - introduce the sub-surface role, which is exclusive - an implementation of the wl_subsurface interface - allow nesting of sub-surfaces - proper surface transformation inheritance from parent to sub-surfaces - two different modes of wl_surface.commit for sub-surfaces - hook sub-surfaces up to repaint by modifying the repaint list code Struct weston_subsurface is dynamically allocated. For sub-surfaces, it is completely populated. For parent surfaces, weston_subsurface acts only as a link for stacking order purposes. The wl_resource is unused, parent_destroy_listener is not registered, the transform is not linked, etc. Sub-surfaces are not added directly into layers for display or input. Instead, they are hooked up via the sub-surface list present in parent weston_surface. This way sub-surfaces are inherently linked to the parent surface, and cannot be displayed unless the parent is mapped, too. This also eases restacking, as only the parent will be in a layer list. Also, only the main surface should be subject to shell actions. The surface list rebuilding in weston_output_repaint() is modified to process sub-surface lists, if they are non-empty. The sub-surface list always contains the parent, too, unless empty. The collection of frame_callback_list is moved to a later loop, to streamline the surface list rebuild functions. Features still lacking are: - full-surface alpha support for compound windows Changes in v2: - fix a bug in surface mapping: commit a sub-surface would cause the main surface to never be mapped. - remove debug printfs - detect attempt of making a surface its own parent - always zero-alloc weston_subsurface - apply wl_subsurface.set_position in commit, not immediately - add weston_surface_to_subsurface() - implement sub-surface commit modes parent-cached and independent - implement wl_subcompositor.destroy and wl_subsurface.destroy Changes in v3: - rebased, and use the new transform inheritance code - squashed the commit "add sub-surfaces to repaint list" - fixed a buffer reference leak in commit_from_cache() - Rewrite the sub-surface destructor code, and make it leave the wl_subsurface protocol object inert, if one destroys the corresponding wl_surface. - replaced set_commit_mode with set_sync and set_desync - allowed sub-surface nesting, and fixed repaint accordingly - implemented nested sub-surface commit modes - Made the sub-surface order changes from wl_subsurface.place_above and .place_below to be applied when the parent surface state is applied, instead of immediately. This conforms with the protocol specification now. Signed-off-by: Pekka Paalanen <ppaalanen@gmail.com>
2013-04-25 14:57:42 +04:00
static void
subsurface_committed(struct weston_surface *surface, int32_t dx, int32_t dy)
compositor: introduce sub-surfaces Implement the basic protocol for sub-surfaces: - expose wl_subcompositor global interface - error checking on protocol calls - associate a parent wl_surface to a sub-surface - introduce the sub-surface role, which is exclusive - an implementation of the wl_subsurface interface - allow nesting of sub-surfaces - proper surface transformation inheritance from parent to sub-surfaces - two different modes of wl_surface.commit for sub-surfaces - hook sub-surfaces up to repaint by modifying the repaint list code Struct weston_subsurface is dynamically allocated. For sub-surfaces, it is completely populated. For parent surfaces, weston_subsurface acts only as a link for stacking order purposes. The wl_resource is unused, parent_destroy_listener is not registered, the transform is not linked, etc. Sub-surfaces are not added directly into layers for display or input. Instead, they are hooked up via the sub-surface list present in parent weston_surface. This way sub-surfaces are inherently linked to the parent surface, and cannot be displayed unless the parent is mapped, too. This also eases restacking, as only the parent will be in a layer list. Also, only the main surface should be subject to shell actions. The surface list rebuilding in weston_output_repaint() is modified to process sub-surface lists, if they are non-empty. The sub-surface list always contains the parent, too, unless empty. The collection of frame_callback_list is moved to a later loop, to streamline the surface list rebuild functions. Features still lacking are: - full-surface alpha support for compound windows Changes in v2: - fix a bug in surface mapping: commit a sub-surface would cause the main surface to never be mapped. - remove debug printfs - detect attempt of making a surface its own parent - always zero-alloc weston_subsurface - apply wl_subsurface.set_position in commit, not immediately - add weston_surface_to_subsurface() - implement sub-surface commit modes parent-cached and independent - implement wl_subcompositor.destroy and wl_subsurface.destroy Changes in v3: - rebased, and use the new transform inheritance code - squashed the commit "add sub-surfaces to repaint list" - fixed a buffer reference leak in commit_from_cache() - Rewrite the sub-surface destructor code, and make it leave the wl_subsurface protocol object inert, if one destroys the corresponding wl_surface. - replaced set_commit_mode with set_sync and set_desync - allowed sub-surface nesting, and fixed repaint accordingly - implemented nested sub-surface commit modes - Made the sub-surface order changes from wl_subsurface.place_above and .place_below to be applied when the parent surface state is applied, instead of immediately. This conforms with the protocol specification now. Signed-off-by: Pekka Paalanen <ppaalanen@gmail.com>
2013-04-25 14:57:42 +04:00
{
Split the geometry information from weston_surface out into weston_view The weston_surface structure is split into two structures: * The weston_surface structure storres everything required for a client-side or server-side surface. This includes buffers; callbacks; backend private data; input, damage, and opaque regions; and a few other bookkeeping bits. * The weston_view structure represents an entity in the scenegraph and storres all of the geometry information. This includes clip region, alpha, position, and the transformation list as well as all of the temporary information derived from the geometry state. Because a view, and not a surface, is a scenegraph element, the view is what is placed in layers and planes. There are a few things worth noting about the surface/view split: 1. This is *not* a modification to the protocol. It is, instead, a modification to Weston's internal scenegraph to allow a single surface to exist in multiple places at a time. Clients are completely unaware of how many views to a particular surface exist. 2. A view is considered a direct child of a surface and is destroyed when the surface is destroyed. Because of this, the view.surface pointer is always valid and non-null. 3. The compositor's surface_list is replaced with a view_list. Due to subsurfaces, building the view list is a little more complicated than it used to be and involves building a tree of views on the fly whenever subsurfaces are used. However, this means that backends can remain completely subsurface-agnostic. 4. Surfaces and views both keep track of which outputs they are on. 5. The weston_surface structure now has width and height fields. These are populated when a new buffer is attached before surface.configure is called. This is because there are many surface-based operations that really require the width and height and digging through the views didn't work well. Signed-off-by: Jason Ekstrand <jason@jlekstrand.net>
2013-10-13 07:38:11 +04:00
struct weston_view *view;
compositor: introduce sub-surfaces Implement the basic protocol for sub-surfaces: - expose wl_subcompositor global interface - error checking on protocol calls - associate a parent wl_surface to a sub-surface - introduce the sub-surface role, which is exclusive - an implementation of the wl_subsurface interface - allow nesting of sub-surfaces - proper surface transformation inheritance from parent to sub-surfaces - two different modes of wl_surface.commit for sub-surfaces - hook sub-surfaces up to repaint by modifying the repaint list code Struct weston_subsurface is dynamically allocated. For sub-surfaces, it is completely populated. For parent surfaces, weston_subsurface acts only as a link for stacking order purposes. The wl_resource is unused, parent_destroy_listener is not registered, the transform is not linked, etc. Sub-surfaces are not added directly into layers for display or input. Instead, they are hooked up via the sub-surface list present in parent weston_surface. This way sub-surfaces are inherently linked to the parent surface, and cannot be displayed unless the parent is mapped, too. This also eases restacking, as only the parent will be in a layer list. Also, only the main surface should be subject to shell actions. The surface list rebuilding in weston_output_repaint() is modified to process sub-surface lists, if they are non-empty. The sub-surface list always contains the parent, too, unless empty. The collection of frame_callback_list is moved to a later loop, to streamline the surface list rebuild functions. Features still lacking are: - full-surface alpha support for compound windows Changes in v2: - fix a bug in surface mapping: commit a sub-surface would cause the main surface to never be mapped. - remove debug printfs - detect attempt of making a surface its own parent - always zero-alloc weston_subsurface - apply wl_subsurface.set_position in commit, not immediately - add weston_surface_to_subsurface() - implement sub-surface commit modes parent-cached and independent - implement wl_subcompositor.destroy and wl_subsurface.destroy Changes in v3: - rebased, and use the new transform inheritance code - squashed the commit "add sub-surfaces to repaint list" - fixed a buffer reference leak in commit_from_cache() - Rewrite the sub-surface destructor code, and make it leave the wl_subsurface protocol object inert, if one destroys the corresponding wl_surface. - replaced set_commit_mode with set_sync and set_desync - allowed sub-surface nesting, and fixed repaint accordingly - implemented nested sub-surface commit modes - Made the sub-surface order changes from wl_subsurface.place_above and .place_below to be applied when the parent surface state is applied, instead of immediately. This conforms with the protocol specification now. Signed-off-by: Pekka Paalanen <ppaalanen@gmail.com>
2013-04-25 14:57:42 +04:00
Split the geometry information from weston_surface out into weston_view The weston_surface structure is split into two structures: * The weston_surface structure storres everything required for a client-side or server-side surface. This includes buffers; callbacks; backend private data; input, damage, and opaque regions; and a few other bookkeeping bits. * The weston_view structure represents an entity in the scenegraph and storres all of the geometry information. This includes clip region, alpha, position, and the transformation list as well as all of the temporary information derived from the geometry state. Because a view, and not a surface, is a scenegraph element, the view is what is placed in layers and planes. There are a few things worth noting about the surface/view split: 1. This is *not* a modification to the protocol. It is, instead, a modification to Weston's internal scenegraph to allow a single surface to exist in multiple places at a time. Clients are completely unaware of how many views to a particular surface exist. 2. A view is considered a direct child of a surface and is destroyed when the surface is destroyed. Because of this, the view.surface pointer is always valid and non-null. 3. The compositor's surface_list is replaced with a view_list. Due to subsurfaces, building the view list is a little more complicated than it used to be and involves building a tree of views on the fly whenever subsurfaces are used. However, this means that backends can remain completely subsurface-agnostic. 4. Surfaces and views both keep track of which outputs they are on. 5. The weston_surface structure now has width and height fields. These are populated when a new buffer is attached before surface.configure is called. This is because there are many surface-based operations that really require the width and height and digging through the views didn't work well. Signed-off-by: Jason Ekstrand <jason@jlekstrand.net>
2013-10-13 07:38:11 +04:00
wl_list_for_each(view, &surface->views, surface_link)
weston_view_set_position(view,
view->geometry.x + dx,
view->geometry.y + dy);
compositor: introduce sub-surfaces Implement the basic protocol for sub-surfaces: - expose wl_subcompositor global interface - error checking on protocol calls - associate a parent wl_surface to a sub-surface - introduce the sub-surface role, which is exclusive - an implementation of the wl_subsurface interface - allow nesting of sub-surfaces - proper surface transformation inheritance from parent to sub-surfaces - two different modes of wl_surface.commit for sub-surfaces - hook sub-surfaces up to repaint by modifying the repaint list code Struct weston_subsurface is dynamically allocated. For sub-surfaces, it is completely populated. For parent surfaces, weston_subsurface acts only as a link for stacking order purposes. The wl_resource is unused, parent_destroy_listener is not registered, the transform is not linked, etc. Sub-surfaces are not added directly into layers for display or input. Instead, they are hooked up via the sub-surface list present in parent weston_surface. This way sub-surfaces are inherently linked to the parent surface, and cannot be displayed unless the parent is mapped, too. This also eases restacking, as only the parent will be in a layer list. Also, only the main surface should be subject to shell actions. The surface list rebuilding in weston_output_repaint() is modified to process sub-surface lists, if they are non-empty. The sub-surface list always contains the parent, too, unless empty. The collection of frame_callback_list is moved to a later loop, to streamline the surface list rebuild functions. Features still lacking are: - full-surface alpha support for compound windows Changes in v2: - fix a bug in surface mapping: commit a sub-surface would cause the main surface to never be mapped. - remove debug printfs - detect attempt of making a surface its own parent - always zero-alloc weston_subsurface - apply wl_subsurface.set_position in commit, not immediately - add weston_surface_to_subsurface() - implement sub-surface commit modes parent-cached and independent - implement wl_subcompositor.destroy and wl_subsurface.destroy Changes in v3: - rebased, and use the new transform inheritance code - squashed the commit "add sub-surfaces to repaint list" - fixed a buffer reference leak in commit_from_cache() - Rewrite the sub-surface destructor code, and make it leave the wl_subsurface protocol object inert, if one destroys the corresponding wl_surface. - replaced set_commit_mode with set_sync and set_desync - allowed sub-surface nesting, and fixed repaint accordingly - implemented nested sub-surface commit modes - Made the sub-surface order changes from wl_subsurface.place_above and .place_below to be applied when the parent surface state is applied, instead of immediately. This conforms with the protocol specification now. Signed-off-by: Pekka Paalanen <ppaalanen@gmail.com>
2013-04-25 14:57:42 +04:00
/* No need to check parent mappedness, because if parent is not
* mapped, parent is not in a visible layer, so this sub-surface
* will not be drawn either.
*/
compositor: introduce sub-surfaces Implement the basic protocol for sub-surfaces: - expose wl_subcompositor global interface - error checking on protocol calls - associate a parent wl_surface to a sub-surface - introduce the sub-surface role, which is exclusive - an implementation of the wl_subsurface interface - allow nesting of sub-surfaces - proper surface transformation inheritance from parent to sub-surfaces - two different modes of wl_surface.commit for sub-surfaces - hook sub-surfaces up to repaint by modifying the repaint list code Struct weston_subsurface is dynamically allocated. For sub-surfaces, it is completely populated. For parent surfaces, weston_subsurface acts only as a link for stacking order purposes. The wl_resource is unused, parent_destroy_listener is not registered, the transform is not linked, etc. Sub-surfaces are not added directly into layers for display or input. Instead, they are hooked up via the sub-surface list present in parent weston_surface. This way sub-surfaces are inherently linked to the parent surface, and cannot be displayed unless the parent is mapped, too. This also eases restacking, as only the parent will be in a layer list. Also, only the main surface should be subject to shell actions. The surface list rebuilding in weston_output_repaint() is modified to process sub-surface lists, if they are non-empty. The sub-surface list always contains the parent, too, unless empty. The collection of frame_callback_list is moved to a later loop, to streamline the surface list rebuild functions. Features still lacking are: - full-surface alpha support for compound windows Changes in v2: - fix a bug in surface mapping: commit a sub-surface would cause the main surface to never be mapped. - remove debug printfs - detect attempt of making a surface its own parent - always zero-alloc weston_subsurface - apply wl_subsurface.set_position in commit, not immediately - add weston_surface_to_subsurface() - implement sub-surface commit modes parent-cached and independent - implement wl_subcompositor.destroy and wl_subsurface.destroy Changes in v3: - rebased, and use the new transform inheritance code - squashed the commit "add sub-surfaces to repaint list" - fixed a buffer reference leak in commit_from_cache() - Rewrite the sub-surface destructor code, and make it leave the wl_subsurface protocol object inert, if one destroys the corresponding wl_surface. - replaced set_commit_mode with set_sync and set_desync - allowed sub-surface nesting, and fixed repaint accordingly - implemented nested sub-surface commit modes - Made the sub-surface order changes from wl_subsurface.place_above and .place_below to be applied when the parent surface state is applied, instead of immediately. This conforms with the protocol specification now. Signed-off-by: Pekka Paalanen <ppaalanen@gmail.com>
2013-04-25 14:57:42 +04:00
if (!weston_surface_is_mapped(surface)) {
surface->is_mapped = true;
compositor: quick fix for sub-surface mapping If a client does this: 1. create a main window and map it 2. create a wl_surface, and make it a sub-surface of the main window 3. set the sub-surface to desync 4. commit content to the sub-surface to map it Then step 4 should cause the sub-surface to become mapped. However, Weston fails to schedule a repaint in that case, so the sub-surface will not appear until something else causes a repaint on that output, e.g. the main window. A quick and dirty fix is to set the output mask for the surface in Weston, which allows the repaint to be scheduled. This patch implements that, and might only work right on single-output systems. A proper fix would involve rewriting the whole "is surface mapped" mechanism in Weston, to not rely on output assignments but to have a separate flag for "mapped", and figuring out how to schedule repaints for the right outputs. Following is the actual protocol sequence used to trigger the problem: [3224648.125] -> wl_compositor@4.create_surface(new id wl_surface@3) [3224648.206] -> xdg_shell@7.get_xdg_surface(new id xdg_surface@8, wl_surface@3) [3224648.311] -> xdg_surface@8.set_title("simple-shm") [3224648.378] -> wl_surface@3.damage(0, 0, 250, 250) [3224649.888] -> wl_shm@6.create_pool(new id wl_shm_pool@9, fd 6, 250000) [3224650.031] -> wl_shm_pool@9.create_buffer(new id wl_buffer@10, 0, 250, 250, 1000, 1) [3224650.244] -> wl_shm_pool@9.destroy() [3224651.975] -> wl_surface@3.attach(wl_buffer@10, 0, 0) [3224652.100] -> wl_surface@3.damage(20, 20, 210, 210) [3224652.243] -> wl_surface@3.frame(new id wl_callback@11) [3224652.317] -> wl_surface@3.commit() [3228652.535] -> wl_compositor@4.create_surface(new id wl_surface@12) [3228652.610] -> wl_subcompositor@5.get_subsurface(new id wl_subsurface@13, wl_surface@12, wl_surface@3) [3228652.644] -> wl_subsurface@13.set_desync() [3228652.659] -> wl_subsurface@13.set_position(100, 100) [3228654.090] -> wl_shm@6.create_pool(new id wl_shm_pool@14, fd 6, 250000) [3228654.140] -> wl_shm_pool@14.create_buffer(new id wl_buffer@15, 0, 250, 250, 1000, 1) [3228654.180] -> wl_shm_pool@14.destroy() [3228654.408] -> wl_surface@12.attach(wl_buffer@15, 0, 0) [3228654.436] -> wl_surface@12.damage(0, 0, 250, 250) [3228654.462] -> wl_surface@12.commit() Signed-off-by: Pekka Paalanen <pekka.paalanen@collabora.co.uk> Cc: George Kiagiadakis <george.kiagiadakis@collabora.com> Cc: Jason Ekstrand <jason.ekstrand@intel.com>
2014-06-30 12:52:07 +04:00
/* Cannot call weston_view_update_transform(),
compositor: introduce sub-surfaces Implement the basic protocol for sub-surfaces: - expose wl_subcompositor global interface - error checking on protocol calls - associate a parent wl_surface to a sub-surface - introduce the sub-surface role, which is exclusive - an implementation of the wl_subsurface interface - allow nesting of sub-surfaces - proper surface transformation inheritance from parent to sub-surfaces - two different modes of wl_surface.commit for sub-surfaces - hook sub-surfaces up to repaint by modifying the repaint list code Struct weston_subsurface is dynamically allocated. For sub-surfaces, it is completely populated. For parent surfaces, weston_subsurface acts only as a link for stacking order purposes. The wl_resource is unused, parent_destroy_listener is not registered, the transform is not linked, etc. Sub-surfaces are not added directly into layers for display or input. Instead, they are hooked up via the sub-surface list present in parent weston_surface. This way sub-surfaces are inherently linked to the parent surface, and cannot be displayed unless the parent is mapped, too. This also eases restacking, as only the parent will be in a layer list. Also, only the main surface should be subject to shell actions. The surface list rebuilding in weston_output_repaint() is modified to process sub-surface lists, if they are non-empty. The sub-surface list always contains the parent, too, unless empty. The collection of frame_callback_list is moved to a later loop, to streamline the surface list rebuild functions. Features still lacking are: - full-surface alpha support for compound windows Changes in v2: - fix a bug in surface mapping: commit a sub-surface would cause the main surface to never be mapped. - remove debug printfs - detect attempt of making a surface its own parent - always zero-alloc weston_subsurface - apply wl_subsurface.set_position in commit, not immediately - add weston_surface_to_subsurface() - implement sub-surface commit modes parent-cached and independent - implement wl_subcompositor.destroy and wl_subsurface.destroy Changes in v3: - rebased, and use the new transform inheritance code - squashed the commit "add sub-surfaces to repaint list" - fixed a buffer reference leak in commit_from_cache() - Rewrite the sub-surface destructor code, and make it leave the wl_subsurface protocol object inert, if one destroys the corresponding wl_surface. - replaced set_commit_mode with set_sync and set_desync - allowed sub-surface nesting, and fixed repaint accordingly - implemented nested sub-surface commit modes - Made the sub-surface order changes from wl_subsurface.place_above and .place_below to be applied when the parent surface state is applied, instead of immediately. This conforms with the protocol specification now. Signed-off-by: Pekka Paalanen <ppaalanen@gmail.com>
2013-04-25 14:57:42 +04:00
* because that would call it also for the parent surface,
* which might not be mapped yet. That would lead to
* inconsistent state, where the window could never be
* mapped.
*
* Instead just force the is_mapped flag on, to make
compositor: introduce sub-surfaces Implement the basic protocol for sub-surfaces: - expose wl_subcompositor global interface - error checking on protocol calls - associate a parent wl_surface to a sub-surface - introduce the sub-surface role, which is exclusive - an implementation of the wl_subsurface interface - allow nesting of sub-surfaces - proper surface transformation inheritance from parent to sub-surfaces - two different modes of wl_surface.commit for sub-surfaces - hook sub-surfaces up to repaint by modifying the repaint list code Struct weston_subsurface is dynamically allocated. For sub-surfaces, it is completely populated. For parent surfaces, weston_subsurface acts only as a link for stacking order purposes. The wl_resource is unused, parent_destroy_listener is not registered, the transform is not linked, etc. Sub-surfaces are not added directly into layers for display or input. Instead, they are hooked up via the sub-surface list present in parent weston_surface. This way sub-surfaces are inherently linked to the parent surface, and cannot be displayed unless the parent is mapped, too. This also eases restacking, as only the parent will be in a layer list. Also, only the main surface should be subject to shell actions. The surface list rebuilding in weston_output_repaint() is modified to process sub-surface lists, if they are non-empty. The sub-surface list always contains the parent, too, unless empty. The collection of frame_callback_list is moved to a later loop, to streamline the surface list rebuild functions. Features still lacking are: - full-surface alpha support for compound windows Changes in v2: - fix a bug in surface mapping: commit a sub-surface would cause the main surface to never be mapped. - remove debug printfs - detect attempt of making a surface its own parent - always zero-alloc weston_subsurface - apply wl_subsurface.set_position in commit, not immediately - add weston_surface_to_subsurface() - implement sub-surface commit modes parent-cached and independent - implement wl_subcompositor.destroy and wl_subsurface.destroy Changes in v3: - rebased, and use the new transform inheritance code - squashed the commit "add sub-surfaces to repaint list" - fixed a buffer reference leak in commit_from_cache() - Rewrite the sub-surface destructor code, and make it leave the wl_subsurface protocol object inert, if one destroys the corresponding wl_surface. - replaced set_commit_mode with set_sync and set_desync - allowed sub-surface nesting, and fixed repaint accordingly - implemented nested sub-surface commit modes - Made the sub-surface order changes from wl_subsurface.place_above and .place_below to be applied when the parent surface state is applied, instead of immediately. This conforms with the protocol specification now. Signed-off-by: Pekka Paalanen <ppaalanen@gmail.com>
2013-04-25 14:57:42 +04:00
* weston_surface_is_mapped() return true, so that when the
* parent surface does get mapped, this one will get
compositor: quick fix for sub-surface mapping If a client does this: 1. create a main window and map it 2. create a wl_surface, and make it a sub-surface of the main window 3. set the sub-surface to desync 4. commit content to the sub-surface to map it Then step 4 should cause the sub-surface to become mapped. However, Weston fails to schedule a repaint in that case, so the sub-surface will not appear until something else causes a repaint on that output, e.g. the main window. A quick and dirty fix is to set the output mask for the surface in Weston, which allows the repaint to be scheduled. This patch implements that, and might only work right on single-output systems. A proper fix would involve rewriting the whole "is surface mapped" mechanism in Weston, to not rely on output assignments but to have a separate flag for "mapped", and figuring out how to schedule repaints for the right outputs. Following is the actual protocol sequence used to trigger the problem: [3224648.125] -> wl_compositor@4.create_surface(new id wl_surface@3) [3224648.206] -> xdg_shell@7.get_xdg_surface(new id xdg_surface@8, wl_surface@3) [3224648.311] -> xdg_surface@8.set_title("simple-shm") [3224648.378] -> wl_surface@3.damage(0, 0, 250, 250) [3224649.888] -> wl_shm@6.create_pool(new id wl_shm_pool@9, fd 6, 250000) [3224650.031] -> wl_shm_pool@9.create_buffer(new id wl_buffer@10, 0, 250, 250, 1000, 1) [3224650.244] -> wl_shm_pool@9.destroy() [3224651.975] -> wl_surface@3.attach(wl_buffer@10, 0, 0) [3224652.100] -> wl_surface@3.damage(20, 20, 210, 210) [3224652.243] -> wl_surface@3.frame(new id wl_callback@11) [3224652.317] -> wl_surface@3.commit() [3228652.535] -> wl_compositor@4.create_surface(new id wl_surface@12) [3228652.610] -> wl_subcompositor@5.get_subsurface(new id wl_subsurface@13, wl_surface@12, wl_surface@3) [3228652.644] -> wl_subsurface@13.set_desync() [3228652.659] -> wl_subsurface@13.set_position(100, 100) [3228654.090] -> wl_shm@6.create_pool(new id wl_shm_pool@14, fd 6, 250000) [3228654.140] -> wl_shm_pool@14.create_buffer(new id wl_buffer@15, 0, 250, 250, 1000, 1) [3228654.180] -> wl_shm_pool@14.destroy() [3228654.408] -> wl_surface@12.attach(wl_buffer@15, 0, 0) [3228654.436] -> wl_surface@12.damage(0, 0, 250, 250) [3228654.462] -> wl_surface@12.commit() Signed-off-by: Pekka Paalanen <pekka.paalanen@collabora.co.uk> Cc: George Kiagiadakis <george.kiagiadakis@collabora.com> Cc: Jason Ekstrand <jason.ekstrand@intel.com>
2014-06-30 12:52:07 +04:00
* included, too. See view_list_add().
