- Added ping_timer_destroy() to simplify cleanup.
- Changed timeout and fade step to more realistic values.
- Renamed ping_timeout_fade_frame() to unresponsive_fade_frame().
If a client is not responding, lower the brightness and
saturation to indicate it's stalled. The surface is restored
to it's original color values if the client later becomes
responsive.
Switching display mode may happen when:
1. The fullscreen surface is at top most in fullscreen layer and with
"driver" method. Shell will switch output mode to match the surface
size. If no matched mode found, fall back to "fill" method.
2. The top fullscreen surface is destroyed or unset. Switch back to the
origin mode.
Specially the "either" word there is essential, so users don't confuse
thinking that both steps are needed.
Signed-off-by: Tiago Vignatti <tiago.vignatti@intel.com>
Some GL implementations do not provide GL_EXT_read_format_bgra
extension.
Set a glReadPixels format based on whether the extensions is supported
or not, and use that format in all backends.
Add RGBA->BGRA swapping copy to screenshooter to keep the shm buffer
data format as BGRA.
Signed-off-by: Pekka Paalanen <ppaalanen@gmail.com>
During a bring-up of a new backend, it would be nice to get a real error
message, when the EGL and GL contexts have not been properly set up.
Signed-off-by: Pekka Paalanen <ppaalanen@gmail.com>
This is a workaround for platforms, whose EGL headers miss the
definitions for EGL_WL_bind_wayland_display.
Signed-off-by: Pekka Paalanen <ppaalanen@gmail.com>
Compositor core does not do anything with udev, so the header is not
needed there. Move the #include into evdev.h, from where it gets used by
compositor-drm.c, too.
Also fix the fallout:
tty.c: In function 'tty_create':
tty.c:143:2: warning: implicit declaration of function 'fstat'
Signed-off-by: Pekka Paalanen <ppaalanen@gmail.com>
This reverts a snip from commit d012e9df. On that commit, it was lost the
ability of calling X applications from desktop panel; xserver module
was setting DISPLAY only later, after panel was already launched.
Signed-off-by: Tiago Vignatti <tiago.vignatti@intel.com>
wl_shell is more likely to be used for core protocol specific. Now it follows
pretty much the same style of what tablet-shell is using.
No functional changes.
Signed-off-by: Tiago Vignatti <tiago.vignatti@intel.com>
When the surface being moved, resized or rotated was destroyed, the
compositor would crash.
Fix this by using a destroy listener on the referenced surface. To
reduce code duplication, the surface reference and the destroy
listener is added to a new struct shell_grab.
Removing the input devices will trigger events and activity that will cause
a redraw. Do this before we put the compositor to sleep so we can cancel it
properly and avoid immediately waking the compositor again.
We dont want to receive hotplug events while being inactive.
When getting back active we enumerate all devices and would end up
with two sources for one device that may be hotplugged in the mean time.
weston-launch starts weston and provides mechanism
for weston to set/drop drm master, open a tty,
and read input devices without being root.
Execution is allowed for local-active sessions
or users in the group weston-launch.
Use shell_stack_fullscreen() to raise fullscreen surface in activate(),
and before activate() a regular surface, move all the fullscreen
surfaces from fullscreen_layer to toplevel_layer.
Also add a void *private into struct wesont_surface since we already
have a configure() vfunc. That helps to get the associated fullscreen
surface of black surface.
The remaining use case was making our context current before we had any
output surfaces. We can do that now using a dummy surface, so let's stop
relying on surfaceless.
Instead of using the hacks of gbm_bos, EGLImage, FBOs and surfaceless
we switch to using the new gbm surface API. This gives us an EGL
native window type for gbm and lets us use a real EGL surface.
A couple of fixes were made: Weston tablet-shell needed to use weston_layer,
so the compositor could rebuild the surface list correctly when repainting;
homescreen and locking are using the widget + window abstration of toytoolkit;
and widget_set_redraw_handler are being set for widgets redraw.
Also, it was given some basic meaning for lockscreen_button_handler, which
was completely disabled before. As a clean up, I updated the global listener
mechanism on tablet-shell client, using the regular way of registering a
handler instead wl_display_roundtrip -> wl_display_get_global.
Switcher still without code to proper work and the same for tablet-shell
clients, which are not launched.
krh: Edited to not scale down homescreen icons, use new load_cairo_surface()
for image loading.
Signed-off-by: Tiago Vignatti <tiago.vignatti@intel.com>
Just like device_added, now the routines to close the compositor and vt switch
leave are