compositor: introduce sub-surfaces Implement the basic protocol for sub-surfaces: - expose wl_subcompositor global interface - error checking on protocol calls - associate a parent wl_surface to a sub-surface - introduce the sub-surface role, which is exclusive - an implementation of the wl_subsurface interface - allow nesting of sub-surfaces - proper surface transformation inheritance from parent to sub-surfaces - two different modes of wl_surface.commit for sub-surfaces - hook sub-surfaces up to repaint by modifying the repaint list code Struct weston_subsurface is dynamically allocated. For sub-surfaces, it is completely populated. For parent surfaces, weston_subsurface acts only as a link for stacking order purposes. The wl_resource is unused, parent_destroy_listener is not registered, the transform is not linked, etc. Sub-surfaces are not added directly into layers for display or input. Instead, they are hooked up via the sub-surface list present in parent weston_surface. This way sub-surfaces are inherently linked to the parent surface, and cannot be displayed unless the parent is mapped, too. This also eases restacking, as only the parent will be in a layer list. Also, only the main surface should be subject to shell actions. The surface list rebuilding in weston_output_repaint() is modified to process sub-surface lists, if they are non-empty. The sub-surface list always contains the parent, too, unless empty. The collection of frame_callback_list is moved to a later loop, to streamline the surface list rebuild functions. Features still lacking are: - full-surface alpha support for compound windows Changes in v2: - fix a bug in surface mapping: commit a sub-surface would cause the main surface to never be mapped. - remove debug printfs - detect attempt of making a surface its own parent - always zero-alloc weston_subsurface - apply wl_subsurface.set_position in commit, not immediately - add weston_surface_to_subsurface() - implement sub-surface commit modes parent-cached and independent - implement wl_subcompositor.destroy and wl_subsurface.destroy Changes in v3: - rebased, and use the new transform inheritance code - squashed the commit "add sub-surfaces to repaint list" - fixed a buffer reference leak in commit_from_cache() - Rewrite the sub-surface destructor code, and make it leave the wl_subsurface protocol object inert, if one destroys the corresponding wl_surface. - replaced set_commit_mode with set_sync and set_desync - allowed sub-surface nesting, and fixed repaint accordingly - implemented nested sub-surface commit modes - Made the sub-surface order changes from wl_subsurface.place_above and .place_below to be applied when the parent surface state is applied, instead of immediately. This conforms with the protocol specification now. Signed-off-by: Pekka Paalanen <ppaalanen@gmail.com>
2013-04-25 14:57:42 +04:00
*/
}
}
static struct weston_subsurface *
weston_surface_to_subsurface(struct weston_surface *surface)
{
if (surface->committed == subsurface_committed)
return surface->committed_private;
compositor: introduce sub-surfaces Implement the basic protocol for sub-surfaces: - expose wl_subcompositor global interface - error checking on protocol calls - associate a parent wl_surface to a sub-surface - introduce the sub-surface role, which is exclusive - an implementation of the wl_subsurface interface - allow nesting of sub-surfaces - proper surface transformation inheritance from parent to sub-surfaces - two different modes of wl_surface.commit for sub-surfaces - hook sub-surfaces up to repaint by modifying the repaint list code Struct weston_subsurface is dynamically allocated. For sub-surfaces, it is completely populated. For parent surfaces, weston_subsurface acts only as a link for stacking order purposes. The wl_resource is unused, parent_destroy_listener is not registered, the transform is not linked, etc. Sub-surfaces are not added directly into layers for display or input. Instead, they are hooked up via the sub-surface list present in parent weston_surface. This way sub-surfaces are inherently linked to the parent surface, and cannot be displayed unless the parent is mapped, too. This also eases restacking, as only the parent will be in a layer list. Also, only the main surface should be subject to shell actions. The surface list rebuilding in weston_output_repaint() is modified to process sub-surface lists, if they are non-empty. The sub-surface list always contains the parent, too, unless empty. The collection of frame_callback_list is moved to a later loop, to streamline the surface list rebuild functions. Features still lacking are: - full-surface alpha support for compound windows Changes in v2: - fix a bug in surface mapping: commit a sub-surface would cause the main surface to never be mapped. - remove debug printfs - detect attempt of making a surface its own parent - always zero-alloc weston_subsurface - apply wl_subsurface.set_position in commit, not immediately - add weston_surface_to_subsurface() - implement sub-surface commit modes parent-cached and independent - implement wl_subcompositor.destroy and wl_subsurface.destroy Changes in v3: - rebased, and use the new transform inheritance code - squashed the commit "add sub-surfaces to repaint list" - fixed a buffer reference leak in commit_from_cache() - Rewrite the sub-surface destructor code, and make it leave the wl_subsurface protocol object inert, if one destroys the corresponding wl_surface. - replaced set_commit_mode with set_sync and set_desync - allowed sub-surface nesting, and fixed repaint accordingly - implemented nested sub-surface commit modes - Made the sub-surface order changes from wl_subsurface.place_above and .place_below to be applied when the parent surface state is applied, instead of immediately. This conforms with the protocol specification now. Signed-off-by: Pekka Paalanen <ppaalanen@gmail.com>
2013-04-25 14:57:42 +04:00
return NULL;
}
shell: keyboard focus and restacking fixes for sub-surfaces The shell needs to redirect some actions to the parent surface, when they originally target a sub-surface. This patch implements the following: - Move, resize, and rotate bindings always target the parent surface. - Opacity (full-surface alpha) binding targets the parent surface. This is broken, because it should change the opacity of the whole compound window, which is difficult to implement in the renderer. - click_to_activate_binding() needs to check the shell surface type from the main surface, because sub-surface would produce SHELL_SURFACE_NONE and prevent activation. - Also activate() needs to check the type from the main surface, and restack the main surface. Keyboard focus is assigned to the original (sub-)surface. - focus_state_surface_destroy() needs to handle sub-surfaces: only the main surface will be in a layer list. If the destroyed surface is indeed a sub-surface, activate the main surface next. This way a client that destroys a focused sub-surface still retains focus in the same window. - The workspace_manager.move_surface request can accept also sub-surfaces, and it will move the corresponding main surface. Changes in v2: - do not special-case keyboard focus for sub-surfaces - fix surface type checks for sub-surfaces in shell, fix restacking of sub-surfaces in shell, fix focus_state_surface_destroy() Changes in v3: - Renamed weston_surface_get_parent() to weston_surface_get_main_surface() to be more explicit that this is about sub-surfaces - Fixed move_surface_to_workspace() to handle keyboard focus on a sub-surface. - Used a temporary variable in several places to clarify code, instead of reassigning a variable. - Fixed workspace_manager_move_surface() to deal with sub-surfaces. Signed-off-by: Pekka Paalanen <ppaalanen@gmail.com>
2013-04-25 14:57:44 +04:00
WL_EXPORT struct weston_surface *
weston_surface_get_main_surface(struct weston_surface *surface)
{
struct weston_subsurface *sub;
while (surface && (sub = weston_surface_to_subsurface(surface)))
surface = sub->parent;
return surface;
}
2014-10-01 16:02:41 +04:00
WL_EXPORT int
weston_surface_set_role(struct weston_surface *surface,
const char *role_name,
struct wl_resource *error_resource,
uint32_t error_code)
{
assert(role_name);
if (surface->role_name == NULL ||
surface->role_name == role_name ||
strcmp(surface->role_name, role_name) == 0) {
surface->role_name = role_name;
return 0;
}
wl_resource_post_error(error_resource, error_code,
"Cannot assign role %s to wl_surface@%d,"
" already has role %s\n",
role_name,
wl_resource_get_id(surface->resource),
surface->role_name);
return -1;
}
WL_EXPORT const char *
weston_surface_get_role(struct weston_surface *surface)
{
return surface->role_name;
}
WL_EXPORT void
weston_surface_set_label_func(struct weston_surface *surface,
int (*desc)(struct weston_surface *,
char *, size_t))
{
surface->get_label = desc;
compositor: Implement JSON-timeline logging Logging is activated and deactivated with the debug key binding 't'. When activated, it creates a new log file, where it records the events. The log file contains events and detailed object information entries in JSON format, and is meant to be parsed in sequence from beginning to the end. The emitted events are mostly related to the output repaint cycle, like when repaint begins, is submitted to GPU, and when it completes on a vblank. This is recorded per-output. Also some per-surface events are recorded, including when surface damage is flushed. To reduce the log size, events refer to objects like outputs and surfaces by id numbers. Detailed object information is emitted only as needed: on the first object occurrence, and afterwards only if weston_timeline_object::force_refresh asks for it. The detailed information for surfaces includes the string returned by weston_surface::get_label. Therefore it is important to set weston_timeline_object::force_refresh = 1 whenever the string would change, so that the new details get recorded. A rudimentary parser and SVG generator can be found at: https://github.com/ppaalanen/wesgr The timeline logs can answer questions including: - How does the compositor repaint cycle work timing-wise? - When was the vblank deadline missed? - What is the latency from surface commit to showing the new content on screen? - How long does it take to process the scenegraph? v2: weston_surface::get_description renamed to get_label. v3: reafctor a bit into fprint_quoted_string(). Signed-off-by: Pekka Paalanen <pekka.paalanen@collabora.co.uk>
2014-11-12 16:09:24 +03:00
surface->timeline.force_refresh = 1;
}
/** Get the size of surface contents
*
* \param surface The surface to query.
* \param width Returns the width of raw contents.
* \param height Returns the height of raw contents.
*
* Retrieves the raw surface content size in pixels for the given surface.
* This is the whole content size in buffer pixels. If the surface
* has no content or the renderer does not implement this feature,
* zeroes are returned.
*
* This function is used to determine the buffer size needed for
* a weston_surface_copy_content() call.
*/
WL_EXPORT void
weston_surface_get_content_size(struct weston_surface *surface,
int *width, int *height)
{
struct weston_renderer *rer = surface->compositor->renderer;
if (!rer->surface_get_content_size) {
*width = 0;
*height = 0;
return;
}
rer->surface_get_content_size(surface, width, height);
}
/** Get the bounding box of a surface and its subsurfaces
*
* \param surface The surface to query.
* \return The bounding box relative to the surface origin.
*
*/
WL_EXPORT struct weston_geometry
weston_surface_get_bounding_box(struct weston_surface *surface)
{
pixman_region32_t region;
pixman_box32_t *box;
struct weston_subsurface *subsurface;
pixman_region32_init_rect(&region,
0, 0,
surface->width, surface->height);
wl_list_for_each(subsurface, &surface->subsurface_list, parent_link)
pixman_region32_union_rect(&region, &region,
subsurface->position.x,
subsurface->position.y,
subsurface->surface->width,
subsurface->surface->height);
box = pixman_region32_extents(&region);
struct weston_geometry geometry = {
.x = box->x1,
.y = box->y1,
.width = box->x2 - box->x1,
.height = box->y2 - box->y1,
};
pixman_region32_fini(&region);
return geometry;
}
/** Copy surface contents to system memory.
*
* \param surface The surface to copy from.
* \param target Pointer to the target memory buffer.
* \param size Size of the target buffer in bytes.
* \param src_x X location on contents to copy from.
* \param src_y Y location on contents to copy from.
* \param width Width in pixels of the area to copy.
* \param height Height in pixels of the area to copy.
* \return 0 for success, -1 for failure.
*
* Surface contents are maintained by the renderer. They can be in a
* reserved weston_buffer or as a copy, e.g. a GL texture, or something
* else.
*
* Surface contents are copied into memory pointed to by target,
* which has size bytes of space available. The target memory
* may be larger than needed, but being smaller returns an error.
* The extra bytes in target may or may not be written; their content is
* unspecified. Size must be large enough to hold the image.
*
* The image in the target memory will be arranged in rows from
* top to bottom, and pixels on a row from left to right. The pixel
* format is PIXMAN_a8b8g8r8, 4 bytes per pixel, and stride is exactly
* width * 4.
*
* Parameters src_x and src_y define the upper-left corner in buffer
* coordinates (pixels) to copy from. Parameters width and height
* define the size of the area to copy in pixels.
*
* The rectangle defined by src_x, src_y, width, height must fit in
* the surface contents. Otherwise an error is returned.
*
* Use surface_get_data_size to determine the content size; the
* needed target buffer size and rectangle limits.
*
* CURRENT IMPLEMENTATION RESTRICTIONS:
* - the machine must be little-endian due to Pixman formats.
*
* NOTE: Pixman formats are premultiplied.
*/
WL_EXPORT int
weston_surface_copy_content(struct weston_surface *surface,
void *target, size_t size,
int src_x, int src_y,
int width, int height)
{
struct weston_renderer *rer = surface->compositor->renderer;
int cw, ch;
const size_t bytespp = 4; /* PIXMAN_a8b8g8r8 */
if (!rer->surface_copy_content)
return -1;
weston_surface_get_content_size(surface, &cw, &ch);
if (src_x < 0 || src_y < 0)
return -1;
if (width <= 0 || height <= 0)
return -1;
if (src_x + width > cw || src_y + height > ch)
return -1;
if (width * bytespp * height > size)
return -1;
return rer->surface_copy_content(surface, target, size,
src_x, src_y, width, height);
}
compositor: introduce sub-surfaces Implement the basic protocol for sub-surfaces: - expose wl_subcompositor global interface - error checking on protocol calls - associate a parent wl_surface to a sub-surface - introduce the sub-surface role, which is exclusive - an implementation of the wl_subsurface interface - allow nesting of sub-surfaces - proper surface transformation inheritance from parent to sub-surfaces - two different modes of wl_surface.commit for sub-surfaces - hook sub-surfaces up to repaint by modifying the repaint list code Struct weston_subsurface is dynamically allocated. For sub-surfaces, it is completely populated. For parent surfaces, weston_subsurface acts only as a link for stacking order purposes. The wl_resource is unused, parent_destroy_listener is not registered, the transform is not linked, etc. Sub-surfaces are not added directly into layers for display or input. Instead, they are hooked up via the sub-surface list present in parent weston_surface. This way sub-surfaces are inherently linked to the parent surface, and cannot be displayed unless the parent is mapped, too. This also eases restacking, as only the parent will be in a layer list. Also, only the main surface should be subject to shell actions. The surface list rebuilding in weston_output_repaint() is modified to process sub-surface lists, if they are non-empty. The sub-surface list always contains the parent, too, unless empty. The collection of frame_callback_list is moved to a later loop, to streamline the surface list rebuild functions. Features still lacking are: - full-surface alpha support for compound windows Changes in v2: - fix a bug in surface mapping: commit a sub-surface would cause the main surface to never be mapped. - remove debug printfs - detect attempt of making a surface its own parent - always zero-alloc weston_subsurface - apply wl_subsurface.set_position in commit, not immediately - add weston_surface_to_subsurface() - implement sub-surface commit modes parent-cached and independent - implement wl_subcompositor.destroy and wl_subsurface.destroy Changes in v3: - rebased, and use the new transform inheritance code - squashed the commit "add sub-surfaces to repaint list" - fixed a buffer reference leak in commit_from_cache() - Rewrite the sub-surface destructor code, and make it leave the wl_subsurface protocol object inert, if one destroys the corresponding wl_surface. - replaced set_commit_mode with set_sync and set_desync - allowed sub-surface nesting, and fixed repaint accordingly - implemented nested sub-surface commit modes - Made the sub-surface order changes from wl_subsurface.place_above and .place_below to be applied when the parent surface state is applied, instead of immediately. This conforms with the protocol specification now. Signed-off-by: Pekka Paalanen <ppaalanen@gmail.com>
2013-04-25 14:57:42 +04:00
static void
subsurface_set_position(struct wl_client *client,
struct wl_resource *resource, int32_t x, int32_t y)
{
struct weston_subsurface *sub = wl_resource_get_user_data(resource);
compositor: introduce sub-surfaces Implement the basic protocol for sub-surfaces: - expose wl_subcompositor global interface - error checking on protocol calls - associate a parent wl_surface to a sub-surface - introduce the sub-surface role, which is exclusive - an implementation of the wl_subsurface interface - allow nesting of sub-surfaces - proper surface transformation inheritance from parent to sub-surfaces - two different modes of wl_surface.commit for sub-surfaces - hook sub-surfaces up to repaint by modifying the repaint list code Struct weston_subsurface is dynamically allocated. For sub-surfaces, it is completely populated. For parent surfaces, weston_subsurface acts only as a link for stacking order purposes. The wl_resource is unused, parent_destroy_listener is not registered, the transform is not linked, etc. Sub-surfaces are not added directly into layers for display or input. Instead, they are hooked up via the sub-surface list present in parent weston_surface. This way sub-surfaces are inherently linked to the parent surface, and cannot be displayed unless the parent is mapped, too. This also eases restacking, as only the parent will be in a layer list. Also, only the main surface should be subject to shell actions. The surface list rebuilding in weston_output_repaint() is modified to process sub-surface lists, if they are non-empty. The sub-surface list always contains the parent, too, unless empty. The collection of frame_callback_list is moved to a later loop, to streamline the surface list rebuild functions. Features still lacking are: - full-surface alpha support for compound windows Changes in v2: - fix a bug in surface mapping: commit a sub-surface would cause the main surface to never be mapped. - remove debug printfs - detect attempt of making a surface its own parent - always zero-alloc weston_subsurface - apply wl_subsurface.set_position in commit, not immediately - add weston_surface_to_subsurface() - implement sub-surface commit modes parent-cached and independent - implement wl_subcompositor.destroy and wl_subsurface.destroy Changes in v3: - rebased, and use the new transform inheritance code - squashed the commit "add sub-surfaces to repaint list" - fixed a buffer reference leak in commit_from_cache() - Rewrite the sub-surface destructor code, and make it leave the wl_subsurface protocol object inert, if one destroys the corresponding wl_surface. - replaced set_commit_mode with set_sync and set_desync - allowed sub-surface nesting, and fixed repaint accordingly - implemented nested sub-surface commit modes - Made the sub-surface order changes from wl_subsurface.place_above and .place_below to be applied when the parent surface state is applied, instead of immediately. This conforms with the protocol specification now. Signed-off-by: Pekka Paalanen <ppaalanen@gmail.com>
2013-04-25 14:57:42 +04:00
if (!sub)
return;
sub->position.x = x;
sub->position.y = y;
sub->position.set = 1;
}
static struct weston_subsurface *
subsurface_find_sibling(struct weston_subsurface *sub,
struct weston_surface *surface)
compositor: introduce sub-surfaces Implement the basic protocol for sub-surfaces: - expose wl_subcompositor global interface - error checking on protocol calls - associate a parent wl_surface to a sub-surface - introduce the sub-surface role, which is exclusive - an implementation of the wl_subsurface interface - allow nesting of sub-surfaces - proper surface transformation inheritance from parent to sub-surfaces - two different modes of wl_surface.commit for sub-surfaces - hook sub-surfaces up to repaint by modifying the repaint list code Struct weston_subsurface is dynamically allocated. For sub-surfaces, it is completely populated. For parent surfaces, weston_subsurface acts only as a link for stacking order purposes. The wl_resource is unused, parent_destroy_listener is not registered, the transform is not linked, etc. Sub-surfaces are not added directly into layers for display or input. Instead, they are hooked up via the sub-surface list present in parent weston_surface. This way sub-surfaces are inherently linked to the parent surface, and cannot be displayed unless the parent is mapped, too. This also eases restacking, as only the parent will be in a layer list. Also, only the main surface should be subject to shell actions. The surface list rebuilding in weston_output_repaint() is modified to process sub-surface lists, if they are non-empty. The sub-surface list always contains the parent, too, unless empty. The collection of frame_callback_list is moved to a later loop, to streamline the surface list rebuild functions. Features still lacking are: - full-surface alpha support for compound windows Changes in v2: - fix a bug in surface mapping: commit a sub-surface would cause the main surface to never be mapped. - remove debug printfs - detect attempt of making a surface its own parent - always zero-alloc weston_subsurface - apply wl_subsurface.set_position in commit, not immediately - add weston_surface_to_subsurface() - implement sub-surface commit modes parent-cached and independent - implement wl_subcompositor.destroy and wl_subsurface.destroy Changes in v3: - rebased, and use the new transform inheritance code - squashed the commit "add sub-surfaces to repaint list" - fixed a buffer reference leak in commit_from_cache() - Rewrite the sub-surface destructor code, and make it leave the wl_subsurface protocol object inert, if one destroys the corresponding wl_surface. - replaced set_commit_mode with set_sync and set_desync - allowed sub-surface nesting, and fixed repaint accordingly - implemented nested sub-surface commit modes - Made the sub-surface order changes from wl_subsurface.place_above and .place_below to be applied when the parent surface state is applied, instead of immediately. This conforms with the protocol specification now. Signed-off-by: Pekka Paalanen <ppaalanen@gmail.com>
2013-04-25 14:57:42 +04:00
{
struct weston_surface *parent = sub->parent;
struct weston_subsurface *sibling;
compositor: introduce sub-surfaces Implement the basic protocol for sub-surfaces: - expose wl_subcompositor global interface - error checking on protocol calls - associate a parent wl_surface to a sub-surface - introduce the sub-surface role, which is exclusive - an implementation of the wl_subsurface interface - allow nesting of sub-surfaces - proper surface transformation inheritance from parent to sub-surfaces - two different modes of wl_surface.commit for sub-surfaces - hook sub-surfaces up to repaint by modifying the repaint list code Struct weston_subsurface is dynamically allocated. For sub-surfaces, it is completely populated. For parent surfaces, weston_subsurface acts only as a link for stacking order purposes. The wl_resource is unused, parent_destroy_listener is not registered, the transform is not linked, etc. Sub-surfaces are not added directly into layers for display or input. Instead, they are hooked up via the sub-surface list present in parent weston_surface. This way sub-surfaces are inherently linked to the parent surface, and cannot be displayed unless the parent is mapped, too. This also eases restacking, as only the parent will be in a layer list. Also, only the main surface should be subject to shell actions. The surface list rebuilding in weston_output_repaint() is modified to process sub-surface lists, if they are non-empty. The sub-surface list always contains the parent, too, unless empty. The collection of frame_callback_list is moved to a later loop, to streamline the surface list rebuild functions. Features still lacking are: - full-surface alpha support for compound windows Changes in v2: - fix a bug in surface mapping: commit a sub-surface would cause the main surface to never be mapped. - remove debug printfs - detect attempt of making a surface its own parent - always zero-alloc weston_subsurface - apply wl_subsurface.set_position in commit, not immediately - add weston_surface_to_subsurface() - implement sub-surface commit modes parent-cached and independent - implement wl_subcompositor.destroy and wl_subsurface.destroy Changes in v3: - rebased, and use the new transform inheritance code - squashed the commit "add sub-surfaces to repaint list" - fixed a buffer reference leak in commit_from_cache() - Rewrite the sub-surface destructor code, and make it leave the wl_subsurface protocol object inert, if one destroys the corresponding wl_surface. - replaced set_commit_mode with set_sync and set_desync - allowed sub-surface nesting, and fixed repaint accordingly - implemented nested sub-surface commit modes - Made the sub-surface order changes from wl_subsurface.place_above and .place_below to be applied when the parent surface state is applied, instead of immediately. This conforms with the protocol specification now. Signed-off-by: Pekka Paalanen <ppaalanen@gmail.com>
2013-04-25 14:57:42 +04:00
wl_list_for_each(sibling, &parent->subsurface_list, parent_link) {
if (sibling->surface == surface && sibling != sub)
return sibling;
}
compositor: introduce sub-surfaces Implement the basic protocol for sub-surfaces: - expose wl_subcompositor global interface - error checking on protocol calls - associate a parent wl_surface to a sub-surface - introduce the sub-surface role, which is exclusive - an implementation of the wl_subsurface interface - allow nesting of sub-surfaces - proper surface transformation inheritance from parent to sub-surfaces - two different modes of wl_surface.commit for sub-surfaces - hook sub-surfaces up to repaint by modifying the repaint list code Struct weston_subsurface is dynamically allocated. For sub-surfaces, it is completely populated. For parent surfaces, weston_subsurface acts only as a link for stacking order purposes. The wl_resource is unused, parent_destroy_listener is not registered, the transform is not linked, etc. Sub-surfaces are not added directly into layers for display or input. Instead, they are hooked up via the sub-surface list present in parent weston_surface. This way sub-surfaces are inherently linked to the parent surface, and cannot be displayed unless the parent is mapped, too. This also eases restacking, as only the parent will be in a layer list. Also, only the main surface should be subject to shell actions. The surface list rebuilding in weston_output_repaint() is modified to process sub-surface lists, if they are non-empty. The sub-surface list always contains the parent, too, unless empty. The collection of frame_callback_list is moved to a later loop, to streamline the surface list rebuild functions. Features still lacking are: - full-surface alpha support for compound windows Changes in v2: - fix a bug in surface mapping: commit a sub-surface would cause the main surface to never be mapped. - remove debug printfs - detect attempt of making a surface its own parent - always zero-alloc weston_subsurface - apply wl_subsurface.set_position in commit, not immediately - add weston_surface_to_subsurface() - implement sub-surface commit modes parent-cached and independent - implement wl_subcompositor.destroy and wl_subsurface.destroy Changes in v3: - rebased, and use the new transform inheritance code - squashed the commit "add sub-surfaces to repaint list" - fixed a buffer reference leak in commit_from_cache() - Rewrite the sub-surface destructor code, and make it leave the wl_subsurface protocol object inert, if one destroys the corresponding wl_surface. - replaced set_commit_mode with set_sync and set_desync - allowed sub-surface nesting, and fixed repaint accordingly - implemented nested sub-surface commit modes - Made the sub-surface order changes from wl_subsurface.place_above and .place_below to be applied when the parent surface state is applied, instead of immediately. This conforms with the protocol specification now. Signed-off-by: Pekka Paalanen <ppaalanen@gmail.com>
2013-04-25 14:57:42 +04:00
return NULL;
}
static struct weston_subsurface *
subsurface_sibling_check(struct weston_subsurface *sub,
struct weston_surface *surface,
const char *request)
{
struct weston_subsurface *sibling;
sibling = subsurface_find_sibling(sub, surface);
compositor: introduce sub-surfaces Implement the basic protocol for sub-surfaces: - expose wl_subcompositor global interface - error checking on protocol calls - associate a parent wl_surface to a sub-surface - introduce the sub-surface role, which is exclusive - an implementation of the wl_subsurface interface - allow nesting of sub-surfaces - proper surface transformation inheritance from parent to sub-surfaces - two different modes of wl_surface.commit for sub-surfaces - hook sub-surfaces up to repaint by modifying the repaint list code Struct weston_subsurface is dynamically allocated. For sub-surfaces, it is completely populated. For parent surfaces, weston_subsurface acts only as a link for stacking order purposes. The wl_resource is unused, parent_destroy_listener is not registered, the transform is not linked, etc. Sub-surfaces are not added directly into layers for display or input. Instead, they are hooked up via the sub-surface list present in parent weston_surface. This way sub-surfaces are inherently linked to the parent surface, and cannot be displayed unless the parent is mapped, too. This also eases restacking, as only the parent will be in a layer list. Also, only the main surface should be subject to shell actions. The surface list rebuilding in weston_output_repaint() is modified to process sub-surface lists, if they are non-empty. The sub-surface list always contains the parent, too, unless empty. The collection of frame_callback_list is moved to a later loop, to streamline the surface list rebuild functions. Features still lacking are: - full-surface alpha support for compound windows Changes in v2: - fix a bug in surface mapping: commit a sub-surface would cause the main surface to never be mapped. - remove debug printfs - detect attempt of making a surface its own parent - always zero-alloc weston_subsurface - apply wl_subsurface.set_position in commit, not immediately - add weston_surface_to_subsurface() - implement sub-surface commit modes parent-cached and independent - implement wl_subcompositor.destroy and wl_subsurface.destroy Changes in v3: - rebased, and use the new transform inheritance code - squashed the commit "add sub-surfaces to repaint list" - fixed a buffer reference leak in commit_from_cache() - Rewrite the sub-surface destructor code, and make it leave the wl_subsurface protocol object inert, if one destroys the corresponding wl_surface. - replaced set_commit_mode with set_sync and set_desync - allowed sub-surface nesting, and fixed repaint accordingly - implemented nested sub-surface commit modes - Made the sub-surface order changes from wl_subsurface.place_above and .place_below to be applied when the parent surface state is applied, instead of immediately. This conforms with the protocol specification now. Signed-off-by: Pekka Paalanen <ppaalanen@gmail.com>
2013-04-25 14:57:42 +04:00
if (!sibling) {
wl_resource_post_error(sub->resource,
WL_SUBSURFACE_ERROR_BAD_SURFACE,
"%s: wl_surface@%d is not a parent or sibling",
request, wl_resource_get_id(surface->resource));
compositor: introduce sub-surfaces Implement the basic protocol for sub-surfaces: - expose wl_subcompositor global interface - error checking on protocol calls - associate a parent wl_surface to a sub-surface - introduce the sub-surface role, which is exclusive - an implementation of the wl_subsurface interface - allow nesting of sub-surfaces - proper surface transformation inheritance from parent to sub-surfaces - two different modes of wl_surface.commit for sub-surfaces - hook sub-surfaces up to repaint by modifying the repaint list code Struct weston_subsurface is dynamically allocated. For sub-surfaces, it is completely populated. For parent surfaces, weston_subsurface acts only as a link for stacking order purposes. The wl_resource is unused, parent_destroy_listener is not registered, the transform is not linked, etc. Sub-surfaces are not added directly into layers for display or input. Instead, they are hooked up via the sub-surface list present in parent weston_surface. This way sub-surfaces are inherently linked to the parent surface, and cannot be displayed unless the parent is mapped, too. This also eases restacking, as only the parent will be in a layer list. Also, only the main surface should be subject to shell actions. The surface list rebuilding in weston_output_repaint() is modified to process sub-surface lists, if they are non-empty. The sub-surface list always contains the parent, too, unless empty. The collection of frame_callback_list is moved to a later loop, to streamline the surface list rebuild functions. Features still lacking are: - full-surface alpha support for compound windows Changes in v2: - fix a bug in surface mapping: commit a sub-surface would cause the main surface to never be mapped. - remove debug printfs - detect attempt of making a surface its own parent - always zero-alloc weston_subsurface - apply wl_subsurface.set_position in commit, not immediately - add weston_surface_to_subsurface() - implement sub-surface commit modes parent-cached and independent - implement wl_subcompositor.destroy and wl_subsurface.destroy Changes in v3: - rebased, and use the new transform inheritance code - squashed the commit "add sub-surfaces to repaint list" - fixed a buffer reference leak in commit_from_cache() - Rewrite the sub-surface destructor code, and make it leave the wl_subsurface protocol object inert, if one destroys the corresponding wl_surface. - replaced set_commit_mode with set_sync and set_desync - allowed sub-surface nesting, and fixed repaint accordingly - implemented nested sub-surface commit modes - Made the sub-surface order changes from wl_subsurface.place_above and .place_below to be applied when the parent surface state is applied, instead of immediately. This conforms with the protocol specification now. Signed-off-by: Pekka Paalanen <ppaalanen@gmail.com>
2013-04-25 14:57:42 +04:00
return NULL;
}
assert(sibling->parent == sub->parent);
compositor: introduce sub-surfaces Implement the basic protocol for sub-surfaces: - expose wl_subcompositor global interface - error checking on protocol calls - associate a parent wl_surface to a sub-surface - introduce the sub-surface role, which is exclusive - an implementation of the wl_subsurface interface - allow nesting of sub-surfaces - proper surface transformation inheritance from parent to sub-surfaces - two different modes of wl_surface.commit for sub-surfaces - hook sub-surfaces up to repaint by modifying the repaint list code Struct weston_subsurface is dynamically allocated. For sub-surfaces, it is completely populated. For parent surfaces, weston_subsurface acts only as a link for stacking order purposes. The wl_resource is unused, parent_destroy_listener is not registered, the transform is not linked, etc. Sub-surfaces are not added directly into layers for display or input. Instead, they are hooked up via the sub-surface list present in parent weston_surface. This way sub-surfaces are inherently linked to the parent surface, and cannot be displayed unless the parent is mapped, too. This also eases restacking, as only the parent will be in a layer list. Also, only the main surface should be subject to shell actions. The surface list rebuilding in weston_output_repaint() is modified to process sub-surface lists, if they are non-empty. The sub-surface list always contains the parent, too, unless empty. The collection of frame_callback_list is moved to a later loop, to streamline the surface list rebuild functions. Features still lacking are: - full-surface alpha support for compound windows Changes in v2: - fix a bug in surface mapping: commit a sub-surface would cause the main surface to never be mapped. - remove debug printfs - detect attempt of making a surface its own parent - always zero-alloc weston_subsurface - apply wl_subsurface.set_position in commit, not immediately - add weston_surface_to_subsurface() - implement sub-surface commit modes parent-cached and independent - implement wl_subcompositor.destroy and wl_subsurface.destroy Changes in v3: - rebased, and use the new transform inheritance code - squashed the commit "add sub-surfaces to repaint list" - fixed a buffer reference leak in commit_from_cache() - Rewrite the sub-surface destructor code, and make it leave the wl_subsurface protocol object inert, if one destroys the corresponding wl_surface. - replaced set_commit_mode with set_sync and set_desync - allowed sub-surface nesting, and fixed repaint accordingly - implemented nested sub-surface commit modes - Made the sub-surface order changes from wl_subsurface.place_above and .place_below to be applied when the parent surface state is applied, instead of immediately. This conforms with the protocol specification now. Signed-off-by: Pekka Paalanen <ppaalanen@gmail.com>
2013-04-25 14:57:42 +04:00
return sibling;
}
static void
subsurface_place_above(struct wl_client *client,
struct wl_resource *resource,
struct wl_resource *sibling_resource)
{
struct weston_subsurface *sub = wl_resource_get_user_data(resource);
struct weston_surface *surface =
wl_resource_get_user_data(sibling_resource);
compositor: introduce sub-surfaces Implement the basic protocol for sub-surfaces: - expose wl_subcompositor global interface - error checking on protocol calls - associate a parent wl_surface to a sub-surface - introduce the sub-surface role, which is exclusive - an implementation of the wl_subsurface interface - allow nesting of sub-surfaces - proper surface transformation inheritance from parent to sub-surfaces - two different modes of wl_surface.commit for sub-surfaces - hook sub-surfaces up to repaint by modifying the repaint list code Struct weston_subsurface is dynamically allocated. For sub-surfaces, it is completely populated. For parent surfaces, weston_subsurface acts only as a link for stacking order purposes. The wl_resource is unused, parent_destroy_listener is not registered, the transform is not linked, etc. Sub-surfaces are not added directly into layers for display or input. Instead, they are hooked up via the sub-surface list present in parent weston_surface. This way sub-surfaces are inherently linked to the parent surface, and cannot be displayed unless the parent is mapped, too. This also eases restacking, as only the parent will be in a layer list. Also, only the main surface should be subject to shell actions. The surface list rebuilding in weston_output_repaint() is modified to process sub-surface lists, if they are non-empty. The sub-surface list always contains the parent, too, unless empty. The collection of frame_callback_list is moved to a later loop, to streamline the surface list rebuild functions. Features still lacking are: - full-surface alpha support for compound windows Changes in v2: - fix a bug in surface mapping: commit a sub-surface would cause the main surface to never be mapped. - remove debug printfs - detect attempt of making a surface its own parent - always zero-alloc weston_subsurface - apply wl_subsurface.set_position in commit, not immediately - add weston_surface_to_subsurface() - implement sub-surface commit modes parent-cached and independent - implement wl_subcompositor.destroy and wl_subsurface.destroy Changes in v3: - rebased, and use the new transform inheritance code - squashed the commit "add sub-surfaces to repaint list" - fixed a buffer reference leak in commit_from_cache() - Rewrite the sub-surface destructor code, and make it leave the wl_subsurface protocol object inert, if one destroys the corresponding wl_surface. - replaced set_commit_mode with set_sync and set_desync - allowed sub-surface nesting, and fixed repaint accordingly - implemented nested sub-surface commit modes - Made the sub-surface order changes from wl_subsurface.place_above and .place_below to be applied when the parent surface state is applied, instead of immediately. This conforms with the protocol specification now. Signed-off-by: Pekka Paalanen <ppaalanen@gmail.com>
2013-04-25 14:57:42 +04:00
struct weston_subsurface *sibling;
if (!sub)
return;
sibling = subsurface_sibling_check(sub, surface, "place_above");
if (!sibling)
return;
wl_list_remove(&sub->parent_link_pending);
wl_list_insert(sibling->parent_link_pending.prev,
&sub->parent_link_pending);
sub->reordered = true;
compositor: introduce sub-surfaces Implement the basic protocol for sub-surfaces: - expose wl_subcompositor global interface - error checking on protocol calls - associate a parent wl_surface to a sub-surface - introduce the sub-surface role, which is exclusive - an implementation of the wl_subsurface interface - allow nesting of sub-surfaces - proper surface transformation inheritance from parent to sub-surfaces - two different modes of wl_surface.commit for sub-surfaces - hook sub-surfaces up to repaint by modifying the repaint list code Struct weston_subsurface is dynamically allocated. For sub-surfaces, it is completely populated. For parent surfaces, weston_subsurface acts only as a link for stacking order purposes. The wl_resource is unused, parent_destroy_listener is not registered, the transform is not linked, etc. Sub-surfaces are not added directly into layers for display or input. Instead, they are hooked up via the sub-surface list present in parent weston_surface. This way sub-surfaces are inherently linked to the parent surface, and cannot be displayed unless the parent is mapped, too. This also eases restacking, as only the parent will be in a layer list. Also, only the main surface should be subject to shell actions. The surface list rebuilding in weston_output_repaint() is modified to process sub-surface lists, if they are non-empty. The sub-surface list always contains the parent, too, unless empty. The collection of frame_callback_list is moved to a later loop, to streamline the surface list rebuild functions. Features still lacking are: - full-surface alpha support for compound windows Changes in v2: - fix a bug in surface mapping: commit a sub-surface would cause the main surface to never be mapped. - remove debug printfs - detect attempt of making a surface its own parent - always zero-alloc weston_subsurface - apply wl_subsurface.set_position in commit, not immediately - add weston_surface_to_subsurface() - implement sub-surface commit modes parent-cached and independent - implement wl_subcompositor.destroy and wl_subsurface.destroy Changes in v3: - rebased, and use the new transform inheritance code - squashed the commit "add sub-surfaces to repaint list" - fixed a buffer reference leak in commit_from_cache() - Rewrite the sub-surface destructor code, and make it leave the wl_subsurface protocol object inert, if one destroys the corresponding wl_surface. - replaced set_commit_mode with set_sync and set_desync - allowed sub-surface nesting, and fixed repaint accordingly - implemented nested sub-surface commit modes - Made the sub-surface order changes from wl_subsurface.place_above and .place_below to be applied when the parent surface state is applied, instead of immediately. This conforms with the protocol specification now. Signed-off-by: Pekka Paalanen <ppaalanen@gmail.com>
2013-04-25 14:57:42 +04:00
}
static void
subsurface_place_below(struct wl_client *client,
struct wl_resource *resource,
struct wl_resource *sibling_resource)
{
struct weston_subsurface *sub = wl_resource_get_user_data(resource);
struct weston_surface *surface =
wl_resource_get_user_data(sibling_resource);
compositor: introduce sub-surfaces Implement the basic protocol for sub-surfaces: - expose wl_subcompositor global interface - error checking on protocol calls - associate a parent wl_surface to a sub-surface - introduce the sub-surface role, which is exclusive - an implementation of the wl_subsurface interface - allow nesting of sub-surfaces - proper surface transformation inheritance from parent to sub-surfaces - two different modes of wl_surface.commit for sub-surfaces - hook sub-surfaces up to repaint by modifying the repaint list code Struct weston_subsurface is dynamically allocated. For sub-surfaces, it is completely populated. For parent surfaces, weston_subsurface acts only as a link for stacking order purposes. The wl_resource is unused, parent_destroy_listener is not registered, the transform is not linked, etc. Sub-surfaces are not added directly into layers for display or input. Instead, they are hooked up via the sub-surface list present in parent weston_surface. This way sub-surfaces are inherently linked to the parent surface, and cannot be displayed unless the parent is mapped, too. This also eases restacking, as only the parent will be in a layer list. Also, only the main surface should be subject to shell actions. The surface list rebuilding in weston_output_repaint() is modified to process sub-surface lists, if they are non-empty. The sub-surface list always contains the parent, too, unless empty. The collection of frame_callback_list is moved to a later loop, to streamline the surface list rebuild functions. Features still lacking are: - full-surface alpha support for compound windows Changes in v2: - fix a bug in surface mapping: commit a sub-surface would cause the main surface to never be mapped. - remove debug printfs - detect attempt of making a surface its own parent - always zero-alloc weston_subsurface - apply wl_subsurface.set_position in commit, not immediately - add weston_surface_to_subsurface() - implement sub-surface commit modes parent-cached and independent - implement wl_subcompositor.destroy and wl_subsurface.destroy Changes in v3: - rebased, and use the new transform inheritance code - squashed the commit "add sub-surfaces to repaint list" - fixed a buffer reference leak in commit_from_cache() - Rewrite the sub-surface destructor code, and make it leave the wl_subsurface protocol object inert, if one destroys the corresponding wl_surface. - replaced set_commit_mode with set_sync and set_desync - allowed sub-surface nesting, and fixed repaint accordingly - implemented nested sub-surface commit modes - Made the sub-surface order changes from wl_subsurface.place_above and .place_below to be applied when the parent surface state is applied, instead of immediately. This conforms with the protocol specification now. Signed-off-by: Pekka Paalanen <ppaalanen@gmail.com>
2013-04-25 14:57:42 +04:00
struct weston_subsurface *sibling;
if (!sub)
return;
sibling = subsurface_sibling_check(sub, surface, "place_below");
if (!sibling)
return;
wl_list_remove(&sub->parent_link_pending);
wl_list_insert(&sibling->parent_link_pending,
&sub->parent_link_pending);
sub->reordered = true;
compositor: introduce sub-surfaces Implement the basic protocol for sub-surfaces: - expose wl_subcompositor global interface - error checking on protocol calls - associate a parent wl_surface to a sub-surface - introduce the sub-surface role, which is exclusive - an implementation of the wl_subsurface interface - allow nesting of sub-surfaces - proper surface transformation inheritance from parent to sub-surfaces - two different modes of wl_surface.commit for sub-surfaces - hook sub-surfaces up to repaint by modifying the repaint list code Struct weston_subsurface is dynamically allocated. For sub-surfaces, it is completely populated. For parent surfaces, weston_subsurface acts only as a link for stacking order purposes. The wl_resource is unused, parent_destroy_listener is not registered, the transform is not linked, etc. Sub-surfaces are not added directly into layers for display or input. Instead, they are hooked up via the sub-surface list present in parent weston_surface. This way sub-surfaces are inherently linked to the parent surface, and cannot be displayed unless the parent is mapped, too. This also eases restacking, as only the parent will be in a layer list. Also, only the main surface should be subject to shell actions. The surface list rebuilding in weston_output_repaint() is modified to process sub-surface lists, if they are non-empty. The sub-surface list always contains the parent, too, unless empty. The collection of frame_callback_list is moved to a later loop, to streamline the surface list rebuild functions. Features still lacking are: - full-surface alpha support for compound windows Changes in v2: - fix a bug in surface mapping: commit a sub-surface would cause the main surface to never be mapped. - remove debug printfs - detect attempt of making a surface its own parent - always zero-alloc weston_subsurface - apply wl_subsurface.set_position in commit, not immediately - add weston_surface_to_subsurface() - implement sub-surface commit modes parent-cached and independent - implement wl_subcompositor.destroy and wl_subsurface.destroy Changes in v3: - rebased, and use the new transform inheritance code - squashed the commit "add sub-surfaces to repaint list" - fixed a buffer reference leak in commit_from_cache() - Rewrite the sub-surface destructor code, and make it leave the wl_subsurface protocol object inert, if one destroys the corresponding wl_surface. - replaced set_commit_mode with set_sync and set_desync - allowed sub-surface nesting, and fixed repaint accordingly - implemented nested sub-surface commit modes - Made the sub-surface order changes from wl_subsurface.place_above and .place_below to be applied when the parent surface state is applied, instead of immediately. This conforms with the protocol specification now. Signed-off-by: Pekka Paalanen <ppaalanen@gmail.com>
2013-04-25 14:57:42 +04:00
}
static void
subsurface_set_sync(struct wl_client *client, struct wl_resource *resource)
{
struct weston_subsurface *sub = wl_resource_get_user_data(resource);
compositor: introduce sub-surfaces Implement the basic protocol for sub-surfaces: - expose wl_subcompositor global interface - error checking on protocol calls - associate a parent wl_surface to a sub-surface - introduce the sub-surface role, which is exclusive - an implementation of the wl_subsurface interface - allow nesting of sub-surfaces - proper surface transformation inheritance from parent to sub-surfaces - two different modes of wl_surface.commit for sub-surfaces - hook sub-surfaces up to repaint by modifying the repaint list code Struct weston_subsurface is dynamically allocated. For sub-surfaces, it is completely populated. For parent surfaces, weston_subsurface acts only as a link for stacking order purposes. The wl_resource is unused, parent_destroy_listener is not registered, the transform is not linked, etc. Sub-surfaces are not added directly into layers for display or input. Instead, they are hooked up via the sub-surface list present in parent weston_surface. This way sub-surfaces are inherently linked to the parent surface, and cannot be displayed unless the parent is mapped, too. This also eases restacking, as only the parent will be in a layer list. Also, only the main surface should be subject to shell actions. The surface list rebuilding in weston_output_repaint() is modified to process sub-surface lists, if they are non-empty. The sub-surface list always contains the parent, too, unless empty. The collection of frame_callback_list is moved to a later loop, to streamline the surface list rebuild functions. Features still lacking are: - full-surface alpha support for compound windows Changes in v2: - fix a bug in surface mapping: commit a sub-surface would cause the main surface to never be mapped. - remove debug printfs - detect attempt of making a surface its own parent - always zero-alloc weston_subsurface - apply wl_subsurface.set_position in commit, not immediately - add weston_surface_to_subsurface() - implement sub-surface commit modes parent-cached and independent - implement wl_subcompositor.destroy and wl_subsurface.destroy Changes in v3: - rebased, and use the new transform inheritance code - squashed the commit "add sub-surfaces to repaint list" - fixed a buffer reference leak in commit_from_cache() - Rewrite the sub-surface destructor code, and make it leave the wl_subsurface protocol object inert, if one destroys the corresponding wl_surface. - replaced set_commit_mode with set_sync and set_desync - allowed sub-surface nesting, and fixed repaint accordingly - implemented nested sub-surface commit modes - Made the sub-surface order changes from wl_subsurface.place_above and .place_below to be applied when the parent surface state is applied, instead of immediately. This conforms with the protocol specification now. Signed-off-by: Pekka Paalanen <ppaalanen@gmail.com>
2013-04-25 14:57:42 +04:00
if (sub)
sub->synchronized = 1;
}
static void
subsurface_set_desync(struct wl_client *client, struct wl_resource *resource)
{
struct weston_subsurface *sub = wl_resource_get_user_data(resource);
compositor: introduce sub-surfaces Implement the basic protocol for sub-surfaces: - expose wl_subcompositor global interface - error checking on protocol calls - associate a parent wl_surface to a sub-surface - introduce the sub-surface role, which is exclusive - an implementation of the wl_subsurface interface - allow nesting of sub-surfaces - proper surface transformation inheritance from parent to sub-surfaces - two different modes of wl_surface.commit for sub-surfaces - hook sub-surfaces up to repaint by modifying the repaint list code Struct weston_subsurface is dynamically allocated. For sub-surfaces, it is completely populated. For parent surfaces, weston_subsurface acts only as a link for stacking order purposes. The wl_resource is unused, parent_destroy_listener is not registered, the transform is not linked, etc. Sub-surfaces are not added directly into layers for display or input. Instead, they are hooked up via the sub-surface list present in parent weston_surface. This way sub-surfaces are inherently linked to the parent surface, and cannot be displayed unless the parent is mapped, too. This also eases restacking, as only the parent will be in a layer list. Also, only the main surface should be subject to shell actions. The surface list rebuilding in weston_output_repaint() is modified to process sub-surface lists, if they are non-empty. The sub-surface list always contains the parent, too, unless empty. The collection of frame_callback_list is moved to a later loop, to streamline the surface list rebuild functions. Features still lacking are: - full-surface alpha support for compound windows Changes in v2: - fix a bug in surface mapping: commit a sub-surface would cause the main surface to never be mapped. - remove debug printfs - detect attempt of making a surface its own parent - always zero-alloc weston_subsurface - apply wl_subsurface.set_position in commit, not immediately - add weston_surface_to_subsurface() - implement sub-surface commit modes parent-cached and independent - implement wl_subcompositor.destroy and wl_subsurface.destroy Changes in v3: - rebased, and use the new transform inheritance code - squashed the commit "add sub-surfaces to repaint list" - fixed a buffer reference leak in commit_from_cache() - Rewrite the sub-surface destructor code, and make it leave the wl_subsurface protocol object inert, if one destroys the corresponding wl_surface. - replaced set_commit_mode with set_sync and set_desync - allowed sub-surface nesting, and fixed repaint accordingly - implemented nested sub-surface commit modes - Made the sub-surface order changes from wl_subsurface.place_above and .place_below to be applied when the parent surface state is applied, instead of immediately. This conforms with the protocol specification now. Signed-off-by: Pekka Paalanen <ppaalanen@gmail.com>
2013-04-25 14:57:42 +04:00
if (sub && sub->synchronized) {
compositor: introduce sub-surfaces Implement the basic protocol for sub-surfaces: - expose wl_subcompositor global interface - error checking on protocol calls - associate a parent wl_surface to a sub-surface - introduce the sub-surface role, which is exclusive - an implementation of the wl_subsurface interface - allow nesting of sub-surfaces - proper surface transformation inheritance from parent to sub-surfaces - two different modes of wl_surface.commit for sub-surfaces - hook sub-surfaces up to repaint by modifying the repaint list code Struct weston_subsurface is dynamically allocated. For sub-surfaces, it is completely populated. For parent surfaces, weston_subsurface acts only as a link for stacking order purposes. The wl_resource is unused, parent_destroy_listener is not registered, the transform is not linked, etc. Sub-surfaces are not added directly into layers for display or input. Instead, they are hooked up via the sub-surface list present in parent weston_surface. This way sub-surfaces are inherently linked to the parent surface, and cannot be displayed unless the parent is mapped, too. This also eases restacking, as only the parent will be in a layer list. Also, only the main surface should be subject to shell actions. The surface list rebuilding in weston_output_repaint() is modified to process sub-surface lists, if they are non-empty. The sub-surface list always contains the parent, too, unless empty. The collection of frame_callback_list is moved to a later loop, to streamline the surface list rebuild functions. Features still lacking are: - full-surface alpha support for compound windows Changes in v2: - fix a bug in surface mapping: commit a sub-surface would cause the main surface to never be mapped. - remove debug printfs - detect attempt of making a surface its own parent - always zero-alloc weston_subsurface - apply wl_subsurface.set_position in commit, not immediately - add weston_surface_to_subsurface() - implement sub-surface commit modes parent-cached and independent - implement wl_subcompositor.destroy and wl_subsurface.destroy Changes in v3: - rebased, and use the new transform inheritance code - squashed the commit "add sub-surfaces to repaint list" - fixed a buffer reference leak in commit_from_cache() - Rewrite the sub-surface destructor code, and make it leave the wl_subsurface protocol object inert, if one destroys the corresponding wl_surface. - replaced set_commit_mode with set_sync and set_desync - allowed sub-surface nesting, and fixed repaint accordingly - implemented nested sub-surface commit modes - Made the sub-surface order changes from wl_subsurface.place_above and .place_below to be applied when the parent surface state is applied, instead of immediately. This conforms with the protocol specification now. Signed-off-by: Pekka Paalanen <ppaalanen@gmail.com>
2013-04-25 14:57:42 +04:00
sub->synchronized = 0;
/* If sub became effectively desynchronized, flush. */
if (!weston_subsurface_is_synchronized(sub))
weston_subsurface_synchronized_commit(sub);
}
compositor: introduce sub-surfaces Implement the basic protocol for sub-surfaces: - expose wl_subcompositor global interface - error checking on protocol calls - associate a parent wl_surface to a sub-surface - introduce the sub-surface role, which is exclusive - an implementation of the wl_subsurface interface - allow nesting of sub-surfaces - proper surface transformation inheritance from parent to sub-surfaces - two different modes of wl_surface.commit for sub-surfaces - hook sub-surfaces up to repaint by modifying the repaint list code Struct weston_subsurface is dynamically allocated. For sub-surfaces, it is completely populated. For parent surfaces, weston_subsurface acts only as a link for stacking order purposes. The wl_resource is unused, parent_destroy_listener is not registered, the transform is not linked, etc. Sub-surfaces are not added directly into layers for display or input. Instead, they are hooked up via the sub-surface list present in parent weston_surface. This way sub-surfaces are inherently linked to the parent surface, and cannot be displayed unless the parent is mapped, too. This also eases restacking, as only the parent will be in a layer list. Also, only the main surface should be subject to shell actions. The surface list rebuilding in weston_output_repaint() is modified to process sub-surface lists, if they are non-empty. The sub-surface list always contains the parent, too, unless empty. The collection of frame_callback_list is moved to a later loop, to streamline the surface list rebuild functions. Features still lacking are: - full-surface alpha support for compound windows Changes in v2: - fix a bug in surface mapping: commit a sub-surface would cause the main surface to never be mapped. - remove debug printfs - detect attempt of making a surface its own parent - always zero-alloc weston_subsurface - apply wl_subsurface.set_position in commit, not immediately - add weston_surface_to_subsurface() - implement sub-surface commit modes parent-cached and independent - implement wl_subcompositor.destroy and wl_subsurface.destroy Changes in v3: - rebased, and use the new transform inheritance code - squashed the commit "add sub-surfaces to repaint list" - fixed a buffer reference leak in commit_from_cache() - Rewrite the sub-surface destructor code, and make it leave the wl_subsurface protocol object inert, if one destroys the corresponding wl_surface. - replaced set_commit_mode with set_sync and set_desync - allowed sub-surface nesting, and fixed repaint accordingly - implemented nested sub-surface commit modes - Made the sub-surface order changes from wl_subsurface.place_above and .place_below to be applied when the parent surface state is applied, instead of immediately. This conforms with the protocol specification now. Signed-off-by: Pekka Paalanen <ppaalanen@gmail.com>
2013-04-25 14:57:42 +04:00
}
static void
weston_subsurface_unlink_parent(struct weston_subsurface *sub)
{
wl_list_remove(&sub->parent_link);
wl_list_remove(&sub->parent_link_pending);
wl_list_remove(&sub->parent_destroy_listener.link);
sub->parent = NULL;
}
static void
weston_subsurface_destroy(struct weston_subsurface *sub);
static void
subsurface_handle_surface_destroy(struct wl_listener *listener, void *data)
{
struct weston_subsurface *sub =
container_of(listener, struct weston_subsurface,
surface_destroy_listener);
assert(data == sub->surface);
compositor: introduce sub-surfaces Implement the basic protocol for sub-surfaces: - expose wl_subcompositor global interface - error checking on protocol calls - associate a parent wl_surface to a sub-surface - introduce the sub-surface role, which is exclusive - an implementation of the wl_subsurface interface - allow nesting of sub-surfaces - proper surface transformation inheritance from parent to sub-surfaces - two different modes of wl_surface.commit for sub-surfaces - hook sub-surfaces up to repaint by modifying the repaint list code Struct weston_subsurface is dynamically allocated. For sub-surfaces, it is completely populated. For parent surfaces, weston_subsurface acts only as a link for stacking order purposes. The wl_resource is unused, parent_destroy_listener is not registered, the transform is not linked, etc. Sub-surfaces are not added directly into layers for display or input. Instead, they are hooked up via the sub-surface list present in parent weston_surface. This way sub-surfaces are inherently linked to the parent surface, and cannot be displayed unless the parent is mapped, too. This also eases restacking, as only the parent will be in a layer list. Also, only the main surface should be subject to shell actions. The surface list rebuilding in weston_output_repaint() is modified to process sub-surface lists, if they are non-empty. The sub-surface list always contains the parent, too, unless empty. The collection of frame_callback_list is moved to a later loop, to streamline the surface list rebuild functions. Features still lacking are: - full-surface alpha support for compound windows Changes in v2: - fix a bug in surface mapping: commit a sub-surface would cause the main surface to never be mapped. - remove debug printfs - detect attempt of making a surface its own parent - always zero-alloc weston_subsurface - apply wl_subsurface.set_position in commit, not immediately - add weston_surface_to_subsurface() - implement sub-surface commit modes parent-cached and independent - implement wl_subcompositor.destroy and wl_subsurface.destroy Changes in v3: - rebased, and use the new transform inheritance code - squashed the commit "add sub-surfaces to repaint list" - fixed a buffer reference leak in commit_from_cache() - Rewrite the sub-surface destructor code, and make it leave the wl_subsurface protocol object inert, if one destroys the corresponding wl_surface. - replaced set_commit_mode with set_sync and set_desync - allowed sub-surface nesting, and fixed repaint accordingly - implemented nested sub-surface commit modes - Made the sub-surface order changes from wl_subsurface.place_above and .place_below to be applied when the parent surface state is applied, instead of immediately. This conforms with the protocol specification now. Signed-off-by: Pekka Paalanen <ppaalanen@gmail.com>
2013-04-25 14:57:42 +04:00
/* The protocol object (wl_resource) is left inert. */
if (sub->resource)
wl_resource_set_user_data(sub->resource, NULL);
compositor: introduce sub-surfaces Implement the basic protocol for sub-surfaces: - expose wl_subcompositor global interface - error checking on protocol calls - associate a parent wl_surface to a sub-surface - introduce the sub-surface role, which is exclusive - an implementation of the wl_subsurface interface - allow nesting of sub-surfaces - proper surface transformation inheritance from parent to sub-surfaces - two different modes of wl_surface.commit for sub-surfaces - hook sub-surfaces up to repaint by modifying the repaint list code Struct weston_subsurface is dynamically allocated. For sub-surfaces, it is completely populated. For parent surfaces, weston_subsurface acts only as a link for stacking order purposes. The wl_resource is unused, parent_destroy_listener is not registered, the transform is not linked, etc. Sub-surfaces are not added directly into layers for display or input. Instead, they are hooked up via the sub-surface list present in parent weston_surface. This way sub-surfaces are inherently linked to the parent surface, and cannot be displayed unless the parent is mapped, too. This also eases restacking, as only the parent will be in a layer list. Also, only the main surface should be subject to shell actions. The surface list rebuilding in weston_output_repaint() is modified to process sub-surface lists, if they are non-empty. The sub-surface list always contains the parent, too, unless empty. The collection of frame_callback_list is moved to a later loop, to streamline the surface list rebuild functions. Features still lacking are: - full-surface alpha support for compound windows Changes in v2: - fix a bug in surface mapping: commit a sub-surface would cause the main surface to never be mapped. - remove debug printfs - detect attempt of making a surface its own parent - always zero-alloc weston_subsurface - apply wl_subsurface.set_position in commit, not immediately - add weston_surface_to_subsurface() - implement sub-surface commit modes parent-cached and independent - implement wl_subcompositor.destroy and wl_subsurface.destroy Changes in v3: - rebased, and use the new transform inheritance code - squashed the commit "add sub-surfaces to repaint list" - fixed a buffer reference leak in commit_from_cache() - Rewrite the sub-surface destructor code, and make it leave the wl_subsurface protocol object inert, if one destroys the corresponding wl_surface. - replaced set_commit_mode with set_sync and set_desync - allowed sub-surface nesting, and fixed repaint accordingly - implemented nested sub-surface commit modes - Made the sub-surface order changes from wl_subsurface.place_above and .place_below to be applied when the parent surface state is applied, instead of immediately. This conforms with the protocol specification now. Signed-off-by: Pekka Paalanen <ppaalanen@gmail.com>
2013-04-25 14:57:42 +04:00
weston_subsurface_destroy(sub);
}
static void
subsurface_handle_parent_destroy(struct wl_listener *listener, void *data)
{
struct weston_subsurface *sub =
container_of(listener, struct weston_subsurface,
parent_destroy_listener);
assert(data == sub->parent);
compositor: introduce sub-surfaces Implement the basic protocol for sub-surfaces: - expose wl_subcompositor global interface - error checking on protocol calls - associate a parent wl_surface to a sub-surface - introduce the sub-surface role, which is exclusive - an implementation of the wl_subsurface interface - allow nesting of sub-surfaces - proper surface transformation inheritance from parent to sub-surfaces - two different modes of wl_surface.commit for sub-surfaces - hook sub-surfaces up to repaint by modifying the repaint list code Struct weston_subsurface is dynamically allocated. For sub-surfaces, it is completely populated. For parent surfaces, weston_subsurface acts only as a link for stacking order purposes. The wl_resource is unused, parent_destroy_listener is not registered, the transform is not linked, etc. Sub-surfaces are not added directly into layers for display or input. Instead, they are hooked up via the sub-surface list present in parent weston_surface. This way sub-surfaces are inherently linked to the parent surface, and cannot be displayed unless the parent is mapped, too. This also eases restacking, as only the parent will be in a layer list. Also, only the main surface should be subject to shell actions. The surface list rebuilding in weston_output_repaint() is modified to process sub-surface lists, if they are non-empty. The sub-surface list always contains the parent, too, unless empty. The collection of frame_callback_list is moved to a later loop, to streamline the surface list rebuild functions. Features still lacking are: - full-surface alpha support for compound windows Changes in v2: - fix a bug in surface mapping: commit a sub-surface would cause the main surface to never be mapped. - remove debug printfs - detect attempt of making a surface its own parent - always zero-alloc weston_subsurface - apply wl_subsurface.set_position in commit, not immediately - add weston_surface_to_subsurface() - implement sub-surface commit modes parent-cached and independent - implement wl_subcompositor.destroy and wl_subsurface.destroy Changes in v3: - rebased, and use the new transform inheritance code - squashed the commit "add sub-surfaces to repaint list" - fixed a buffer reference leak in commit_from_cache() - Rewrite the sub-surface destructor code, and make it leave the wl_subsurface protocol object inert, if one destroys the corresponding wl_surface. - replaced set_commit_mode with set_sync and set_desync - allowed sub-surface nesting, and fixed repaint accordingly - implemented nested sub-surface commit modes - Made the sub-surface order changes from wl_subsurface.place_above and .place_below to be applied when the parent surface state is applied, instead of immediately. This conforms with the protocol specification now. Signed-off-by: Pekka Paalanen <ppaalanen@gmail.com>
2013-04-25 14:57:42 +04:00
assert(sub->surface != sub->parent);
if (weston_surface_is_mapped(sub->surface))
weston_surface_unmap(sub->surface);
weston_subsurface_unlink_parent(sub);
}
static void
subsurface_resource_destroy(struct wl_resource *resource)
{
struct weston_subsurface *sub = wl_resource_get_user_data(resource);
compositor: introduce sub-surfaces Implement the basic protocol for sub-surfaces: - expose wl_subcompositor global interface - error checking on protocol calls - associate a parent wl_surface to a sub-surface - introduce the sub-surface role, which is exclusive - an implementation of the wl_subsurface interface - allow nesting of sub-surfaces - proper surface transformation inheritance from parent to sub-surfaces - two different modes of wl_surface.commit for sub-surfaces - hook sub-surfaces up to repaint by modifying the repaint list code Struct weston_subsurface is dynamically allocated. For sub-surfaces, it is completely populated. For parent surfaces, weston_subsurface acts only as a link for stacking order purposes. The wl_resource is unused, parent_destroy_listener is not registered, the transform is not linked, etc. Sub-surfaces are not added directly into layers for display or input. Instead, they are hooked up via the sub-surface list present in parent weston_surface. This way sub-surfaces are inherently linked to the parent surface, and cannot be displayed unless the parent is mapped, too. This also eases restacking, as only the parent will be in a layer list. Also, only the main surface should be subject to shell actions. The surface list rebuilding in weston_output_repaint() is modified to process sub-surface lists, if they are non-empty. The sub-surface list always contains the parent, too, unless empty. The collection of frame_callback_list is moved to a later loop, to streamline the surface list rebuild functions. Features still lacking are: - full-surface alpha support for compound windows Changes in v2: - fix a bug in surface mapping: commit a sub-surface would cause the main surface to never be mapped. - remove debug printfs - detect attempt of making a surface its own parent - always zero-alloc weston_subsurface - apply wl_subsurface.set_position in commit, not immediately - add weston_surface_to_subsurface() - implement sub-surface commit modes parent-cached and independent - implement wl_subcompositor.destroy and wl_subsurface.destroy Changes in v3: - rebased, and use the new transform inheritance code - squashed the commit "add sub-surfaces to repaint list" - fixed a buffer reference leak in commit_from_cache() - Rewrite the sub-surface destructor code, and make it leave the wl_subsurface protocol object inert, if one destroys the corresponding wl_surface. - replaced set_commit_mode with set_sync and set_desync - allowed sub-surface nesting, and fixed repaint accordingly - implemented nested sub-surface commit modes - Made the sub-surface order changes from wl_subsurface.place_above and .place_below to be applied when the parent surface state is applied, instead of immediately. This conforms with the protocol specification now. Signed-off-by: Pekka Paalanen <ppaalanen@gmail.com>
2013-04-25 14:57:42 +04:00
if (sub)
weston_subsurface_destroy(sub);
}
static void
subsurface_destroy(struct wl_client *client, struct wl_resource *resource)
{
wl_resource_destroy(resource);
}
static void
weston_subsurface_link_parent(struct weston_subsurface *sub,
struct weston_surface *parent)
{
sub->parent = parent;
sub->parent_destroy_listener.notify = subsurface_handle_parent_destroy;
wl_signal_add(&parent->destroy_signal,
compositor: introduce sub-surfaces Implement the basic protocol for sub-surfaces: - expose wl_subcompositor global interface - error checking on protocol calls - associate a parent wl_surface to a sub-surface - introduce the sub-surface role, which is exclusive - an implementation of the wl_subsurface interface - allow nesting of sub-surfaces - proper surface transformation inheritance from parent to sub-surfaces - two different modes of wl_surface.commit for sub-surfaces - hook sub-surfaces up to repaint by modifying the repaint list code Struct weston_subsurface is dynamically allocated. For sub-surfaces, it is completely populated. For parent surfaces, weston_subsurface acts only as a link for stacking order purposes. The wl_resource is unused, parent_destroy_listener is not registered, the transform is not linked, etc. Sub-surfaces are not added directly into layers for display or input. Instead, they are hooked up via the sub-surface list present in parent weston_surface. This way sub-surfaces are inherently linked to the parent surface, and cannot be displayed unless the parent is mapped, too. This also eases restacking, as only the parent will be in a layer list. Also, only the main surface should be subject to shell actions. The surface list rebuilding in weston_output_repaint() is modified to process sub-surface lists, if they are non-empty. The sub-surface list always contains the parent, too, unless empty. The collection of frame_callback_list is moved to a later loop, to streamline the surface list rebuild functions. Features still lacking are: - full-surface alpha support for compound windows Changes in v2: - fix a bug in surface mapping: commit a sub-surface would cause the main surface to never be mapped. - remove debug printfs - detect attempt of making a surface its own parent - always zero-alloc weston_subsurface - apply wl_subsurface.set_position in commit, not immediately - add weston_surface_to_subsurface() - implement sub-surface commit modes parent-cached and independent - implement wl_subcompositor.destroy and wl_subsurface.destroy Changes in v3: - rebased, and use the new transform inheritance code - squashed the commit "add sub-surfaces to repaint list" - fixed a buffer reference leak in commit_from_cache() - Rewrite the sub-surface destructor code, and make it leave the wl_subsurface protocol object inert, if one destroys the corresponding wl_surface. - replaced set_commit_mode with set_sync and set_desync - allowed sub-surface nesting, and fixed repaint accordingly - implemented nested sub-surface commit modes - Made the sub-surface order changes from wl_subsurface.place_above and .place_below to be applied when the parent surface state is applied, instead of immediately. This conforms with the protocol specification now. Signed-off-by: Pekka Paalanen <ppaalanen@gmail.com>
2013-04-25 14:57:42 +04:00
&sub->parent_destroy_listener);
wl_list_insert(&parent->subsurface_list, &sub->parent_link);
wl_list_insert(&parent->subsurface_list_pending,
&sub->parent_link_pending);
}
static void
weston_subsurface_link_surface(struct weston_subsurface *sub,
struct weston_surface *surface)
{
sub->surface = surface;
sub->surface_destroy_listener.notify =
subsurface_handle_surface_destroy;
wl_signal_add(&surface->destroy_signal,
compositor: introduce sub-surfaces Implement the basic protocol for sub-surfaces: - expose wl_subcompositor global interface - error checking on protocol calls - associate a parent wl_surface to a sub-surface - introduce the sub-surface role, which is exclusive - an implementation of the wl_subsurface interface - allow nesting of sub-surfaces - proper surface transformation inheritance from parent to sub-surfaces - two different modes of wl_surface.commit for sub-surfaces - hook sub-surfaces up to repaint by modifying the repaint list code Struct weston_subsurface is dynamically allocated. For sub-surfaces, it is completely populated. For parent surfaces, weston_subsurface acts only as a link for stacking order purposes. The wl_resource is unused, parent_destroy_listener is not registered, the transform is not linked, etc. Sub-surfaces are not added directly into layers for display or input. Instead, they are hooked up via the sub-surface list present in parent weston_surface. This way sub-surfaces are inherently linked to the parent surface, and cannot be displayed unless the parent is mapped, too. This also eases restacking, as only the parent will be in a layer list. Also, only the main surface should be subject to shell actions. The surface list rebuilding in weston_output_repaint() is modified to process sub-surface lists, if they are non-empty. The sub-surface list always contains the parent, too, unless empty. The collection of frame_callback_list is moved to a later loop, to streamline the surface list rebuild functions. Features still lacking are: - full-surface alpha support for compound windows Changes in v2: - fix a bug in surface mapping: commit a sub-surface would cause the main surface to never be mapped. - remove debug printfs - detect attempt of making a surface its own parent - always zero-alloc weston_subsurface - apply wl_subsurface.set_position in commit, not immediately - add weston_surface_to_subsurface() - implement sub-surface commit modes parent-cached and independent - implement wl_subcompositor.destroy and wl_subsurface.destroy Changes in v3: - rebased, and use the new transform inheritance code - squashed the commit "add sub-surfaces to repaint list" - fixed a buffer reference leak in commit_from_cache() - Rewrite the sub-surface destructor code, and make it leave the wl_subsurface protocol object inert, if one destroys the corresponding wl_surface. - replaced set_commit_mode with set_sync and set_desync - allowed sub-surface nesting, and fixed repaint accordingly - implemented nested sub-surface commit modes - Made the sub-surface order changes from wl_subsurface.place_above and .place_below to be applied when the parent surface state is applied, instead of immediately. This conforms with the protocol specification now. Signed-off-by: Pekka Paalanen <ppaalanen@gmail.com>
2013-04-25 14:57:42 +04:00
&sub->surface_destroy_listener);
}
static void
weston_subsurface_destroy(struct weston_subsurface *sub)
{
Split the geometry information from weston_surface out into weston_view The weston_surface structure is split into two structures: * The weston_surface structure storres everything required for a client-side or server-side surface. This includes buffers; callbacks; backend private data; input, damage, and opaque regions; and a few other bookkeeping bits. * The weston_view structure represents an entity in the scenegraph and storres all of the geometry information. This includes clip region, alpha, position, and the transformation list as well as all of the temporary information derived from the geometry state. Because a view, and not a surface, is a scenegraph element, the view is what is placed in layers and planes. There are a few things worth noting about the surface/view split: 1. This is *not* a modification to the protocol. It is, instead, a modification to Weston's internal scenegraph to allow a single surface to exist in multiple places at a time. Clients are completely unaware of how many views to a particular surface exist. 2. A view is considered a direct child of a surface and is destroyed when the surface is destroyed. Because of this, the view.surface pointer is always valid and non-null. 3. The compositor's surface_list is replaced with a view_list. Due to subsurfaces, building the view list is a little more complicated than it used to be and involves building a tree of views on the fly whenever subsurfaces are used. However, this means that backends can remain completely subsurface-agnostic. 4. Surfaces and views both keep track of which outputs they are on. 5. The weston_surface structure now has width and height fields. These are populated when a new buffer is attached before surface.configure is called. This is because there are many surface-based operations that really require the width and height and digging through the views didn't work well. Signed-off-by: Jason Ekstrand <jason@jlekstrand.net>
2013-10-13 07:38:11 +04:00
struct weston_view *view, *next;
compositor: introduce sub-surfaces Implement the basic protocol for sub-surfaces: - expose wl_subcompositor global interface - error checking on protocol calls - associate a parent wl_surface to a sub-surface - introduce the sub-surface role, which is exclusive - an implementation of the wl_subsurface interface - allow nesting of sub-surfaces - proper surface transformation inheritance from parent to sub-surfaces - two different modes of wl_surface.commit for sub-surfaces - hook sub-surfaces up to repaint by modifying the repaint list code Struct weston_subsurface is dynamically allocated. For sub-surfaces, it is completely populated. For parent surfaces, weston_subsurface acts only as a link for stacking order purposes. The wl_resource is unused, parent_destroy_listener is not registered, the transform is not linked, etc. Sub-surfaces are not added directly into layers for display or input. Instead, they are hooked up via the sub-surface list present in parent weston_surface. This way sub-surfaces are inherently linked to the parent surface, and cannot be displayed unless the parent is mapped, too. This also eases restacking, as only the parent will be in a layer list. Also, only the main surface should be subject to shell actions. The surface list rebuilding in weston_output_repaint() is modified to process sub-surface lists, if they are non-empty. The sub-surface list always contains the parent, too, unless empty. The collection of frame_callback_list is moved to a later loop, to streamline the surface list rebuild functions. Features still lacking are: - full-surface alpha support for compound windows Changes in v2: - fix a bug in surface mapping: commit a sub-surface would cause the main surface to never be mapped. - remove debug printfs - detect attempt of making a surface its own parent - always zero-alloc weston_subsurface - apply wl_subsurface.set_position in commit, not immediately - add weston_surface_to_subsurface() - implement sub-surface commit modes parent-cached and independent - implement wl_subcompositor.destroy and wl_subsurface.destroy Changes in v3: - rebased, and use the new transform inheritance code - squashed the commit "add sub-surfaces to repaint list" - fixed a buffer reference leak in commit_from_cache() - Rewrite the sub-surface destructor code, and make it leave the wl_subsurface protocol object inert, if one destroys the corresponding wl_surface. - replaced set_commit_mode with set_sync and set_desync - allowed sub-surface nesting, and fixed repaint accordingly - implemented nested sub-surface commit modes - Made the sub-surface order changes from wl_subsurface.place_above and .place_below to be applied when the parent surface state is applied, instead of immediately. This conforms with the protocol specification now. Signed-off-by: Pekka Paalanen <ppaalanen@gmail.com>
2013-04-25 14:57:42 +04:00
assert(sub->surface);
if (sub->resource) {
assert(weston_surface_to_subsurface(sub->surface) == sub);
assert(sub->parent_destroy_listener.notify ==
subsurface_handle_parent_destroy);
wl_list_for_each_safe(view, next, &sub->surface->views, surface_link) {
weston_view_unmap(view);
Split the geometry information from weston_surface out into weston_view The weston_surface structure is split into two structures: * The weston_surface structure storres everything required for a client-side or server-side surface. This includes buffers; callbacks; backend private data; input, damage, and opaque regions; and a few other bookkeeping bits. * The weston_view structure represents an entity in the scenegraph and storres all of the geometry information. This includes clip region, alpha, position, and the transformation list as well as all of the temporary information derived from the geometry state. Because a view, and not a surface, is a scenegraph element, the view is what is placed in layers and planes. There are a few things worth noting about the surface/view split: 1. This is *not* a modification to the protocol. It is, instead, a modification to Weston's internal scenegraph to allow a single surface to exist in multiple places at a time. Clients are completely unaware of how many views to a particular surface exist. 2. A view is considered a direct child of a surface and is destroyed when the surface is destroyed. Because of this, the view.surface pointer is always valid and non-null. 3. The compositor's surface_list is replaced with a view_list. Due to subsurfaces, building the view list is a little more complicated than it used to be and involves building a tree of views on the fly whenever subsurfaces are used. However, this means that backends can remain completely subsurface-agnostic. 4. Surfaces and views both keep track of which outputs they are on. 5. The weston_surface structure now has width and height fields. These are populated when a new buffer is attached before surface.configure is called. This is because there are many surface-based operations that really require the width and height and digging through the views didn't work well. Signed-off-by: Jason Ekstrand <jason@jlekstrand.net>
2013-10-13 07:38:11 +04:00
weston_view_destroy(view);
}
Split the geometry information from weston_surface out into weston_view The weston_surface structure is split into two structures: * The weston_surface structure storres everything required for a client-side or server-side surface. This includes buffers; callbacks; backend private data; input, damage, and opaque regions; and a few other bookkeeping bits. * The weston_view structure represents an entity in the scenegraph and storres all of the geometry information. This includes clip region, alpha, position, and the transformation list as well as all of the temporary information derived from the geometry state. Because a view, and not a surface, is a scenegraph element, the view is what is placed in layers and planes. There are a few things worth noting about the surface/view split: 1. This is *not* a modification to the protocol. It is, instead, a modification to Weston's internal scenegraph to allow a single surface to exist in multiple places at a time. Clients are completely unaware of how many views to a particular surface exist. 2. A view is considered a direct child of a surface and is destroyed when the surface is destroyed. Because of this, the view.surface pointer is always valid and non-null. 3. The compositor's surface_list is replaced with a view_list. Due to subsurfaces, building the view list is a little more complicated than it used to be and involves building a tree of views on the fly whenever subsurfaces are used. However, this means that backends can remain completely subsurface-agnostic. 4. Surfaces and views both keep track of which outputs they are on. 5. The weston_surface structure now has width and height fields. These are populated when a new buffer is attached before surface.configure is called. This is because there are many surface-based operations that really require the width and height and digging through the views didn't work well. Signed-off-by: Jason Ekstrand <jason@jlekstrand.net>
2013-10-13 07:38:11 +04:00
compositor: introduce sub-surfaces Implement the basic protocol for sub-surfaces: - expose wl_subcompositor global interface - error checking on protocol calls - associate a parent wl_surface to a sub-surface - introduce the sub-surface role, which is exclusive - an implementation of the wl_subsurface interface - allow nesting of sub-surfaces - proper surface transformation inheritance from parent to sub-surfaces - two different modes of wl_surface.commit for sub-surfaces - hook sub-surfaces up to repaint by modifying the repaint list code Struct weston_subsurface is dynamically allocated. For sub-surfaces, it is completely populated. For parent surfaces, weston_subsurface acts only as a link for stacking order purposes. The wl_resource is unused, parent_destroy_listener is not registered, the transform is not linked, etc. Sub-surfaces are not added directly into layers for display or input. Instead, they are hooked up via the sub-surface list present in parent weston_surface. This way sub-surfaces are inherently linked to the parent surface, and cannot be displayed unless the parent is mapped, too. This also eases restacking, as only the parent will be in a layer list. Also, only the main surface should be subject to shell actions. The surface list rebuilding in weston_output_repaint() is modified to process sub-surface lists, if they are non-empty. The sub-surface list always contains the parent, too, unless empty. The collection of frame_callback_list is moved to a later loop, to streamline the surface list rebuild functions. Features still lacking are: - full-surface alpha support for compound windows Changes in v2: - fix a bug in surface mapping: commit a sub-surface would cause the main surface to never be mapped. - remove debug printfs - detect attempt of making a surface its own parent - always zero-alloc weston_subsurface - apply wl_subsurface.set_position in commit, not immediately - add weston_surface_to_subsurface() - implement sub-surface commit modes parent-cached and independent - implement wl_subcompositor.destroy and wl_subsurface.destroy Changes in v3: - rebased, and use the new transform inheritance code - squashed the commit "add sub-surfaces to repaint list" - fixed a buffer reference leak in commit_from_cache() - Rewrite the sub-surface destructor code, and make it leave the wl_subsurface protocol object inert, if one destroys the corresponding wl_surface. - replaced set_commit_mode with set_sync and set_desync - allowed sub-surface nesting, and fixed repaint accordingly - implemented nested sub-surface commit modes - Made the sub-surface order changes from wl_subsurface.place_above and .place_below to be applied when the parent surface state is applied, instead of immediately. This conforms with the protocol specification now. Signed-off-by: Pekka Paalanen <ppaalanen@gmail.com>
2013-04-25 14:57:42 +04:00
if (sub->parent)
weston_subsurface_unlink_parent(sub);
weston_surface_state_fini(&sub->cached);
weston_buffer_reference(&sub->cached_buffer_ref, NULL);
compositor: introduce sub-surfaces Implement the basic protocol for sub-surfaces: - expose wl_subcompositor global interface - error checking on protocol calls - associate a parent wl_surface to a sub-surface - introduce the sub-surface role, which is exclusive - an implementation of the wl_subsurface interface - allow nesting of sub-surfaces - proper surface transformation inheritance from parent to sub-surfaces - two different modes of wl_surface.commit for sub-surfaces - hook sub-surfaces up to repaint by modifying the repaint list code Struct weston_subsurface is dynamically allocated. For sub-surfaces, it is completely populated. For parent surfaces, weston_subsurface acts only as a link for stacking order purposes. The wl_resource is unused, parent_destroy_listener is not registered, the transform is not linked, etc. Sub-surfaces are not added directly into layers for display or input. Instead, they are hooked up via the sub-surface list present in parent weston_surface. This way sub-surfaces are inherently linked to the parent surface, and cannot be displayed unless the parent is mapped, too. This also eases restacking, as only the parent will be in a layer list. Also, only the main surface should be subject to shell actions. The surface list rebuilding in weston_output_repaint() is modified to process sub-surface lists, if they are non-empty. The sub-surface list always contains the parent, too, unless empty. The collection of frame_callback_list is moved to a later loop, to streamline the surface list rebuild functions. Features still lacking are: - full-surface alpha support for compound windows Changes in v2: - fix a bug in surface mapping: commit a sub-surface would cause the main surface to never be mapped. - remove debug printfs - detect attempt of making a surface its own parent - always zero-alloc weston_subsurface - apply wl_subsurface.set_position in commit, not immediately - add weston_surface_to_subsurface() - implement sub-surface commit modes parent-cached and independent - implement wl_subcompositor.destroy and wl_subsurface.destroy Changes in v3: - rebased, and use the new transform inheritance code - squashed the commit "add sub-surfaces to repaint list" - fixed a buffer reference leak in commit_from_cache() - Rewrite the sub-surface destructor code, and make it leave the wl_subsurface protocol object inert, if one destroys the corresponding wl_surface. - replaced set_commit_mode with set_sync and set_desync - allowed sub-surface nesting, and fixed repaint accordingly - implemented nested sub-surface commit modes - Made the sub-surface order changes from wl_subsurface.place_above and .place_below to be applied when the parent surface state is applied, instead of immediately. This conforms with the protocol specification now. Signed-off-by: Pekka Paalanen <ppaalanen@gmail.com>
2013-04-25 14:57:42 +04:00
sub->surface->committed = NULL;
sub->surface->committed_private = NULL;
weston_surface_set_label_func(sub->surface, NULL);
compositor: introduce sub-surfaces Implement the basic protocol for sub-surfaces: - expose wl_subcompositor global interface - error checking on protocol calls - associate a parent wl_surface to a sub-surface - introduce the sub-surface role, which is exclusive - an implementation of the wl_subsurface interface - allow nesting of sub-surfaces - proper surface transformation inheritance from parent to sub-surfaces - two different modes of wl_surface.commit for sub-surfaces - hook sub-surfaces up to repaint by modifying the repaint list code Struct weston_subsurface is dynamically allocated. For sub-surfaces, it is completely populated. For parent surfaces, weston_subsurface acts only as a link for stacking order purposes. The wl_resource is unused, parent_destroy_listener is not registered, the transform is not linked, etc. Sub-surfaces are not added directly into layers for display or input. Instead, they are hooked up via the sub-surface list present in parent weston_surface. This way sub-surfaces are inherently linked to the parent surface, and cannot be displayed unless the parent is mapped, too. This also eases restacking, as only the parent will be in a layer list. Also, only the main surface should be subject to shell actions. The surface list rebuilding in weston_output_repaint() is modified to process sub-surface lists, if they are non-empty. The sub-surface list always contains the parent, too, unless empty. The collection of frame_callback_list is moved to a later loop, to streamline the surface list rebuild functions. Features still lacking are: - full-surface alpha support for compound windows Changes in v2: - fix a bug in surface mapping: commit a sub-surface would cause the main surface to never be mapped. - remove debug printfs - detect attempt of making a surface its own parent - always zero-alloc weston_subsurface - apply wl_subsurface.set_position in commit, not immediately - add weston_surface_to_subsurface() - implement sub-surface commit modes parent-cached and independent - implement wl_subcompositor.destroy and wl_subsurface.destroy Changes in v3: - rebased, and use the new transform inheritance code - squashed the commit "add sub-surfaces to repaint list" - fixed a buffer reference leak in commit_from_cache() - Rewrite the sub-surface destructor code, and make it leave the wl_subsurface protocol object inert, if one destroys the corresponding wl_surface. - replaced set_commit_mode with set_sync and set_desync - allowed sub-surface nesting, and fixed repaint accordingly - implemented nested sub-surface commit modes - Made the sub-surface order changes from wl_subsurface.place_above and .place_below to be applied when the parent surface state is applied, instead of immediately. This conforms with the protocol specification now. Signed-off-by: Pekka Paalanen <ppaalanen@gmail.com>
2013-04-25 14:57:42 +04:00
} else {
/* the dummy weston_subsurface for the parent itself */
assert(sub->parent_destroy_listener.notify == NULL);
wl_list_remove(&sub->parent_link);
wl_list_remove(&sub->parent_link_pending);
}
wl_list_remove(&sub->surface_destroy_listener.link);
free(sub);
}
static const struct wl_subsurface_interface subsurface_implementation = {
subsurface_destroy,
subsurface_set_position,
subsurface_place_above,
subsurface_place_below,
subsurface_set_sync,
subsurface_set_desync
};
static struct weston_subsurface *
weston_subsurface_create(uint32_t id, struct weston_surface *surface,
struct weston_surface *parent)
{
struct weston_subsurface *sub;
struct wl_client *client = wl_resource_get_client(surface->resource);
compositor: introduce sub-surfaces Implement the basic protocol for sub-surfaces: - expose wl_subcompositor global interface - error checking on protocol calls - associate a parent wl_surface to a sub-surface - introduce the sub-surface role, which is exclusive - an implementation of the wl_subsurface interface - allow nesting of sub-surfaces - proper surface transformation inheritance from parent to sub-surfaces - two different modes of wl_surface.commit for sub-surfaces - hook sub-surfaces up to repaint by modifying the repaint list code Struct weston_subsurface is dynamically allocated. For sub-surfaces, it is completely populated. For parent surfaces, weston_subsurface acts only as a link for stacking order purposes. The wl_resource is unused, parent_destroy_listener is not registered, the transform is not linked, etc. Sub-surfaces are not added directly into layers for display or input. Instead, they are hooked up via the sub-surface list present in parent weston_surface. This way sub-surfaces are inherently linked to the parent surface, and cannot be displayed unless the parent is mapped, too. This also eases restacking, as only the parent will be in a layer list. Also, only the main surface should be subject to shell actions. The surface list rebuilding in weston_output_repaint() is modified to process sub-surface lists, if they are non-empty. The sub-surface list always contains the parent, too, unless empty. The collection of frame_callback_list is moved to a later loop, to streamline the surface list rebuild functions. Features still lacking are: - full-surface alpha support for compound windows Changes in v2: - fix a bug in surface mapping: commit a sub-surface would cause the main surface to never be mapped. - remove debug printfs - detect attempt of making a surface its own parent - always zero-alloc weston_subsurface - apply wl_subsurface.set_position in commit, not immediately - add weston_surface_to_subsurface() - implement sub-surface commit modes parent-cached and independent - implement wl_subcompositor.destroy and wl_subsurface.destroy Changes in v3: - rebased, and use the new transform inheritance code - squashed the commit "add sub-surfaces to repaint list" - fixed a buffer reference leak in commit_from_cache() - Rewrite the sub-surface destructor code, and make it leave the wl_subsurface protocol object inert, if one destroys the corresponding wl_surface. - replaced set_commit_mode with set_sync and set_desync - allowed sub-surface nesting, and fixed repaint accordingly - implemented nested sub-surface commit modes - Made the sub-surface order changes from wl_subsurface.place_above and .place_below to be applied when the parent surface state is applied, instead of immediately. This conforms with the protocol specification now. Signed-off-by: Pekka Paalanen <ppaalanen@gmail.com>
2013-04-25 14:57:42 +04:00
sub = zalloc(sizeof *sub);
if (sub == NULL)
compositor: introduce sub-surfaces Implement the basic protocol for sub-surfaces: - expose wl_subcompositor global interface - error checking on protocol calls - associate a parent wl_surface to a sub-surface - introduce the sub-surface role, which is exclusive - an implementation of the wl_subsurface interface - allow nesting of sub-surfaces - proper surface transformation inheritance from parent to sub-surfaces - two different modes of wl_surface.commit for sub-surfaces - hook sub-surfaces up to repaint by modifying the repaint list code Struct weston_subsurface is dynamically allocated. For sub-surfaces, it is completely populated. For parent surfaces, weston_subsurface acts only as a link for stacking order purposes. The wl_resource is unused, parent_destroy_listener is not registered, the transform is not linked, etc. Sub-surfaces are not added directly into layers for display or input. Instead, they are hooked up via the sub-surface list present in parent weston_surface. This way sub-surfaces are inherently linked to the parent surface, and cannot be displayed unless the parent is mapped, too. This also eases restacking, as only the parent will be in a layer list. Also, only the main surface should be subject to shell actions. The surface list rebuilding in weston_output_repaint() is modified to process sub-surface lists, if they are non-empty. The sub-surface list always contains the parent, too, unless empty. The collection of frame_callback_list is moved to a later loop, to streamline the surface list rebuild functions. Features still lacking are: - full-surface alpha support for compound windows Changes in v2: - fix a bug in surface mapping: commit a sub-surface would cause the main surface to never be mapped. - remove debug printfs - detect attempt of making a surface its own parent - always zero-alloc weston_subsurface - apply wl_subsurface.set_position in commit, not immediately - add weston_surface_to_subsurface() - implement sub-surface commit modes parent-cached and independent - implement wl_subcompositor.destroy and wl_subsurface.destroy Changes in v3: - rebased, and use the new transform inheritance code - squashed the commit "add sub-surfaces to repaint list" - fixed a buffer reference leak in commit_from_cache() - Rewrite the sub-surface destructor code, and make it leave the wl_subsurface protocol object inert, if one destroys the corresponding wl_surface. - replaced set_commit_mode with set_sync and set_desync - allowed sub-surface nesting, and fixed repaint accordingly - implemented nested sub-surface commit modes - Made the sub-surface order changes from wl_subsurface.place_above and .place_below to be applied when the parent surface state is applied, instead of immediately. This conforms with the protocol specification now. Signed-off-by: Pekka Paalanen <ppaalanen@gmail.com>
2013-04-25 14:57:42 +04:00
return NULL;
Split the geometry information from weston_surface out into weston_view The weston_surface structure is split into two structures: * The weston_surface structure storres everything required for a client-side or server-side surface. This includes buffers; callbacks; backend private data; input, damage, and opaque regions; and a few other bookkeeping bits. * The weston_view structure represents an entity in the scenegraph and storres all of the geometry information. This includes clip region, alpha, position, and the transformation list as well as all of the temporary information derived from the geometry state. Because a view, and not a surface, is a scenegraph element, the view is what is placed in layers and planes. There are a few things worth noting about the surface/view split: 1. This is *not* a modification to the protocol. It is, instead, a modification to Weston's internal scenegraph to allow a single surface to exist in multiple places at a time. Clients are completely unaware of how many views to a particular surface exist. 2. A view is considered a direct child of a surface and is destroyed when the surface is destroyed. Because of this, the view.surface pointer is always valid and non-null. 3. The compositor's surface_list is replaced with a view_list. Due to subsurfaces, building the view list is a little more complicated than it used to be and involves building a tree of views on the fly whenever subsurfaces are used. However, this means that backends can remain completely subsurface-agnostic. 4. Surfaces and views both keep track of which outputs they are on. 5. The weston_surface structure now has width and height fields. These are populated when a new buffer is attached before surface.configure is called. This is because there are many surface-based operations that really require the width and height and digging through the views didn't work well. Signed-off-by: Jason Ekstrand <jason@jlekstrand.net>
2013-10-13 07:38:11 +04:00
wl_list_init(&sub->unused_views);
sub->resource =
wl_resource_create(client, &wl_subsurface_interface, 1, id);
compositor: introduce sub-surfaces Implement the basic protocol for sub-surfaces: - expose wl_subcompositor global interface - error checking on protocol calls - associate a parent wl_surface to a sub-surface - introduce the sub-surface role, which is exclusive - an implementation of the wl_subsurface interface - allow nesting of sub-surfaces - proper surface transformation inheritance from parent to sub-surfaces - two different modes of wl_surface.commit for sub-surfaces - hook sub-surfaces up to repaint by modifying the repaint list code Struct weston_subsurface is dynamically allocated. For sub-surfaces, it is completely populated. For parent surfaces, weston_subsurface acts only as a link for stacking order purposes. The wl_resource is unused, parent_destroy_listener is not registered, the transform is not linked, etc. Sub-surfaces are not added directly into layers for display or input. Instead, they are hooked up via the sub-surface list present in parent weston_surface. This way sub-surfaces are inherently linked to the parent surface, and cannot be displayed unless the parent is mapped, too. This also eases restacking, as only the parent will be in a layer list. Also, only the main surface should be subject to shell actions. The surface list rebuilding in weston_output_repaint() is modified to process sub-surface lists, if they are non-empty. The sub-surface list always contains the parent, too, unless empty. The collection of frame_callback_list is moved to a later loop, to streamline the surface list rebuild functions. Features still lacking are: - full-surface alpha support for compound windows Changes in v2: - fix a bug in surface mapping: commit a sub-surface would cause the main surface to never be mapped. - remove debug printfs - detect attempt of making a surface its own parent - always zero-alloc weston_subsurface - apply wl_subsurface.set_position in commit, not immediately - add weston_surface_to_subsurface() - implement sub-surface commit modes parent-cached and independent - implement wl_subcompositor.destroy and wl_subsurface.destroy Changes in v3: - rebased, and use the new transform inheritance code - squashed the commit "add sub-surfaces to repaint list" - fixed a buffer reference leak in commit_from_cache() - Rewrite the sub-surface destructor code, and make it leave the wl_subsurface protocol object inert, if one destroys the corresponding wl_surface. - replaced set_commit_mode with set_sync and set_desync - allowed sub-surface nesting, and fixed repaint accordingly - implemented nested sub-surface commit modes - Made the sub-surface order changes from wl_subsurface.place_above and .place_below to be applied when the parent surface state is applied, instead of immediately. This conforms with the protocol specification now. Signed-off-by: Pekka Paalanen <ppaalanen@gmail.com>
2013-04-25 14:57:42 +04:00
if (!sub->resource) {
free(sub);
return NULL;
}
wl_resource_set_implementation(sub->resource,
&subsurface_implementation,
sub, subsurface_resource_destroy);
compositor: introduce sub-surfaces Implement the basic protocol for sub-surfaces: - expose wl_subcompositor global interface - error checking on protocol calls - associate a parent wl_surface to a sub-surface - introduce the sub-surface role, which is exclusive - an implementation of the wl_subsurface interface - allow nesting of sub-surfaces - proper surface transformation inheritance from parent to sub-surfaces - two different modes of wl_surface.commit for sub-surfaces - hook sub-surfaces up to repaint by modifying the repaint list code Struct weston_subsurface is dynamically allocated. For sub-surfaces, it is completely populated. For parent surfaces, weston_subsurface acts only as a link for stacking order purposes. The wl_resource is unused, parent_destroy_listener is not registered, the transform is not linked, etc. Sub-surfaces are not added directly into layers for display or input. Instead, they are hooked up via the sub-surface list present in parent weston_surface. This way sub-surfaces are inherently linked to the parent surface, and cannot be displayed unless the parent is mapped, too. This also eases restacking, as only the parent will be in a layer list. Also, only the main surface should be subject to shell actions. The surface list rebuilding in weston_output_repaint() is modified to process sub-surface lists, if they are non-empty. The sub-surface list always contains the parent, too, unless empty. The collection of frame_callback_list is moved to a later loop, to streamline the surface list rebuild functions. Features still lacking are: - full-surface alpha support for compound windows Changes in v2: - fix a bug in surface mapping: commit a sub-surface would cause the main surface to never be mapped. - remove debug printfs - detect attempt of making a surface its own parent - always zero-alloc weston_subsurface - apply wl_subsurface.set_position in commit, not immediately - add weston_surface_to_subsurface() - implement sub-surface commit modes parent-cached and independent - implement wl_subcompositor.destroy and wl_subsurface.destroy Changes in v3: - rebased, and use the new transform inheritance code - squashed the commit "add sub-surfaces to repaint list" - fixed a buffer reference leak in commit_from_cache() - Rewrite the sub-surface destructor code, and make it leave the wl_subsurface protocol object inert, if one destroys the corresponding wl_surface. - replaced set_commit_mode with set_sync and set_desync - allowed sub-surface nesting, and fixed repaint accordingly - implemented nested sub-surface commit modes - Made the sub-surface order changes from wl_subsurface.place_above and .place_below to be applied when the parent surface state is applied, instead of immediately. This conforms with the protocol specification now. Signed-off-by: Pekka Paalanen <ppaalanen@gmail.com>
2013-04-25 14:57:42 +04:00
weston_subsurface_link_surface(sub, surface);
weston_subsurface_link_parent(sub, parent);
weston_surface_state_init(&sub->cached);
sub->cached_buffer_ref.buffer = NULL;
compositor: introduce sub-surfaces Implement the basic protocol for sub-surfaces: - expose wl_subcompositor global interface - error checking on protocol calls - associate a parent wl_surface to a sub-surface - introduce the sub-surface role, which is exclusive - an implementation of the wl_subsurface interface - allow nesting of sub-surfaces - proper surface transformation inheritance from parent to sub-surfaces - two different modes of wl_surface.commit for sub-surfaces - hook sub-surfaces up to repaint by modifying the repaint list code Struct weston_subsurface is dynamically allocated. For sub-surfaces, it is completely populated. For parent surfaces, weston_subsurface acts only as a link for stacking order purposes. The wl_resource is unused, parent_destroy_listener is not registered, the transform is not linked, etc. Sub-surfaces are not added directly into layers for display or input. Instead, they are hooked up via the sub-surface list present in parent weston_surface. This way sub-surfaces are inherently linked to the parent surface, and cannot be displayed unless the parent is mapped, too. This also eases restacking, as only the parent will be in a layer list. Also, only the main surface should be subject to shell actions. The surface list rebuilding in weston_output_repaint() is modified to process sub-surface lists, if they are non-empty. The sub-surface list always contains the parent, too, unless empty. The collection of frame_callback_list is moved to a later loop, to streamline the surface list rebuild functions. Features still lacking are: - full-surface alpha support for compound windows Changes in v2: - fix a bug in surface mapping: commit a sub-surface would cause the main surface to never be mapped. - remove debug printfs - detect attempt of making a surface its own parent - always zero-alloc weston_subsurface - apply wl_subsurface.set_position in commit, not immediately - add weston_surface_to_subsurface() - implement sub-surface commit modes parent-cached and independent - implement wl_subcompositor.destroy and wl_subsurface.destroy Changes in v3: - rebased, and use the new transform inheritance code - squashed the commit "add sub-surfaces to repaint list" - fixed a buffer reference leak in commit_from_cache() - Rewrite the sub-surface destructor code, and make it leave the wl_subsurface protocol object inert, if one destroys the corresponding wl_surface. - replaced set_commit_mode with set_sync and set_desync - allowed sub-surface nesting, and fixed repaint accordingly - implemented nested sub-surface commit modes - Made the sub-surface order changes from wl_subsurface.place_above and .place_below to be applied when the parent surface state is applied, instead of immediately. This conforms with the protocol specification now. Signed-off-by: Pekka Paalanen <ppaalanen@gmail.com>
2013-04-25 14:57:42 +04:00
sub->synchronized = 1;
return sub;
}
/* Create a dummy subsurface for having the parent itself in its
* sub-surface lists. Makes stacking order manipulation easy.
*/
static struct weston_subsurface *
weston_subsurface_create_for_parent(struct weston_surface *parent)
{
struct weston_subsurface *sub;
sub = zalloc(sizeof *sub);
if (sub == NULL)
compositor: introduce sub-surfaces Implement the basic protocol for sub-surfaces: - expose wl_subcompositor global interface - error checking on protocol calls - associate a parent wl_surface to a sub-surface - introduce the sub-surface role, which is exclusive - an implementation of the wl_subsurface interface - allow nesting of sub-surfaces - proper surface transformation inheritance from parent to sub-surfaces - two different modes of wl_surface.commit for sub-surfaces - hook sub-surfaces up to repaint by modifying the repaint list code Struct weston_subsurface is dynamically allocated. For sub-surfaces, it is completely populated. For parent surfaces, weston_subsurface acts only as a link for stacking order purposes. The wl_resource is unused, parent_destroy_listener is not registered, the transform is not linked, etc. Sub-surfaces are not added directly into layers for display or input. Instead, they are hooked up via the sub-surface list present in parent weston_surface. This way sub-surfaces are inherently linked to the parent surface, and cannot be displayed unless the parent is mapped, too. This also eases restacking, as only the parent will be in a layer list. Also, only the main surface should be subject to shell actions. The surface list rebuilding in weston_output_repaint() is modified to process sub-surface lists, if they are non-empty. The sub-surface list always contains the parent, too, unless empty. The collection of frame_callback_list is moved to a later loop, to streamline the surface list rebuild functions. Features still lacking are: - full-surface alpha support for compound windows Changes in v2: - fix a bug in surface mapping: commit a sub-surface would cause the main surface to never be mapped. - remove debug printfs - detect attempt of making a surface its own parent - always zero-alloc weston_subsurface - apply wl_subsurface.set_position in commit, not immediately - add weston_surface_to_subsurface() - implement sub-surface commit modes parent-cached and independent - implement wl_subcompositor.destroy and wl_subsurface.destroy Changes in v3: - rebased, and use the new transform inheritance code - squashed the commit "add sub-surfaces to repaint list" - fixed a buffer reference leak in commit_from_cache() - Rewrite the sub-surface destructor code, and make it leave the wl_subsurface protocol object inert, if one destroys the corresponding wl_surface. - replaced set_commit_mode with set_sync and set_desync - allowed sub-surface nesting, and fixed repaint accordingly - implemented nested sub-surface commit modes - Made the sub-surface order changes from wl_subsurface.place_above and .place_below to be applied when the parent surface state is applied, instead of immediately. This conforms with the protocol specification now. Signed-off-by: Pekka Paalanen <ppaalanen@gmail.com>
2013-04-25 14:57:42 +04:00
return NULL;
weston_subsurface_link_surface(sub, parent);
sub->parent = parent;
wl_list_insert(&parent->subsurface_list, &sub->parent_link);
wl_list_insert(&parent->subsurface_list_pending,
&sub->parent_link_pending);
return sub;
}
static void
subcompositor_get_subsurface(struct wl_client *client,
struct wl_resource *resource,
uint32_t id,
struct wl_resource *surface_resource,
struct wl_resource *parent_resource)
{
struct weston_surface *surface =
wl_resource_get_user_data(surface_resource);
struct weston_surface *parent =
wl_resource_get_user_data(parent_resource);
compositor: introduce sub-surfaces Implement the basic protocol for sub-surfaces: - expose wl_subcompositor global interface - error checking on protocol calls - associate a parent wl_surface to a sub-surface - introduce the sub-surface role, which is exclusive - an implementation of the wl_subsurface interface - allow nesting of sub-surfaces - proper surface transformation inheritance from parent to sub-surfaces - two different modes of wl_surface.commit for sub-surfaces - hook sub-surfaces up to repaint by modifying the repaint list code Struct weston_subsurface is dynamically allocated. For sub-surfaces, it is completely populated. For parent surfaces, weston_subsurface acts only as a link for stacking order purposes. The wl_resource is unused, parent_destroy_listener is not registered, the transform is not linked, etc. Sub-surfaces are not added directly into layers for display or input. Instead, they are hooked up via the sub-surface list present in parent weston_surface. This way sub-surfaces are inherently linked to the parent surface, and cannot be displayed unless the parent is mapped, too. This also eases restacking, as only the parent will be in a layer list. Also, only the main surface should be subject to shell actions. The surface list rebuilding in weston_output_repaint() is modified to process sub-surface lists, if they are non-empty. The sub-surface list always contains the parent, too, unless empty. The collection of frame_callback_list is moved to a later loop, to streamline the surface list rebuild functions. Features still lacking are: - full-surface alpha support for compound windows Changes in v2: - fix a bug in surface mapping: commit a sub-surface would cause the main surface to never be mapped. - remove debug printfs - detect attempt of making a surface its own parent - always zero-alloc weston_subsurface - apply wl_subsurface.set_position in commit, not immediately - add weston_surface_to_subsurface() - implement sub-surface commit modes parent-cached and independent - implement wl_subcompositor.destroy and wl_subsurface.destroy Changes in v3: - rebased, and use the new transform inheritance code - squashed the commit "add sub-surfaces to repaint list" - fixed a buffer reference leak in commit_from_cache() - Rewrite the sub-surface destructor code, and make it leave the wl_subsurface protocol object inert, if one destroys the corresponding wl_surface. - replaced set_commit_mode with set_sync and set_desync - allowed sub-surface nesting, and fixed repaint accordingly - implemented nested sub-surface commit modes - Made the sub-surface order changes from wl_subsurface.place_above and .place_below to be applied when the parent surface state is applied, instead of immediately. This conforms with the protocol specification now. Signed-off-by: Pekka Paalanen <ppaalanen@gmail.com>
2013-04-25 14:57:42 +04:00
struct weston_subsurface *sub;
static const char where[] = "get_subsurface: wl_subsurface@";
if (surface == parent) {
wl_resource_post_error(resource,
WL_SUBCOMPOSITOR_ERROR_BAD_SURFACE,
"%s%d: wl_surface@%d cannot be its own parent",
where, id, wl_resource_get_id(surface_resource));
compositor: introduce sub-surfaces Implement the basic protocol for sub-surfaces: - expose wl_subcompositor global interface - error checking on protocol calls - associate a parent wl_surface to a sub-surface - introduce the sub-surface role, which is exclusive - an implementation of the wl_subsurface interface - allow nesting of sub-surfaces - proper surface transformation inheritance from parent to sub-surfaces - two different modes of wl_surface.commit for sub-surfaces - hook sub-surfaces up to repaint by modifying the repaint list code Struct weston_subsurface is dynamically allocated. For sub-surfaces, it is completely populated. For parent surfaces, weston_subsurface acts only as a link for stacking order purposes. The wl_resource is unused, parent_destroy_listener is not registered, the transform is not linked, etc. Sub-surfaces are not added directly into layers for display or input. Instead, they are hooked up via the sub-surface list present in parent weston_surface. This way sub-surfaces are inherently linked to the parent surface, and cannot be displayed unless the parent is mapped, too. This also eases restacking, as only the parent will be in a layer list. Also, only the main surface should be subject to shell actions. The surface list rebuilding in weston_output_repaint() is modified to process sub-surface lists, if they are non-empty. The sub-surface list always contains the parent, too, unless empty. The collection of frame_callback_list is moved to a later loop, to streamline the surface list rebuild functions. Features still lacking are: - full-surface alpha support for compound windows Changes in v2: - fix a bug in surface mapping: commit a sub-surface would cause the main surface to never be mapped. - remove debug printfs - detect attempt of making a surface its own parent - always zero-alloc weston_subsurface - apply wl_subsurface.set_position in commit, not immediately - add weston_surface_to_subsurface() - implement sub-surface commit modes parent-cached and independent - implement wl_subcompositor.destroy and wl_subsurface.destroy Changes in v3: - rebased, and use the new transform inheritance code - squashed the commit "add sub-surfaces to repaint list" - fixed a buffer reference leak in commit_from_cache() - Rewrite the sub-surface destructor code, and make it leave the wl_subsurface protocol object inert, if one destroys the corresponding wl_surface. - replaced set_commit_mode with set_sync and set_desync - allowed sub-surface nesting, and fixed repaint accordingly - implemented nested sub-surface commit modes - Made the sub-surface order changes from wl_subsurface.place_above and .place_below to be applied when the parent surface state is applied, instead of immediately. This conforms with the protocol specification now. Signed-off-by: Pekka Paalanen <ppaalanen@gmail.com>
2013-04-25 14:57:42 +04:00
return;
}
if (weston_surface_to_subsurface(surface)) {
wl_resource_post_error(resource,
WL_SUBCOMPOSITOR_ERROR_BAD_SURFACE,
"%s%d: wl_surface@%d is already a sub-surface",
where, id, wl_resource_get_id(surface_resource));
compositor: introduce sub-surfaces Implement the basic protocol for sub-surfaces: - expose wl_subcompositor global interface - error checking on protocol calls - associate a parent wl_surface to a sub-surface - introduce the sub-surface role, which is exclusive - an implementation of the wl_subsurface interface - allow nesting of sub-surfaces - proper surface transformation inheritance from parent to sub-surfaces - two different modes of wl_surface.commit for sub-surfaces - hook sub-surfaces up to repaint by modifying the repaint list code Struct weston_subsurface is dynamically allocated. For sub-surfaces, it is completely populated. For parent surfaces, weston_subsurface acts only as a link for stacking order purposes. The wl_resource is unused, parent_destroy_listener is not registered, the transform is not linked, etc. Sub-surfaces are not added directly into layers for display or input. Instead, they are hooked up via the sub-surface list present in parent weston_surface. This way sub-surfaces are inherently linked to the parent surface, and cannot be displayed unless the parent is mapped, too. This also eases restacking, as only the parent will be in a layer list. Also, only the main surface should be subject to shell actions. The surface list rebuilding in weston_output_repaint() is modified to process sub-surface lists, if they are non-empty. The sub-surface list always contains the parent, too, unless empty. The collection of frame_callback_list is moved to a later loop, to streamline the surface list rebuild functions. Features still lacking are: - full-surface alpha support for compound windows Changes in v2: - fix a bug in surface mapping: commit a sub-surface would cause the main surface to never be mapped. - remove debug printfs - detect attempt of making a surface its own parent - always zero-alloc weston_subsurface - apply wl_subsurface.set_position in commit, not immediately - add weston_surface_to_subsurface() - implement sub-surface commit modes parent-cached and independent - implement wl_subcompositor.destroy and wl_subsurface.destroy Changes in v3: - rebased, and use the new transform inheritance code - squashed the commit "add sub-surfaces to repaint list" - fixed a buffer reference leak in commit_from_cache() - Rewrite the sub-surface destructor code, and make it leave the wl_subsurface protocol object inert, if one destroys the corresponding wl_surface. - replaced set_commit_mode with set_sync and set_desync - allowed sub-surface nesting, and fixed repaint accordingly - implemented nested sub-surface commit modes - Made the sub-surface order changes from wl_subsurface.place_above and .place_below to be applied when the parent surface state is applied, instead of immediately. This conforms with the protocol specification now. Signed-off-by: Pekka Paalanen <ppaalanen@gmail.com>
2013-04-25 14:57:42 +04:00
return;
}
2014-10-01 16:02:41 +04:00
if (weston_surface_set_role(surface, "wl_subsurface", resource,
WL_SUBCOMPOSITOR_ERROR_BAD_SURFACE) < 0)
compositor: introduce sub-surfaces Implement the basic protocol for sub-surfaces: - expose wl_subcompositor global interface - error checking on protocol calls - associate a parent wl_surface to a sub-surface - introduce the sub-surface role, which is exclusive - an implementation of the wl_subsurface interface - allow nesting of sub-surfaces - proper surface transformation inheritance from parent to sub-surfaces - two different modes of wl_surface.commit for sub-surfaces - hook sub-surfaces up to repaint by modifying the repaint list code Struct weston_subsurface is dynamically allocated. For sub-surfaces, it is completely populated. For parent surfaces, weston_subsurface acts only as a link for stacking order purposes. The wl_resource is unused, parent_destroy_listener is not registered, the transform is not linked, etc. Sub-surfaces are not added directly into layers for display or input. Instead, they are hooked up via the sub-surface list present in parent weston_surface. This way sub-surfaces are inherently linked to the parent surface, and cannot be displayed unless the parent is mapped, too. This also eases restacking, as only the parent will be in a layer list. Also, only the main surface should be subject to shell actions. The surface list rebuilding in weston_output_repaint() is modified to process sub-surface lists, if they are non-empty. The sub-surface list always contains the parent, too, unless empty. The collection of frame_callback_list is moved to a later loop, to streamline the surface list rebuild functions. Features still lacking are: - full-surface alpha support for compound windows Changes in v2: - fix a bug in surface mapping: commit a sub-surface would cause the main surface to never be mapped. - remove debug printfs - detect attempt of making a surface its own parent - always zero-alloc weston_subsurface - apply wl_subsurface.set_position in commit, not immediately - add weston_surface_to_subsurface() - implement sub-surface commit modes parent-cached and independent - implement wl_subcompositor.destroy and wl_subsurface.destroy Changes in v3: - rebased, and use the new transform inheritance code - squashed the commit "add sub-surfaces to repaint list" - fixed a buffer reference leak in commit_from_cache() - Rewrite the sub-surface destructor code, and make it leave the wl_subsurface protocol object inert, if one destroys the corresponding wl_surface. - replaced set_commit_mode with set_sync and set_desync - allowed sub-surface nesting, and fixed repaint accordingly - implemented nested sub-surface commit modes - Made the sub-surface order changes from wl_subsurface.place_above and .place_below to be applied when the parent surface state is applied, instead of immediately. This conforms with the protocol specification now. Signed-off-by: Pekka Paalanen <ppaalanen@gmail.com>
2013-04-25 14:57:42 +04:00
return;
if (weston_surface_get_main_surface(parent) == surface) {
wl_resource_post_error(resource,
WL_SUBCOMPOSITOR_ERROR_BAD_SURFACE,
"%s%d: wl_surface@%d is an ancestor of parent",
where, id, wl_resource_get_id(surface_resource));
return;
}
compositor: introduce sub-surfaces Implement the basic protocol for sub-surfaces: - expose wl_subcompositor global interface - error checking on protocol calls - associate a parent wl_surface to a sub-surface - introduce the sub-surface role, which is exclusive - an implementation of the wl_subsurface interface - allow nesting of sub-surfaces - proper surface transformation inheritance from parent to sub-surfaces - two different modes of wl_surface.commit for sub-surfaces - hook sub-surfaces up to repaint by modifying the repaint list code Struct weston_subsurface is dynamically allocated. For sub-surfaces, it is completely populated. For parent surfaces, weston_subsurface acts only as a link for stacking order purposes. The wl_resource is unused, parent_destroy_listener is not registered, the transform is not linked, etc. Sub-surfaces are not added directly into layers for display or input. Instead, they are hooked up via the sub-surface list present in parent weston_surface. This way sub-surfaces are inherently linked to the parent surface, and cannot be displayed unless the parent is mapped, too. This also eases restacking, as only the parent will be in a layer list. Also, only the main surface should be subject to shell actions. The surface list rebuilding in weston_output_repaint() is modified to process sub-surface lists, if they are non-empty. The sub-surface list always contains the parent, too, unless empty. The collection of frame_callback_list is moved to a later loop, to streamline the surface list rebuild functions. Features still lacking are: - full-surface alpha support for compound windows Changes in v2: - fix a bug in surface mapping: commit a sub-surface would cause the main surface to never be mapped. - remove debug printfs - detect attempt of making a surface its own parent - always zero-alloc weston_subsurface - apply wl_subsurface.set_position in commit, not immediately - add weston_surface_to_subsurface() - implement sub-surface commit modes parent-cached and independent - implement wl_subcompositor.destroy and wl_subsurface.destroy Changes in v3: - rebased, and use the new transform inheritance code - squashed the commit "add sub-surfaces to repaint list" - fixed a buffer reference leak in commit_from_cache() - Rewrite the sub-surface destructor code, and make it leave the wl_subsurface protocol object inert, if one destroys the corresponding wl_surface. - replaced set_commit_mode with set_sync and set_desync - allowed sub-surface nesting, and fixed repaint accordingly - implemented nested sub-surface commit modes - Made the sub-surface order changes from wl_subsurface.place_above and .place_below to be applied when the parent surface state is applied, instead of immediately. This conforms with the protocol specification now. Signed-off-by: Pekka Paalanen <ppaalanen@gmail.com>
2013-04-25 14:57:42 +04:00
/* make sure the parent is in its own list */
if (wl_list_empty(&parent->subsurface_list)) {
if (!weston_subsurface_create_for_parent(parent)) {
wl_resource_post_no_memory(resource);
return;
}
}
sub = weston_subsurface_create(id, surface, parent);
if (!sub) {
wl_resource_post_no_memory(resource);
return;
}
surface->committed = subsurface_committed;
surface->committed_private = sub;
weston_surface_set_label_func(surface, subsurface_get_label);
compositor: introduce sub-surfaces Implement the basic protocol for sub-surfaces: - expose wl_subcompositor global interface - error checking on protocol calls - associate a parent wl_surface to a sub-surface - introduce the sub-surface role, which is exclusive - an implementation of the wl_subsurface interface - allow nesting of sub-surfaces - proper surface transformation inheritance from parent to sub-surfaces - two different modes of wl_surface.commit for sub-surfaces - hook sub-surfaces up to repaint by modifying the repaint list code Struct weston_subsurface is dynamically allocated. For sub-surfaces, it is completely populated. For parent surfaces, weston_subsurface acts only as a link for stacking order purposes. The wl_resource is unused, parent_destroy_listener is not registered, the transform is not linked, etc. Sub-surfaces are not added directly into layers for display or input. Instead, they are hooked up via the sub-surface list present in parent weston_surface. This way sub-surfaces are inherently linked to the parent surface, and cannot be displayed unless the parent is mapped, too. This also eases restacking, as only the parent will be in a layer list. Also, only the main surface should be subject to shell actions. The surface list rebuilding in weston_output_repaint() is modified to process sub-surface lists, if they are non-empty. The sub-surface list always contains the parent, too, unless empty. The collection of frame_callback_list is moved to a later loop, to streamline the surface list rebuild functions. Features still lacking are: - full-surface alpha support for compound windows Changes in v2: - fix a bug in surface mapping: commit a sub-surface would cause the main surface to never be mapped. - remove debug printfs - detect attempt of making a surface its own parent - always zero-alloc weston_subsurface - apply wl_subsurface.set_position in commit, not immediately - add weston_surface_to_subsurface() - implement sub-surface commit modes parent-cached and independent - implement wl_subcompositor.destroy and wl_subsurface.destroy Changes in v3: - rebased, and use the new transform inheritance code - squashed the commit "add sub-surfaces to repaint list" - fixed a buffer reference leak in commit_from_cache() - Rewrite the sub-surface destructor code, and make it leave the wl_subsurface protocol object inert, if one destroys the corresponding wl_surface. - replaced set_commit_mode with set_sync and set_desync - allowed sub-surface nesting, and fixed repaint accordingly - implemented nested sub-surface commit modes - Made the sub-surface order changes from wl_subsurface.place_above and .place_below to be applied when the parent surface state is applied, instead of immediately. This conforms with the protocol specification now. Signed-off-by: Pekka Paalanen <ppaalanen@gmail.com>
2013-04-25 14:57:42 +04:00
}
static void
subcompositor_destroy(struct wl_client *client, struct wl_resource *resource)
{
wl_resource_destroy(resource);
}
static const struct wl_subcompositor_interface subcompositor_interface = {
subcompositor_destroy,
subcompositor_get_subsurface
};
static void
bind_subcompositor(struct wl_client *client,
void *data, uint32_t version, uint32_t id)
{
struct weston_compositor *compositor = data;
struct wl_resource *resource;
compositor: introduce sub-surfaces Implement the basic protocol for sub-surfaces: - expose wl_subcompositor global interface - error checking on protocol calls - associate a parent wl_surface to a sub-surface - introduce the sub-surface role, which is exclusive - an implementation of the wl_subsurface interface - allow nesting of sub-surfaces - proper surface transformation inheritance from parent to sub-surfaces - two different modes of wl_surface.commit for sub-surfaces - hook sub-surfaces up to repaint by modifying the repaint list code Struct weston_subsurface is dynamically allocated. For sub-surfaces, it is completely populated. For parent surfaces, weston_subsurface acts only as a link for stacking order purposes. The wl_resource is unused, parent_destroy_listener is not registered, the transform is not linked, etc. Sub-surfaces are not added directly into layers for display or input. Instead, they are hooked up via the sub-surface list present in parent weston_surface. This way sub-surfaces are inherently linked to the parent surface, and cannot be displayed unless the parent is mapped, too. This also eases restacking, as only the parent will be in a layer list. Also, only the main surface should be subject to shell actions. The surface list rebuilding in weston_output_repaint() is modified to process sub-surface lists, if they are non-empty. The sub-surface list always contains the parent, too, unless empty. The collection of frame_callback_list is moved to a later loop, to streamline the surface list rebuild functions. Features still lacking are: - full-surface alpha support for compound windows Changes in v2: - fix a bug in surface mapping: commit a sub-surface would cause the main surface to never be mapped. - remove debug printfs - detect attempt of making a surface its own parent - always zero-alloc weston_subsurface - apply wl_subsurface.set_position in commit, not immediately - add weston_surface_to_subsurface() - implement sub-surface commit modes parent-cached and independent - implement wl_subcompositor.destroy and wl_subsurface.destroy Changes in v3: - rebased, and use the new transform inheritance code - squashed the commit "add sub-surfaces to repaint list" - fixed a buffer reference leak in commit_from_cache() - Rewrite the sub-surface destructor code, and make it leave the wl_subsurface protocol object inert, if one destroys the corresponding wl_surface. - replaced set_commit_mode with set_sync and set_desync - allowed sub-surface nesting, and fixed repaint accordingly - implemented nested sub-surface commit modes - Made the sub-surface order changes from wl_subsurface.place_above and .place_below to be applied when the parent surface state is applied, instead of immediately. This conforms with the protocol specification now. Signed-off-by: Pekka Paalanen <ppaalanen@gmail.com>
2013-04-25 14:57:42 +04:00
resource =
wl_resource_create(client, &wl_subcompositor_interface, 1, id);
if (resource == NULL) {
wl_client_post_no_memory(client);
return;
}
wl_resource_set_implementation(resource, &subcompositor_interface,
compositor, NULL);
compositor: introduce sub-surfaces Implement the basic protocol for sub-surfaces: - expose wl_subcompositor global interface - error checking on protocol calls - associate a parent wl_surface to a sub-surface - introduce the sub-surface role, which is exclusive - an implementation of the wl_subsurface interface - allow nesting of sub-surfaces - proper surface transformation inheritance from parent to sub-surfaces - two different modes of wl_surface.commit for sub-surfaces - hook sub-surfaces up to repaint by modifying the repaint list code Struct weston_subsurface is dynamically allocated. For sub-surfaces, it is completely populated. For parent surfaces, weston_subsurface acts only as a link for stacking order purposes. The wl_resource is unused, parent_destroy_listener is not registered, the transform is not linked, etc. Sub-surfaces are not added directly into layers for display or input. Instead, they are hooked up via the sub-surface list present in parent weston_surface. This way sub-surfaces are inherently linked to the parent surface, and cannot be displayed unless the parent is mapped, too. This also eases restacking, as only the parent will be in a layer list. Also, only the main surface should be subject to shell actions. The surface list rebuilding in weston_output_repaint() is modified to process sub-surface lists, if they are non-empty. The sub-surface list always contains the parent, too, unless empty. The collection of frame_callback_list is moved to a later loop, to streamline the surface list rebuild functions. Features still lacking are: - full-surface alpha support for compound windows Changes in v2: - fix a bug in surface mapping: commit a sub-surface would cause the main surface to never be mapped. - remove debug printfs - detect attempt of making a surface its own parent - always zero-alloc weston_subsurface - apply wl_subsurface.set_position in commit, not immediately - add weston_surface_to_subsurface() - implement sub-surface commit modes parent-cached and independent - implement wl_subcompositor.destroy and wl_subsurface.destroy Changes in v3: - rebased, and use the new transform inheritance code - squashed the commit "add sub-surfaces to repaint list" - fixed a buffer reference leak in commit_from_cache() - Rewrite the sub-surface destructor code, and make it leave the wl_subsurface protocol object inert, if one destroys the corresponding wl_surface. - replaced set_commit_mode with set_sync and set_desync - allowed sub-surface nesting, and fixed repaint accordingly - implemented nested sub-surface commit modes - Made the sub-surface order changes from wl_subsurface.place_above and .place_below to be applied when the parent surface state is applied, instead of immediately. This conforms with the protocol specification now. Signed-off-by: Pekka Paalanen <ppaalanen@gmail.com>
2013-04-25 14:57:42 +04:00
}
/** Set a DPMS mode on all of the compositor's outputs
*
* \param compositor The compositor instance
* \param state The DPMS state the outputs will be set to
*/
static void
weston_compositor_dpms(struct weston_compositor *compositor,
enum dpms_enum state)
{
struct weston_output *output;
wl_list_for_each(output, &compositor->output_list, link)
if (output->set_dpms)
output->set_dpms(output, state);
}
/** Restores the compositor to active status
*
* \param compositor The compositor instance
*
* If the compositor was in a sleeping mode, all outputs are powered
* back on via DPMS. Otherwise if the compositor was inactive
* (idle/locked, offscreen, or sleeping) then the compositor's wake
* signal will fire.
*
* Restarts the idle timer.
*/
WL_EXPORT void
weston_compositor_wake(struct weston_compositor *compositor)
{
uint32_t old_state = compositor->state;
/* The state needs to be changed before emitting the wake
* signal because that may try to schedule a repaint which
* will not work if the compositor is still sleeping */
compositor->state = WESTON_COMPOSITOR_ACTIVE;
switch (old_state) {
case WESTON_COMPOSITOR_SLEEPING:
case WESTON_COMPOSITOR_IDLE:
case WESTON_COMPOSITOR_OFFSCREEN:
weston_compositor_dpms(compositor, WESTON_DPMS_ON);
wl_signal_emit(&compositor->wake_signal, compositor);
/* fall through */
default:
wl_event_source_timer_update(compositor->idle_source,
compositor->idle_time * 1000);
}
}
/** Turns off rendering and frame events for the compositor.
*
* \param compositor The compositor instance
*
* This is used for example to prevent further rendering while the
* compositor is shutting down.
*
* Stops the idle timer.
*/
WL_EXPORT void
weston_compositor_offscreen(struct weston_compositor *compositor)
{
switch (compositor->state) {
case WESTON_COMPOSITOR_OFFSCREEN:
return;
case WESTON_COMPOSITOR_SLEEPING:
default:
compositor->state = WESTON_COMPOSITOR_OFFSCREEN;
wl_event_source_timer_update(compositor->idle_source, 0);
}
}
/** Powers down all attached output devices
*
* \param compositor The compositor instance
*
* Causes rendering to the outputs to cease, and no frame events to be
* sent. Only powers down the outputs if the compositor is not already
* in sleep mode.
*
* Stops the idle timer.
*/
WL_EXPORT void
weston_compositor_sleep(struct weston_compositor *compositor)
{
if (compositor->state == WESTON_COMPOSITOR_SLEEPING)
return;
wl_event_source_timer_update(compositor->idle_source, 0);
compositor->state = WESTON_COMPOSITOR_SLEEPING;
weston_compositor_dpms(compositor, WESTON_DPMS_OFF);
}
/** Sets compositor to idle mode
*
* \param data The compositor instance
*
* This is called when the idle timer fires. Once the compositor is in
* idle mode it requires a wake action (e.g. via
* weston_compositor_wake()) to restore it. The compositor's
* idle_signal will be triggered when the idle event occurs.
*
* Idleness can be inhibited by setting the compositor's idle_inhibit
* property.
*/
static int
idle_handler(void *data)
{
struct weston_compositor *compositor = data;
if (compositor->idle_inhibit)
return 1;
compositor->state = WESTON_COMPOSITOR_IDLE;
wl_signal_emit(&compositor->idle_signal, compositor);
return 1;
}
WL_EXPORT void
weston_plane_init(struct weston_plane *plane,
struct weston_compositor *ec,
int32_t x, int32_t y)
{
pixman_region32_init(&plane->damage);
pixman_region32_init(&plane->clip);
plane->x = x;
plane->y = y;
plane->compositor = ec;
/* Init the link so that the call to wl_list_remove() when releasing
* the plane without ever stacking doesn't lead to a crash */
wl_list_init(&plane->link);
}
WL_EXPORT void
weston_plane_release(struct weston_plane *plane)
{
struct weston_view *view;
pixman_region32_fini(&plane->damage);
pixman_region32_fini(&plane->clip);
wl_list_for_each(view, &plane->compositor->view_list, link) {
if (view->plane == plane)
view->plane = NULL;
}
wl_list_remove(&plane->link);
}
WL_EXPORT void
weston_compositor_stack_plane(struct weston_compositor *ec,
struct weston_plane *plane,
struct weston_plane *above)
{
if (above)
wl_list_insert(above->link.prev, &plane->link);
else
wl_list_insert(&ec->plane_list, &plane->link);
}
static void
output_release(struct wl_client *client, struct wl_resource *resource)
{
wl_resource_destroy(resource);
}
static const struct wl_output_interface output_interface = {
output_release,
};
static void unbind_resource(struct wl_resource *resource)
{
wl_list_remove(wl_resource_get_link(resource));
}
static void
bind_output(struct wl_client *client,
void *data, uint32_t version, uint32_t id)
{
struct weston_output *output = data;
struct weston_mode *mode;
2011-08-29 23:03:09 +04:00
struct wl_resource *resource;
resource = wl_resource_create(client, &wl_output_interface,
version, id);
if (resource == NULL) {
wl_client_post_no_memory(client);
return;
}
wl_list_insert(&output->resource_list, wl_resource_get_link(resource));
wl_resource_set_implementation(resource, &output_interface, data, unbind_resource);
2012-03-05 06:57:37 +04:00
wl_output_send_geometry(resource,
output->x,
output->y,
output->mm_width,
output->mm_height,
output->subpixel,
output->make, output->model,
output->transform);
if (version >= WL_OUTPUT_SCALE_SINCE_VERSION)
wl_output_send_scale(resource,
output->current_scale);
wl_list_for_each (mode, &output->mode_list, link) {
2012-03-05 06:57:37 +04:00
wl_output_send_mode(resource,
mode->flags,
mode->width,
mode->height,
mode->refresh);
}
if (version >= WL_OUTPUT_DONE_SINCE_VERSION)
wl_output_send_done(resource);
}
/* Move other outputs when one is resized so the space remains contiguous. */
static void
weston_compositor_reflow_outputs(struct weston_compositor *compositor,
struct weston_output *resized_output, int delta_width)
{
struct weston_output *output;
bool start_resizing = false;
if (!delta_width)
return;
wl_list_for_each(output, &compositor->output_list, link) {
if (output == resized_output) {
start_resizing = true;
continue;
}
if (start_resizing) {
weston_output_move(output, output->x + delta_width, output->y);
output->dirty = 1;
}
}
}
WL_EXPORT void
weston_output_update_matrix(struct weston_output *output)
{
float magnification;
weston_matrix_init(&output->matrix);
weston_matrix_translate(&output->matrix, -output->x, -output->y, 0);
if (output->zoom.active) {
magnification = 1 / (1 - output->zoom.spring_z.current);
Split the geometry information from weston_surface out into weston_view The weston_surface structure is split into two structures: * The weston_surface structure storres everything required for a client-side or server-side surface. This includes buffers; callbacks; backend private data; input, damage, and opaque regions; and a few other bookkeeping bits. * The weston_view structure represents an entity in the scenegraph and storres all of the geometry information. This includes clip region, alpha, position, and the transformation list as well as all of the temporary information derived from the geometry state. Because a view, and not a surface, is a scenegraph element, the view is what is placed in layers and planes. There are a few things worth noting about the surface/view split: 1. This is *not* a modification to the protocol. It is, instead, a modification to Weston's internal scenegraph to allow a single surface to exist in multiple places at a time. Clients are completely unaware of how many views to a particular surface exist. 2. A view is considered a direct child of a surface and is destroyed when the surface is destroyed. Because of this, the view.surface pointer is always valid and non-null. 3. The compositor's surface_list is replaced with a view_list. Due to subsurfaces, building the view list is a little more complicated than it used to be and involves building a tree of views on the fly whenever subsurfaces are used. However, this means that backends can remain completely subsurface-agnostic. 4. Surfaces and views both keep track of which outputs they are on. 5. The weston_surface structure now has width and height fields. These are populated when a new buffer is attached before surface.configure is called. This is because there are many surface-based operations that really require the width and height and digging through the views didn't work well. Signed-off-by: Jason Ekstrand <jason@jlekstrand.net>
2013-10-13 07:38:11 +04:00
weston_output_update_zoom(output);
weston_matrix_translate(&output->matrix, -output->zoom.trans_x,
-output->zoom.trans_y, 0);
weston_matrix_scale(&output->matrix, magnification,
magnification, 1.0);
}
switch (output->transform) {
case WL_OUTPUT_TRANSFORM_FLIPPED:
case WL_OUTPUT_TRANSFORM_FLIPPED_90:
case WL_OUTPUT_TRANSFORM_FLIPPED_180:
case WL_OUTPUT_TRANSFORM_FLIPPED_270:
weston_matrix_translate(&output->matrix, -output->width, 0, 0);
weston_matrix_scale(&output->matrix, -1, 1, 1);
break;
}
switch (output->transform) {
default:
case WL_OUTPUT_TRANSFORM_NORMAL:
case WL_OUTPUT_TRANSFORM_FLIPPED:
break;
case WL_OUTPUT_TRANSFORM_90:
case WL_OUTPUT_TRANSFORM_FLIPPED_90:
weston_matrix_translate(&output->matrix, 0, -output->height, 0);
weston_matrix_rotate_xy(&output->matrix, 0, 1);
break;
case WL_OUTPUT_TRANSFORM_180:
case WL_OUTPUT_TRANSFORM_FLIPPED_180:
weston_matrix_translate(&output->matrix,
-output->width, -output->height, 0);
weston_matrix_rotate_xy(&output->matrix, -1, 0);
break;
case WL_OUTPUT_TRANSFORM_270:
case WL_OUTPUT_TRANSFORM_FLIPPED_270:
weston_matrix_translate(&output->matrix, -output->width, 0, 0);
weston_matrix_rotate_xy(&output->matrix, 0, -1);
break;
}
if (output->current_scale != 1)
weston_matrix_scale(&output->matrix,
output->current_scale,
output->current_scale, 1);
output->dirty = 0;
weston_matrix_invert(&output->inverse_matrix, &output->matrix);
}
static void
weston_output_transform_scale_init(struct weston_output *output, uint32_t transform, uint32_t scale)
{
output->transform = transform;
output->native_scale = scale;
output->current_scale = scale;
convert_size_by_transform_scale(&output->width, &output->height,
output->current_mode->width,
output->current_mode->height,
transform, scale);
}
static void
weston_output_init_geometry(struct weston_output *output, int x, int y)
{
output->x = x;
output->y = y;
pixman_region32_init(&output->previous_damage);
pixman_region32_init_rect(&output->region, x, y,
output->width,
output->height);
}
WL_EXPORT void
weston_output_move(struct weston_output *output, int x, int y)
{
struct wl_resource *resource;
output->move_x = x - output->x;
output->move_y = y - output->y;
if (output->move_x == 0 && output->move_y == 0)
return;
weston_output_init_geometry(output, x, y);
output->dirty = 1;
/* Move views on this output. */
wl_signal_emit(&output->compositor->output_moved_signal, output);
/* Notify clients of the change for output position. */
wl_resource_for_each(resource, &output->resource_list) {
wl_output_send_geometry(resource,
output->x,
output->y,
output->mm_width,
output->mm_height,
output->subpixel,
output->make,
output->model,
output->transform);
if (wl_resource_get_version(resource) >= WL_OUTPUT_DONE_SINCE_VERSION)
wl_output_send_done(resource);
}
}
/** Adds an output to the compositor's output list and
* send the compositor's output_created signal.
*
* \param compositor The compositor instance.
* \param output The output to be added.
*/
WL_EXPORT void
weston_compositor_add_output(struct weston_compositor *compositor,
struct weston_output *output)
{
struct weston_view *view, *next;
wl_list_insert(compositor->output_list.prev, &output->link);
wl_signal_emit(&compositor->output_created_signal, output);
wl_list_for_each_safe(view, next, &compositor->view_list, link)
weston_view_geometry_dirty(view);
}
WL_EXPORT void
weston_output_transform_coordinate(struct weston_output *output,
double device_x, double device_y,
double *x, double *y)
{
struct weston_vector p = { {
device_x,
device_y,
0.0,
1.0 } };
weston_matrix_transform(&output->inverse_matrix, &p);
*x = p.f[0] / p.f[3];
*y = p.f[1] / p.f[3];
}
libweston: Add more functionality for handling weston_output objects This patch implements additional functionality that will be used for configuring, enabling and disabling weston's outputs. Its indended use is by the compositors or user programs that want to be able to configure, enable or disable an output at any time. An output can only be configured while it's disabled. The compositor and backend specific functionality is required for these functions to be useful, and those will come later in this series. All the new functions have been documented, so I'll avoid describing them here. v2: - Minor documentation improvements. - Rename output-initialized to output->enabled. - Split weston_output_disable() further into weston_compositor_remove_output(). - Rename weston_output_deinit() to weston_output_enable_undo(). - Make weston_output_disable() call two functions mentioned above instead of calling weston_output_disable() directly. This means that backend needs to take care of doing backend specific disable in backend specific destroy function. v3: - Require output->name to be set before calling weston_output_init_pending(). - Require output->destroying to be set before calling weston_compositor_remove_output(). - Split weston_output_init_pending() into weston_compositor_add_pending_output() so pending outputs can be announced separately. - Require output->disable() to be set in order for weston_output_disable() to be usable. - Fix output removing regression that happened when weston_output_disable() was split. - Minor documentation fix. v4: - Bump libweston version to 2 as this patch breaks the ABI. Reviewed-by: Pekka Paalanen <pekka.paalanen@collabora.co.uk> Signed-off-by: Armin Krezović <krezovic.armin@gmail.com>
2016-09-30 15:11:02 +03:00
/** Undoes changes to an output done by weston_output_enable()
*
* \param output The weston_output object that needs the changes undone.
*
* Removes the repaint timer.
* Destroys the Wayland global assigned to the output.
* Destroys pixman regions allocated to the output.
* Deallocates output's ID and updates compositor's output_id_pool.
*/
static void
weston_output_enable_undo(struct weston_output *output)
{
wl_event_source_remove(output->repaint_timer);
wl_global_destroy(output->global);
pixman_region32_fini(&output->region);
pixman_region32_fini(&output->previous_damage);
output->compositor->output_id_pool &= ~(1u << output->id);
output->enabled = false;
}
/** Removes output from compositor's output list
*
* \param output The weston_output object that is being removed.
*
* Presentation feedback is discarded.
* Compositor is notified that outputs were changed and
* applies the necessary changes.
* All views assigned to the weston_output object are
* moved to a new output.
* Signal is emitted to notify all users of the weston_output
libweston: Add more functionality for handling weston_output objects This patch implements additional functionality that will be used for configuring, enabling and disabling weston's outputs. Its indended use is by the compositors or user programs that want to be able to configure, enable or disable an output at any time. An output can only be configured while it's disabled. The compositor and backend specific functionality is required for these functions to be useful, and those will come later in this series. All the new functions have been documented, so I'll avoid describing them here. v2: - Minor documentation improvements. - Rename output-initialized to output->enabled. - Split weston_output_disable() further into weston_compositor_remove_output(). - Rename weston_output_deinit() to weston_output_enable_undo(). - Make weston_output_disable() call two functions mentioned above instead of calling weston_output_disable() directly. This means that backend needs to take care of doing backend specific disable in backend specific destroy function. v3: - Require output->name to be set before calling weston_output_init_pending(). - Require output->destroying to be set before calling weston_compositor_remove_output(). - Split weston_output_init_pending() into weston_compositor_add_pending_output() so pending outputs can be announced separately. - Require output->disable() to be set in order for weston_output_disable() to be usable. - Fix output removing regression that happened when weston_output_disable() was split. - Minor documentation fix. v4: - Bump libweston version to 2 as this patch breaks the ABI. Reviewed-by: Pekka Paalanen <pekka.paalanen@collabora.co.uk> Signed-off-by: Armin Krezović <krezovic.armin@gmail.com>
2016-09-30 15:11:02 +03:00
* object that the output is being destroyed.
* wl_output protocol objects referencing this weston_output
* are made inert.
*/
static void
weston_compositor_remove_output(struct weston_output *output)
{
struct wl_resource *resource;
struct weston_view *view;
assert(output->destroying);
wl_list_for_each(view, &output->compositor->view_list, link) {
if (view->output_mask & (1u << output->id))
weston_view_assign_output(view);
}
weston_presentation_feedback_discard_list(&output->feedback_list);
weston_compositor_reflow_outputs(output->compositor, output, output->width);
wl_list_remove(&output->link);
wl_signal_emit(&output->compositor->output_destroyed_signal, output);
wl_signal_emit(&output->destroy_signal, output);
wl_resource_for_each(resource, &output->resource_list) {
wl_resource_set_destructor(resource, NULL);
}
}
/** Sets the output scale for a given output.
*
* \param output The weston_output object that the scale is set for.
* \param scale Scale factor for the given output.
*
* It only supports setting scale for an output that
* is not enabled and it can only be ran once.
*/
WL_EXPORT void
weston_output_set_scale(struct weston_output *output,
int32_t scale)
{
/* We can only set scale on a disabled output */
assert(!output->enabled);
/* We only want to set scale once */
assert(!output->scale);
output->scale = scale;
}
/** Sets the output transform for a given output.
*
* \param output The weston_output object that the transform is set for.
* \param transform Transform value for the given output.
*
* It only supports setting transform for an output that is
* not enabled and it can only be ran once.
*
* Refer to wl_output::transform section located at
* https://wayland.freedesktop.org/docs/html/apa.html#protocol-spec-wl_output
* for list of values that can be passed to this function.
*/
WL_EXPORT void
weston_output_set_transform(struct weston_output *output,
uint32_t transform)
{
/* We can only set transform on a disabled output */
assert(!output->enabled);
/* We only want to set transform once */
assert(output->transform == UINT32_MAX);
output->transform = transform;
}
/** Initializes a weston_output object with enough data so
** an output can be configured.
*
* \param output The weston_output object to initialize
* \param compositor The compositor instance.
*
* Sets initial values for fields that are expected to be
* configured either by compositors or backends.
*/
WL_EXPORT void
weston_output_init(struct weston_output *output,
struct weston_compositor *compositor)
libweston: Add more functionality for handling weston_output objects This patch implements additional functionality that will be used for configuring, enabling and disabling weston's outputs. Its indended use is by the compositors or user programs that want to be able to configure, enable or disable an output at any time. An output can only be configured while it's disabled. The compositor and backend specific functionality is required for these functions to be useful, and those will come later in this series. All the new functions have been documented, so I'll avoid describing them here. v2: - Minor documentation improvements. - Rename output-initialized to output->enabled. - Split weston_output_disable() further into weston_compositor_remove_output(). - Rename weston_output_deinit() to weston_output_enable_undo(). - Make weston_output_disable() call two functions mentioned above instead of calling weston_output_disable() directly. This means that backend needs to take care of doing backend specific disable in backend specific destroy function. v3: - Require output->name to be set before calling weston_output_init_pending(). - Require output->destroying to be set before calling weston_compositor_remove_output(). - Split weston_output_init_pending() into weston_compositor_add_pending_output() so pending outputs can be announced separately. - Require output->disable() to be set in order for weston_output_disable() to be usable. - Fix output removing regression that happened when weston_output_disable() was split. - Minor documentation fix. v4: - Bump libweston version to 2 as this patch breaks the ABI. Reviewed-by: Pekka Paalanen <pekka.paalanen@collabora.co.uk> Signed-off-by: Armin Krezović <krezovic.armin@gmail.com>
2016-09-30 15:11:02 +03:00
{
output->compositor = compositor;
output->destroying = 0;
/* Backends must set output->name */
assert(output->name);
wl_list_init(&output->link);
output->enabled = false;
/* Add some (in)sane defaults which can be used
* for checking if an output was properly configured
*/
output->mm_width = 0;
output->mm_height = 0;
output->scale = 0;
/* Can't use -1 on uint32_t and 0 is valid enum value */
output->transform = UINT32_MAX;
}
/** Adds weston_output object to pending output list.
*
* \param output The weston_output object to add
* \param compositor The compositor instance.
*
* Also notifies the compositor that an output is pending for
* configuration.
*/
WL_EXPORT void
weston_compositor_add_pending_output(struct weston_output *output,
struct weston_compositor *compositor)
{
wl_list_insert(compositor->pending_output_list.prev, &output->link);
wl_signal_emit(&compositor->output_pending_signal, output);
}
/** Constructs a weston_output object that can be used by the compositor.
*
* \param output The weston_output object that needs to be enabled.
*
* Output coordinates are calculated and each new output is by default
* assigned to the right of previous one.
*
* Sets up the transformation, zoom, and geometry of the output using
* the properties that need to be configured by the compositor.
*
* Establishes a repaint timer for the output with the relevant display
* object's event loop. See output_repaint_timer_handler().
*
* The output is assigned an ID. Weston can support up to 32 distinct
* outputs, with IDs numbered from 0-31; the compositor's output_id_pool
* is referred to and used to find the first available ID number, and
* then this ID is marked as used in output_id_pool.
*
* The output is also assigned a Wayland global with the wl_output
* external interface.
*
* Backend specific function is called to set up the output output.
*
* Output is added to the compositor's output list
*
* If the backend specific function fails, the weston_output object
* is returned to a state it was before calling this function and
* is added to the compositor's pending_output_list in case it needs
* to be reconfigured or just so it can be destroyed at shutdown.
*
* 0 is returned on success, -1 on failure.
*/
WL_EXPORT int
weston_output_enable(struct weston_output *output)
{
struct weston_compositor *c = output->compositor;
libweston: Add more functionality for handling weston_output objects This patch implements additional functionality that will be used for configuring, enabling and disabling weston's outputs. Its indended use is by the compositors or user programs that want to be able to configure, enable or disable an output at any time. An output can only be configured while it's disabled. The compositor and backend specific functionality is required for these functions to be useful, and those will come later in this series. All the new functions have been documented, so I'll avoid describing them here. v2: - Minor documentation improvements. - Rename output-initialized to output->enabled. - Split weston_output_disable() further into weston_compositor_remove_output(). - Rename weston_output_deinit() to weston_output_enable_undo(). - Make weston_output_disable() call two functions mentioned above instead of calling weston_output_disable() directly. This means that backend needs to take care of doing backend specific disable in backend specific destroy function. v3: - Require output->name to be set before calling weston_output_init_pending(). - Require output->destroying to be set before calling weston_compositor_remove_output(). - Split weston_output_init_pending() into weston_compositor_add_pending_output() so pending outputs can be announced separately. - Require output->disable() to be set in order for weston_output_disable() to be usable. - Fix output removing regression that happened when weston_output_disable() was split. - Minor documentation fix. v4: - Bump libweston version to 2 as this patch breaks the ABI. Reviewed-by: Pekka Paalanen <pekka.paalanen@collabora.co.uk> Signed-off-by: Armin Krezović <krezovic.armin@gmail.com>
2016-09-30 15:11:02 +03:00
struct weston_output *iterator;
struct wl_event_loop *loop;
libweston: Add more functionality for handling weston_output objects This patch implements additional functionality that will be used for configuring, enabling and disabling weston's outputs. Its indended use is by the compositors or user programs that want to be able to configure, enable or disable an output at any time. An output can only be configured while it's disabled. The compositor and backend specific functionality is required for these functions to be useful, and those will come later in this series. All the new functions have been documented, so I'll avoid describing them here. v2: - Minor documentation improvements. - Rename output-initialized to output->enabled. - Split weston_output_disable() further into weston_compositor_remove_output(). - Rename weston_output_deinit() to weston_output_enable_undo(). - Make weston_output_disable() call two functions mentioned above instead of calling weston_output_disable() directly. This means that backend needs to take care of doing backend specific disable in backend specific destroy function. v3: - Require output->name to be set before calling weston_output_init_pending(). - Require output->destroying to be set before calling weston_compositor_remove_output(). - Split weston_output_init_pending() into weston_compositor_add_pending_output() so pending outputs can be announced separately. - Require output->disable() to be set in order for weston_output_disable() to be usable. - Fix output removing regression that happened when weston_output_disable() was split. - Minor documentation fix. v4: - Bump libweston version to 2 as this patch breaks the ABI. Reviewed-by: Pekka Paalanen <pekka.paalanen@collabora.co.uk> Signed-off-by: Armin Krezović <krezovic.armin@gmail.com>
2016-09-30 15:11:02 +03:00
int x = 0, y = 0;
assert(output->enable);
iterator = container_of(c->output_list.prev,
libweston: Add more functionality for handling weston_output objects This patch implements additional functionality that will be used for configuring, enabling and disabling weston's outputs. Its indended use is by the compositors or user programs that want to be able to configure, enable or disable an output at any time. An output can only be configured while it's disabled. The compositor and backend specific functionality is required for these functions to be useful, and those will come later in this series. All the new functions have been documented, so I'll avoid describing them here. v2: - Minor documentation improvements. - Rename output-initialized to output->enabled. - Split weston_output_disable() further into weston_compositor_remove_output(). - Rename weston_output_deinit() to weston_output_enable_undo(). - Make weston_output_disable() call two functions mentioned above instead of calling weston_output_disable() directly. This means that backend needs to take care of doing backend specific disable in backend specific destroy function. v3: - Require output->name to be set before calling weston_output_init_pending(). - Require output->destroying to be set before calling weston_compositor_remove_output(). - Split weston_output_init_pending() into weston_compositor_add_pending_output() so pending outputs can be announced separately. - Require output->disable() to be set in order for weston_output_disable() to be usable. - Fix output removing regression that happened when weston_output_disable() was split. - Minor documentation fix. v4: - Bump libweston version to 2 as this patch breaks the ABI. Reviewed-by: Pekka Paalanen <pekka.paalanen@collabora.co.uk> Signed-off-by: Armin Krezović <krezovic.armin@gmail.com>
2016-09-30 15:11:02 +03:00
struct weston_output, link);
if (!wl_list_empty(&c->output_list))
libweston: Add more functionality for handling weston_output objects This patch implements additional functionality that will be used for configuring, enabling and disabling weston's outputs. Its indended use is by the compositors or user programs that want to be able to configure, enable or disable an output at any time. An output can only be configured while it's disabled. The compositor and backend specific functionality is required for these functions to be useful, and those will come later in this series. All the new functions have been documented, so I'll avoid describing them here. v2: - Minor documentation improvements. - Rename output-initialized to output->enabled. - Split weston_output_disable() further into weston_compositor_remove_output(). - Rename weston_output_deinit() to weston_output_enable_undo(). - Make weston_output_disable() call two functions mentioned above instead of calling weston_output_disable() directly. This means that backend needs to take care of doing backend specific disable in backend specific destroy function. v3: - Require output->name to be set before calling weston_output_init_pending(). - Require output->destroying to be set before calling weston_compositor_remove_output(). - Split weston_output_init_pending() into weston_compositor_add_pending_output() so pending outputs can be announced separately. - Require output->disable() to be set in order for weston_output_disable() to be usable. - Fix output removing regression that happened when weston_output_disable() was split. - Minor documentation fix. v4: - Bump libweston version to 2 as this patch breaks the ABI. Reviewed-by: Pekka Paalanen <pekka.paalanen@collabora.co.uk> Signed-off-by: Armin Krezović <krezovic.armin@gmail.com>
2016-09-30 15:11:02 +03:00
x = iterator->x + iterator->width;
/* Make sure the scale is set up */
assert(output->scale);
/* Make sure we have a transform set */
assert(output->transform != UINT32_MAX);
/* Remove it from pending/disabled output list */
wl_list_remove(&output->link);
/* Verify we haven't reached the limit of 32 available output IDs */
assert(ffs(~c->output_id_pool) > 0);
output->x = x;
output->y = y;
output->dirty = 1;
output->original_scale = output->scale;
weston_output_transform_scale_init(output, output->transform, output->scale);
weston_output_init_zoom(output);
weston_output_init_geometry(output, x, y);
weston_output_damage(output);
wl_signal_init(&output->frame_signal);
wl_signal_init(&output->destroy_signal);
wl_list_init(&output->animation_list);
wl_list_init(&output->resource_list);
wl_list_init(&output->feedback_list);
wl_list_init(&output->link);
loop = wl_display_get_event_loop(c->wl_display);
output->repaint_timer = wl_event_loop_add_timer(loop,
output_repaint_timer_handler, output);
/* Invert the output id pool and look for the lowest numbered
* switch (the least significant bit). Take that bit's position
* as our ID, and mark it used in the compositor's output_id_pool.
*/
output->id = ffs(~output->compositor->output_id_pool) - 1;
output->compositor->output_id_pool |= 1u << output->id;
output->global =
wl_global_create(c->wl_display, &wl_output_interface, 3,
output, bind_output);
output->enabled = true;
libweston: Add more functionality for handling weston_output objects This patch implements additional functionality that will be used for configuring, enabling and disabling weston's outputs. Its indended use is by the compositors or user programs that want to be able to configure, enable or disable an output at any time. An output can only be configured while it's disabled. The compositor and backend specific functionality is required for these functions to be useful, and those will come later in this series. All the new functions have been documented, so I'll avoid describing them here. v2: - Minor documentation improvements. - Rename output-initialized to output->enabled. - Split weston_output_disable() further into weston_compositor_remove_output(). - Rename weston_output_deinit() to weston_output_enable_undo(). - Make weston_output_disable() call two functions mentioned above instead of calling weston_output_disable() directly. This means that backend needs to take care of doing backend specific disable in backend specific destroy function. v3: - Require output->name to be set before calling weston_output_init_pending(). - Require output->destroying to be set before calling weston_compositor_remove_output(). - Split weston_output_init_pending() into weston_compositor_add_pending_output() so pending outputs can be announced separately. - Require output->disable() to be set in order for weston_output_disable() to be usable. - Fix output removing regression that happened when weston_output_disable() was split. - Minor documentation fix. v4: - Bump libweston version to 2 as this patch breaks the ABI. Reviewed-by: Pekka Paalanen <pekka.paalanen@collabora.co.uk> Signed-off-by: Armin Krezović <krezovic.armin@gmail.com>
2016-09-30 15:11:02 +03:00
/* Enable the output (set up the crtc or create a
* window representing the output, set up the
* renderer, etc)
*/
if (output->enable(output) < 0) {
weston_log("Enabling output \"%s\" failed.\n", output->name);
weston_output_enable_undo(output);
wl_list_insert(output->compositor->pending_output_list.prev,
&output->link);
return -1;
}
weston_compositor_add_output(output->compositor, output);
return 0;
}
/** Converts a weston_output object to a pending output state, so it
** can be configured again or destroyed.
*
* \param output The weston_output object that needs to be disabled.
*
* See weston_output_init() for more information on the
libweston: Add more functionality for handling weston_output objects This patch implements additional functionality that will be used for configuring, enabling and disabling weston's outputs. Its indended use is by the compositors or user programs that want to be able to configure, enable or disable an output at any time. An output can only be configured while it's disabled. The compositor and backend specific functionality is required for these functions to be useful, and those will come later in this series. All the new functions have been documented, so I'll avoid describing them here. v2: - Minor documentation improvements. - Rename output-initialized to output->enabled. - Split weston_output_disable() further into weston_compositor_remove_output(). - Rename weston_output_deinit() to weston_output_enable_undo(). - Make weston_output_disable() call two functions mentioned above instead of calling weston_output_disable() directly. This means that backend needs to take care of doing backend specific disable in backend specific destroy function. v3: - Require output->name to be set before calling weston_output_init_pending(). - Require output->destroying to be set before calling weston_compositor_remove_output(). - Split weston_output_init_pending() into weston_compositor_add_pending_output() so pending outputs can be announced separately. - Require output->disable() to be set in order for weston_output_disable() to be usable. - Fix output removing regression that happened when weston_output_disable() was split. - Minor documentation fix. v4: - Bump libweston version to 2 as this patch breaks the ABI. Reviewed-by: Pekka Paalanen <pekka.paalanen@collabora.co.uk> Signed-off-by: Armin Krezović <krezovic.armin@gmail.com>
2016-09-30 15:11:02 +03:00
* state output is returned to.
*
* Calls a backend specific function to disable an output, in case
* such function exists.
*
* If the output is being used by the compositor, it is first removed
* from weston's output_list (see weston_compositor_remove_output())
* and is returned to a state it was before weston_output_enable()
* was ran (see weston_output_enable_undo()).
*
* Output is added to pending_output_list so it will get destroyed
* if the output does not get configured again when the compositor
* shuts down. If an output is to be used immediately, it needs to
* be manually removed from the list (the compositor specific functions
* for handling pending outputs will take care of that).
*
* If backend specific disable function returns negative value,
* this function will return too. It can be used as an indicator
* that output cannot be disabled at the present time. In that case
* backend needs to make sure the output is disabled when it is
* possible.
*/
WL_EXPORT void
weston_output_disable(struct weston_output *output)
{
assert(output->disable);
/* Should we rename this? */
output->destroying = 1;
if (output->disable(output) < 0)
return;
if (output->enabled) {
weston_compositor_remove_output(output);
weston_output_enable_undo(output);
/* We need to preserve it somewhere so it can be destroyed on shutdown
if nobody wants to configure it again */
wl_list_insert(output->compositor->pending_output_list.prev, &output->link);
}
output->destroying = 0;
}
/** Emits a signal to indicate that there are outputs waiting to be configured.
*
* \param compositor The compositor instance
*/
WL_EXPORT void
weston_pending_output_coldplug(struct weston_compositor *compositor)
{
struct weston_output *output, *next;
wl_list_for_each_safe(output, next, &compositor->pending_output_list, link)
wl_signal_emit(&compositor->output_pending_signal, output);
}
WL_EXPORT void
weston_output_destroy(struct weston_output *output)
{
output->destroying = 1;
if (output->enabled) {
weston_compositor_remove_output(output);
weston_output_enable_undo(output);
}
free(output->name);
}
static void
destroy_viewport(struct wl_resource *resource)
{
struct weston_surface *surface =
wl_resource_get_user_data(resource);
if (!surface)
return;
surface->viewport_resource = NULL;
surface->pending.buffer_viewport.buffer.src_width =
wl_fixed_from_int(-1);
surface->pending.buffer_viewport.surface.width = -1;
surface->pending.buffer_viewport.changed = 1;
}
static void
viewport_destroy(struct wl_client *client,
struct wl_resource *resource)
{
wl_resource_destroy(resource);
}
static void
viewport_set_source(struct wl_client *client,
struct wl_resource *resource,
wl_fixed_t src_x,
wl_fixed_t src_y,
wl_fixed_t src_width,
wl_fixed_t src_height)
{
struct weston_surface *surface =
wl_resource_get_user_data(resource);
if (!surface) {
wl_resource_post_error(resource,
WP_VIEWPORT_ERROR_NO_SURFACE,
"wl_surface for this viewport is no longer exists");
return;
}
assert(surface->viewport_resource == resource);
assert(surface->resource);
if (src_width == wl_fixed_from_int(-1) &&
src_height == wl_fixed_from_int(-1) &&
src_x == wl_fixed_from_int(-1) &&
src_y == wl_fixed_from_int(-1)) {
/* unset source rect */
surface->pending.buffer_viewport.buffer.src_width =
wl_fixed_from_int(-1);
surface->pending.buffer_viewport.changed = 1;
return;
}
if (src_width <= 0 || src_height <= 0 || src_x < 0 || src_y < 0) {
wl_resource_post_error(resource,
WP_VIEWPORT_ERROR_BAD_VALUE,
"wl_surface@%d viewport source "
"w=%f <= 0, h=%f <= 0, x=%f < 0, or y=%f < 0",
wl_resource_get_id(surface->resource),
wl_fixed_to_double(src_width),
wl_fixed_to_double(src_height),
wl_fixed_to_double(src_x),
wl_fixed_to_double(src_y));
return;
}
surface->pending.buffer_viewport.buffer.src_x = src_x;
surface->pending.buffer_viewport.buffer.src_y = src_y;
surface->pending.buffer_viewport.buffer.src_width = src_width;
surface->pending.buffer_viewport.buffer.src_height = src_height;
surface->pending.buffer_viewport.changed = 1;
}
static void
viewport_set_destination(struct wl_client *client,
struct wl_resource *resource,
int32_t dst_width,
int32_t dst_height)
{
struct weston_surface *surface =
wl_resource_get_user_data(resource);
if (!surface) {
wl_resource_post_error(resource,
WP_VIEWPORT_ERROR_NO_SURFACE,
"wl_surface for this viewport no longer exists");
return;
}
assert(surface->viewport_resource == resource);
if (dst_width == -1 && dst_height == -1) {
/* unset destination size */
surface->pending.buffer_viewport.surface.width = -1;
surface->pending.buffer_viewport.changed = 1;
return;
}
if (dst_width <= 0 || dst_height <= 0) {
wl_resource_post_error(resource,
WP_VIEWPORT_ERROR_BAD_VALUE,
"destination size must be positive (%dx%d)",
dst_width, dst_height);
return;
}
surface->pending.buffer_viewport.surface.width = dst_width;
surface->pending.buffer_viewport.surface.height = dst_height;
surface->pending.buffer_viewport.changed = 1;
}
static const struct wp_viewport_interface viewport_interface = {
viewport_destroy,
viewport_set_source,
viewport_set_destination
};
static void
viewporter_destroy(struct wl_client *client,
struct wl_resource *resource)
{
wl_resource_destroy(resource);
}
static void
viewporter_get_viewport(struct wl_client *client,
struct wl_resource *viewporter,
uint32_t id,
struct wl_resource *surface_resource)
{
int version = wl_resource_get_version(viewporter);
struct weston_surface *surface =
wl_resource_get_user_data(surface_resource);
struct wl_resource *resource;
if (surface->viewport_resource) {
wl_resource_post_error(viewporter,
WP_VIEWPORTER_ERROR_VIEWPORT_EXISTS,
"a viewport for that surface already exists");
return;
}
resource = wl_resource_create(client, &wp_viewport_interface,
version, id);
if (resource == NULL) {
wl_client_post_no_memory(client);
return;
}
wl_resource_set_implementation(resource, &viewport_interface,
surface, destroy_viewport);
surface->viewport_resource = resource;
}
static const struct wp_viewporter_interface viewporter_interface = {
viewporter_destroy,
viewporter_get_viewport
};
static void
bind_viewporter(struct wl_client *client,
void *data, uint32_t version, uint32_t id)
{
struct wl_resource *resource;
resource = wl_resource_create(client, &wp_viewporter_interface,
version, id);
if (resource == NULL) {
wl_client_post_no_memory(client);
return;
}
wl_resource_set_implementation(resource, &viewporter_interface,
NULL, NULL);
}
static void
destroy_presentation_feedback(struct wl_resource *feedback_resource)
{
struct weston_presentation_feedback *feedback;
feedback = wl_resource_get_user_data(feedback_resource);
wl_list_remove(&feedback->link);
free(feedback);
}
static void
presentation_destroy(struct wl_client *client, struct wl_resource *resource)
{
wl_resource_destroy(resource);
}
static void
presentation_feedback(struct wl_client *client,
struct wl_resource *presentation_resource,
struct wl_resource *surface_resource,
uint32_t callback)
{
struct weston_surface *surface;
struct weston_presentation_feedback *feedback;
surface = wl_resource_get_user_data(surface_resource);
feedback = zalloc(sizeof *feedback);
if (feedback == NULL)
goto err_calloc;
feedback->resource = wl_resource_create(client,
&wp_presentation_feedback_interface,
1, callback);
if (!feedback->resource)
goto err_create;
wl_resource_set_implementation(feedback->resource, NULL, feedback,
destroy_presentation_feedback);
wl_list_insert(&surface->pending.feedback_list, &feedback->link);
return;
err_create:
free(feedback);
err_calloc:
wl_client_post_no_memory(client);
}
static const struct wp_presentation_interface presentation_implementation = {
presentation_destroy,
presentation_feedback
};
static void
bind_presentation(struct wl_client *client,
void *data, uint32_t version, uint32_t id)
{
struct weston_compositor *compositor = data;
struct wl_resource *resource;
resource = wl_resource_create(client, &wp_presentation_interface,
version, id);
if (resource == NULL) {
wl_client_post_no_memory(client);
return;
}
wl_resource_set_implementation(resource, &presentation_implementation,
compositor, NULL);
wp_presentation_send_clock_id(resource, compositor->presentation_clock);
}
static void
compositor_bind(struct wl_client *client,
void *data, uint32_t version, uint32_t id)
{
struct weston_compositor *compositor = data;
struct wl_resource *resource;
resource = wl_resource_create(client, &wl_compositor_interface,
version, id);
if (resource == NULL) {
wl_client_post_no_memory(client);
return;
}
wl_resource_set_implementation(resource, &compositor_interface,
compositor, NULL);
}
WL_EXPORT int
weston_environment_get_fd(const char *env)
{
char *e;
int fd, flags;
e = getenv(env);
if (!e || !safe_strtoint(e, &fd))
return -1;
flags = fcntl(fd, F_GETFD);
if (flags == -1)
return -1;
fcntl(fd, F_SETFD, flags | FD_CLOEXEC);
unsetenv(env);
return fd;
}
compositor: Implement JSON-timeline logging Logging is activated and deactivated with the debug key binding 't'. When activated, it creates a new log file, where it records the events. The log file contains events and detailed object information entries in JSON format, and is meant to be parsed in sequence from beginning to the end. The emitted events are mostly related to the output repaint cycle, like when repaint begins, is submitted to GPU, and when it completes on a vblank. This is recorded per-output. Also some per-surface events are recorded, including when surface damage is flushed. To reduce the log size, events refer to objects like outputs and surfaces by id numbers. Detailed object information is emitted only as needed: on the first object occurrence, and afterwards only if weston_timeline_object::force_refresh asks for it. The detailed information for surfaces includes the string returned by weston_surface::get_label. Therefore it is important to set weston_timeline_object::force_refresh = 1 whenever the string would change, so that the new details get recorded. A rudimentary parser and SVG generator can be found at: https://github.com/ppaalanen/wesgr The timeline logs can answer questions including: - How does the compositor repaint cycle work timing-wise? - When was the vblank deadline missed? - What is the latency from surface commit to showing the new content on screen? - How long does it take to process the scenegraph? v2: weston_surface::get_description renamed to get_label. v3: reafctor a bit into fprint_quoted_string(). Signed-off-by: Pekka Paalanen <pekka.paalanen@collabora.co.uk>
2014-11-12 16:09:24 +03:00
static void
timeline_key_binding_handler(struct weston_keyboard *keyboard, uint32_t time,
compositor: Implement JSON-timeline logging Logging is activated and deactivated with the debug key binding 't'. When activated, it creates a new log file, where it records the events. The log file contains events and detailed object information entries in JSON format, and is meant to be parsed in sequence from beginning to the end. The emitted events are mostly related to the output repaint cycle, like when repaint begins, is submitted to GPU, and when it completes on a vblank. This is recorded per-output. Also some per-surface events are recorded, including when surface damage is flushed. To reduce the log size, events refer to objects like outputs and surfaces by id numbers. Detailed object information is emitted only as needed: on the first object occurrence, and afterwards only if weston_timeline_object::force_refresh asks for it. The detailed information for surfaces includes the string returned by weston_surface::get_label. Therefore it is important to set weston_timeline_object::force_refresh = 1 whenever the string would change, so that the new details get recorded. A rudimentary parser and SVG generator can be found at: https://github.com/ppaalanen/wesgr The timeline logs can answer questions including: - How does the compositor repaint cycle work timing-wise? - When was the vblank deadline missed? - What is the latency from surface commit to showing the new content on screen? - How long does it take to process the scenegraph? v2: weston_surface::get_description renamed to get_label. v3: reafctor a bit into fprint_quoted_string(). Signed-off-by: Pekka Paalanen <pekka.paalanen@collabora.co.uk>
2014-11-12 16:09:24 +03:00
uint32_t key, void *data)
{
struct weston_compositor *compositor = data;
if (weston_timeline_enabled_)
weston_timeline_close();
else
weston_timeline_open(compositor);
}
/** Create the compositor.
*
* This functions creates and initializes a compositor instance.
*
* \param display The Wayland display to be used.
* \param user_data A pointer to an object that can later be retrieved
* using the \ref weston_compositor_get_user_data function.
* \return The compositor instance on success or NULL on failure.
*/
WL_EXPORT struct weston_compositor *
weston_compositor_create(struct wl_display *display, void *user_data)
{
struct weston_compositor *ec;
struct wl_event_loop *loop;
ec = zalloc(sizeof *ec);
if (!ec)
return NULL;
ec->wl_display = display;
ec->user_data = user_data;
wl_signal_init(&ec->destroy_signal);
wl_signal_init(&ec->create_surface_signal);
wl_signal_init(&ec->activate_signal);
wl_signal_init(&ec->transform_signal);
wl_signal_init(&ec->kill_signal);
wl_signal_init(&ec->idle_signal);
wl_signal_init(&ec->wake_signal);
2012-06-21 23:52:17 +04:00
wl_signal_init(&ec->show_input_panel_signal);
wl_signal_init(&ec->hide_input_panel_signal);
wl_signal_init(&ec->update_input_panel_signal);
wl_signal_init(&ec->seat_created_signal);
libweston: Add more functionality for handling weston_output objects This patch implements additional functionality that will be used for configuring, enabling and disabling weston's outputs. Its indended use is by the compositors or user programs that want to be able to configure, enable or disable an output at any time. An output can only be configured while it's disabled. The compositor and backend specific functionality is required for these functions to be useful, and those will come later in this series. All the new functions have been documented, so I'll avoid describing them here. v2: - Minor documentation improvements. - Rename output-initialized to output->enabled. - Split weston_output_disable() further into weston_compositor_remove_output(). - Rename weston_output_deinit() to weston_output_enable_undo(). - Make weston_output_disable() call two functions mentioned above instead of calling weston_output_disable() directly. This means that backend needs to take care of doing backend specific disable in backend specific destroy function. v3: - Require output->name to be set before calling weston_output_init_pending(). - Require output->destroying to be set before calling weston_compositor_remove_output(). - Split weston_output_init_pending() into weston_compositor_add_pending_output() so pending outputs can be announced separately. - Require output->disable() to be set in order for weston_output_disable() to be usable. - Fix output removing regression that happened when weston_output_disable() was split. - Minor documentation fix. v4: - Bump libweston version to 2 as this patch breaks the ABI. Reviewed-by: Pekka Paalanen <pekka.paalanen@collabora.co.uk> Signed-off-by: Armin Krezović <krezovic.armin@gmail.com>
2016-09-30 15:11:02 +03:00
wl_signal_init(&ec->output_pending_signal);
wl_signal_init(&ec->output_created_signal);
wl_signal_init(&ec->output_destroyed_signal);
wl_signal_init(&ec->output_moved_signal);
wl_signal_init(&ec->output_resized_signal);
wl_signal_init(&ec->session_signal);
ec->session_active = 1;
ec->output_id_pool = 0;
ec->repaint_msec = DEFAULT_REPAINT_WINDOW;
ec->activate_serial = 1;
if (!wl_global_create(ec->wl_display, &wl_compositor_interface, 4,
2013-07-09 03:03:57 +04:00
ec, compositor_bind))
goto fail;
if (!wl_global_create(ec->wl_display, &wl_subcompositor_interface, 1,
2013-07-09 03:03:57 +04:00
ec, bind_subcompositor))
goto fail;
compositor: introduce sub-surfaces Implement the basic protocol for sub-surfaces: - expose wl_subcompositor global interface - error checking on protocol calls - associate a parent wl_surface to a sub-surface - introduce the sub-surface role, which is exclusive - an implementation of the wl_subsurface interface - allow nesting of sub-surfaces - proper surface transformation inheritance from parent to sub-surfaces - two different modes of wl_surface.commit for sub-surfaces - hook sub-surfaces up to repaint by modifying the repaint list code Struct weston_subsurface is dynamically allocated. For sub-surfaces, it is completely populated. For parent surfaces, weston_subsurface acts only as a link for stacking order purposes. The wl_resource is unused, parent_destroy_listener is not registered, the transform is not linked, etc. Sub-surfaces are not added directly into layers for display or input. Instead, they are hooked up via the sub-surface list present in parent weston_surface. This way sub-surfaces are inherently linked to the parent surface, and cannot be displayed unless the parent is mapped, too. This also eases restacking, as only the parent will be in a layer list. Also, only the main surface should be subject to shell actions. The surface list rebuilding in weston_output_repaint() is modified to process sub-surface lists, if they are non-empty. The sub-surface list always contains the parent, too, unless empty. The collection of frame_callback_list is moved to a later loop, to streamline the surface list rebuild functions. Features still lacking are: - full-surface alpha support for compound windows Changes in v2: - fix a bug in surface mapping: commit a sub-surface would cause the main surface to never be mapped. - remove debug printfs - detect attempt of making a surface its own parent - always zero-alloc weston_subsurface - apply wl_subsurface.set_position in commit, not immediately - add weston_surface_to_subsurface() - implement sub-surface commit modes parent-cached and independent - implement wl_subcompositor.destroy and wl_subsurface.destroy Changes in v3: - rebased, and use the new transform inheritance code - squashed the commit "add sub-surfaces to repaint list" - fixed a buffer reference leak in commit_from_cache() - Rewrite the sub-surface destructor code, and make it leave the wl_subsurface protocol object inert, if one destroys the corresponding wl_surface. - replaced set_commit_mode with set_sync and set_desync - allowed sub-surface nesting, and fixed repaint accordingly - implemented nested sub-surface commit modes - Made the sub-surface order changes from wl_subsurface.place_above and .place_below to be applied when the parent surface state is applied, instead of immediately. This conforms with the protocol specification now. Signed-off-by: Pekka Paalanen <ppaalanen@gmail.com>
2013-04-25 14:57:42 +04:00
if (!wl_global_create(ec->wl_display, &wp_viewporter_interface, 1,
ec, bind_viewporter))
goto fail;
if (!wl_global_create(ec->wl_display, &wp_presentation_interface, 1,
ec, bind_presentation))
goto fail;
if (weston_input_init(ec) != 0)
goto fail;
Split the geometry information from weston_surface out into weston_view The weston_surface structure is split into two structures: * The weston_surface structure storres everything required for a client-side or server-side surface. This includes buffers; callbacks; backend private data; input, damage, and opaque regions; and a few other bookkeeping bits. * The weston_view structure represents an entity in the scenegraph and storres all of the geometry information. This includes clip region, alpha, position, and the transformation list as well as all of the temporary information derived from the geometry state. Because a view, and not a surface, is a scenegraph element, the view is what is placed in layers and planes. There are a few things worth noting about the surface/view split: 1. This is *not* a modification to the protocol. It is, instead, a modification to Weston's internal scenegraph to allow a single surface to exist in multiple places at a time. Clients are completely unaware of how many views to a particular surface exist. 2. A view is considered a direct child of a surface and is destroyed when the surface is destroyed. Because of this, the view.surface pointer is always valid and non-null. 3. The compositor's surface_list is replaced with a view_list. Due to subsurfaces, building the view list is a little more complicated than it used to be and involves building a tree of views on the fly whenever subsurfaces are used. However, this means that backends can remain completely subsurface-agnostic. 4. Surfaces and views both keep track of which outputs they are on. 5. The weston_surface structure now has width and height fields. These are populated when a new buffer is attached before surface.configure is called. This is because there are many surface-based operations that really require the width and height and digging through the views didn't work well. Signed-off-by: Jason Ekstrand <jason@jlekstrand.net>
2013-10-13 07:38:11 +04:00
wl_list_init(&ec->view_list);
wl_list_init(&ec->plane_list);
wl_list_init(&ec->layer_list);
wl_list_init(&ec->seat_list);
libweston: Add more functionality for handling weston_output objects This patch implements additional functionality that will be used for configuring, enabling and disabling weston's outputs. Its indended use is by the compositors or user programs that want to be able to configure, enable or disable an output at any time. An output can only be configured while it's disabled. The compositor and backend specific functionality is required for these functions to be useful, and those will come later in this series. All the new functions have been documented, so I'll avoid describing them here. v2: - Minor documentation improvements. - Rename output-initialized to output->enabled. - Split weston_output_disable() further into weston_compositor_remove_output(). - Rename weston_output_deinit() to weston_output_enable_undo(). - Make weston_output_disable() call two functions mentioned above instead of calling weston_output_disable() directly. This means that backend needs to take care of doing backend specific disable in backend specific destroy function. v3: - Require output->name to be set before calling weston_output_init_pending(). - Require output->destroying to be set before calling weston_compositor_remove_output(). - Split weston_output_init_pending() into weston_compositor_add_pending_output() so pending outputs can be announced separately. - Require output->disable() to be set in order for weston_output_disable() to be usable. - Fix output removing regression that happened when weston_output_disable() was split. - Minor documentation fix. v4: - Bump libweston version to 2 as this patch breaks the ABI. Reviewed-by: Pekka Paalanen <pekka.paalanen@collabora.co.uk> Signed-off-by: Armin Krezović <krezovic.armin@gmail.com>
2016-09-30 15:11:02 +03:00
wl_list_init(&ec->pending_output_list);
wl_list_init(&ec->output_list);
wl_list_init(&ec->key_binding_list);
wl_list_init(&ec->modifier_binding_list);
wl_list_init(&ec->button_binding_list);
wl_list_init(&ec->touch_binding_list);
wl_list_init(&ec->axis_binding_list);
wl_list_init(&ec->debug_binding_list);
wl_list_init(&ec->plugin_api_list);
weston_plane_init(&ec->primary_plane, ec, 0, 0);
weston_compositor_stack_plane(ec, &ec->primary_plane, NULL);
wl_data_device_manager_init(ec->wl_display);
wl_display_init_shm(ec->wl_display);
loop = wl_display_get_event_loop(ec->wl_display);
ec->idle_source = wl_event_loop_add_timer(loop, idle_handler, ec);
weston_layer_init(&ec->fade_layer, ec);
weston_layer_init(&ec->cursor_layer, ec);
weston_layer_set_position(&ec->fade_layer, WESTON_LAYER_POSITION_FADE);
weston_layer_set_position(&ec->cursor_layer,
WESTON_LAYER_POSITION_CURSOR);
weston_compositor_add_debug_binding(ec, KEY_T,
timeline_key_binding_handler, ec);
return ec;
fail:
free(ec);
return NULL;
}
WL_EXPORT void
weston_compositor_shutdown(struct weston_compositor *ec)
{
struct weston_output *output, *next;
wl_event_source_remove(ec->idle_source);
/* Destroy all outputs associated with this compositor */
wl_list_for_each_safe(output, next, &ec->output_list, link)
output->destroy(output);
libweston: Add more functionality for handling weston_output objects This patch implements additional functionality that will be used for configuring, enabling and disabling weston's outputs. Its indended use is by the compositors or user programs that want to be able to configure, enable or disable an output at any time. An output can only be configured while it's disabled. The compositor and backend specific functionality is required for these functions to be useful, and those will come later in this series. All the new functions have been documented, so I'll avoid describing them here. v2: - Minor documentation improvements. - Rename output-initialized to output->enabled. - Split weston_output_disable() further into weston_compositor_remove_output(). - Rename weston_output_deinit() to weston_output_enable_undo(). - Make weston_output_disable() call two functions mentioned above instead of calling weston_output_disable() directly. This means that backend needs to take care of doing backend specific disable in backend specific destroy function. v3: - Require output->name to be set before calling weston_output_init_pending(). - Require output->destroying to be set before calling weston_compositor_remove_output(). - Split weston_output_init_pending() into weston_compositor_add_pending_output() so pending outputs can be announced separately. - Require output->disable() to be set in order for weston_output_disable() to be usable. - Fix output removing regression that happened when weston_output_disable() was split. - Minor documentation fix. v4: - Bump libweston version to 2 as this patch breaks the ABI. Reviewed-by: Pekka Paalanen <pekka.paalanen@collabora.co.uk> Signed-off-by: Armin Krezović <krezovic.armin@gmail.com>
2016-09-30 15:11:02 +03:00
/* Destroy all pending outputs associated with this compositor */
wl_list_for_each_safe(output, next, &ec->pending_output_list, link)
output->destroy(output);
if (ec->renderer)
ec->renderer->destroy(ec);
weston_binding_list_destroy_all(&ec->key_binding_list);
weston_binding_list_destroy_all(&ec->modifier_binding_list);
weston_binding_list_destroy_all(&ec->button_binding_list);
weston_binding_list_destroy_all(&ec->touch_binding_list);
weston_binding_list_destroy_all(&ec->axis_binding_list);
weston_binding_list_destroy_all(&ec->debug_binding_list);
weston_plane_release(&ec->primary_plane);
}
WL_EXPORT void
weston_compositor_exit_with_code(struct weston_compositor *compositor,
int exit_code)
{
if (compositor->exit_code == EXIT_SUCCESS)
compositor->exit_code = exit_code;
weston_compositor_exit(compositor);
}
WL_EXPORT void
weston_compositor_set_default_pointer_grab(struct weston_compositor *ec,
const struct weston_pointer_grab_interface *interface)
{
struct weston_seat *seat;
ec->default_pointer_grab = interface;
wl_list_for_each(seat, &ec->seat_list, link) {
struct weston_pointer *pointer = weston_seat_get_pointer(seat);
if (pointer)
weston_pointer_set_default_grab(pointer, interface);
}
}
WL_EXPORT int
weston_compositor_set_presentation_clock(struct weston_compositor *compositor,
clockid_t clk_id)
{
struct timespec ts;
if (clock_gettime(clk_id, &ts) < 0)
return -1;
compositor->presentation_clock = clk_id;
return 0;
}
/*
* For choosing the software clock, when the display hardware or API
* does not expose a compatible presentation timestamp.
*/
WL_EXPORT int
weston_compositor_set_presentation_clock_software(
struct weston_compositor *compositor)
{
/* In order of preference */
static const clockid_t clocks[] = {
CLOCK_MONOTONIC_RAW, /* no jumps, no crawling */
CLOCK_MONOTONIC_COARSE, /* no jumps, may crawl, fast & coarse */
CLOCK_MONOTONIC, /* no jumps, may crawl */
CLOCK_REALTIME_COARSE, /* may jump and crawl, fast & coarse */
CLOCK_REALTIME /* may jump and crawl */
};
unsigned i;
for (i = 0; i < ARRAY_LENGTH(clocks); i++)
if (weston_compositor_set_presentation_clock(compositor,
clocks[i]) == 0)
return 0;
weston_log("Error: no suitable presentation clock available.\n");
return -1;
}
/** Read the current time from the Presentation clock
*
* \param compositor
* \param ts[out] The current time.
*
* \note Reading the current time in user space is always imprecise to some
* degree.
*
* This function is never meant to fail. If reading the clock does fail,
* an error message is logged and a zero time is returned. Callers are not
* supposed to detect or react to failures.
*/
WL_EXPORT void
weston_compositor_read_presentation_clock(
const struct weston_compositor *compositor,
struct timespec *ts)
{
static bool warned;
int ret;
ret = clock_gettime(compositor->presentation_clock, ts);
if (ret < 0) {
ts->tv_sec = 0;
ts->tv_nsec = 0;
if (!warned)
weston_log("Error: failure to read "
"the presentation clock %#x: '%m' (%d)\n",
compositor->presentation_clock, errno);
warned = true;
}
}
/** Import dmabuf buffer into current renderer
*
* \param compositor
* \param buffer the dmabuf buffer to import
* \return true on usable buffers, false otherwise
*
* This function tests that the linux_dmabuf_buffer is usable
* for the current renderer. Returns false on unusable buffers. Usually
* usability is tested by importing the dmabufs for composition.
*
* This hook is also used for detecting if the renderer supports
* dmabufs at all. If the renderer hook is NULL, dmabufs are not
* supported.
* */
WL_EXPORT bool
weston_compositor_import_dmabuf(struct weston_compositor *compositor,
struct linux_dmabuf_buffer *buffer)
{
struct weston_renderer *renderer;
renderer = compositor->renderer;
if (renderer->import_dmabuf == NULL)
return false;
return renderer->import_dmabuf(compositor, buffer);
}
WL_EXPORT void
weston_version(int *major, int *minor, int *micro)
{
*major = WESTON_VERSION_MAJOR;
*minor = WESTON_VERSION_MINOR;
*micro = WESTON_VERSION_MICRO;
}
WL_EXPORT void *
weston_load_module(const char *name, const char *entrypoint)
{
const char *builddir = getenv("WESTON_BUILD_DIR");
char path[PATH_MAX];
void *module, *init;
size_t len;
if (name == NULL)
return NULL;
if (name[0] != '/') {
if (builddir)
len = snprintf(path, sizeof path, "%s/.libs/%s",
builddir, name);
else
len = snprintf(path, sizeof path, "%s/%s",
LIBWESTON_MODULEDIR, name);
} else {
len = snprintf(path, sizeof path, "%s", name);
}
/* snprintf returns the length of the string it would've written,
* _excluding_ the NUL byte. So even being equal to the size of
* our buffer is an error here. */
if (len >= sizeof path)
return NULL;
module = dlopen(path, RTLD_NOW | RTLD_NOLOAD);
if (module) {
weston_log("Module '%s' already loaded\n", path);
dlclose(module);
return NULL;
}
weston_log("Loading module '%s'\n", path);
module = dlopen(path, RTLD_NOW);
if (!module) {
weston_log("Failed to load module: %s\n", dlerror());
return NULL;
}
init = dlsym(module, entrypoint);
if (!init) {
weston_log("Failed to lookup init function: %s\n", dlerror());
dlclose(module);
return NULL;
}
return init;
}
/** Destroys the compositor.
*
* This function cleans up the compositor state and destroys it.
*
* \param compositor The compositor to be destroyed.
*/
WL_EXPORT void
weston_compositor_destroy(struct weston_compositor *compositor)
{
/* prevent further rendering while shutting down */
compositor->state = WESTON_COMPOSITOR_OFFSCREEN;
wl_signal_emit(&compositor->destroy_signal, compositor);
weston_compositor_xkb_destroy(compositor);
if (compositor->backend)
compositor->backend->destroy(compositor);
weston_plugin_api_destroy_list(compositor);
free(compositor);
}
/** Instruct the compositor to exit.
*
* This functions does not directly destroy the compositor object, it merely
* command it to start the tear down process. It is not guaranteed that the
* tear down will happen immediately.
*
* \param compositor The compositor to tear down.
*/
WL_EXPORT void
weston_compositor_exit(struct weston_compositor *compositor)
{
compositor->exit(compositor);
}
/** Return the user data stored in the compositor.
*
* This function returns the user data pointer set with user_data parameter
* to the \ref weston_compositor_create function.
*/
WL_EXPORT void *
weston_compositor_get_user_data(struct weston_compositor *compositor)
{
return compositor->user_data;
}
static const char * const backend_map[] = {
[WESTON_BACKEND_DRM] = "drm-backend.so",
[WESTON_BACKEND_FBDEV] = "fbdev-backend.so",
[WESTON_BACKEND_HEADLESS] = "headless-backend.so",
[WESTON_BACKEND_RDP] = "rdp-backend.so",
[WESTON_BACKEND_WAYLAND] = "wayland-backend.so",
[WESTON_BACKEND_X11] = "x11-backend.so",
};
/** Load a backend into a weston_compositor
*
* A backend must be loaded to make a weston_compositor work. A backend
* provides input and output capabilities, and determines the renderer to use.
*
* \param compositor A compositor that has not had a backend loaded yet.
* \param backend Name of the backend file.
* \param config_base A pointer to a backend-specific configuration
* structure's 'base' member.
*
* \return 0 on success, or -1 on error.
*/
WL_EXPORT int
weston_compositor_load_backend(struct weston_compositor *compositor,
enum weston_compositor_backend backend,
struct weston_backend_config *config_base)
{
int (*backend_init)(struct weston_compositor *c,
struct weston_backend_config *config_base);
if (backend >= ARRAY_LENGTH(backend_map))
return -1;
backend_init = weston_load_module(backend_map[backend], "weston_backend_init");
if (!backend_init)
return -1;
return backend_init(compositor, config_base);
}
WL_EXPORT int
weston_compositor_load_xwayland(struct weston_compositor *compositor)
{
int (*module_init)(struct weston_compositor *ec);
module_init = weston_load_module("xwayland.so", "weston_module_init");
if (!module_init)
return -1;
if (module_init(compositor) < 0)
return -1;
return 0;
}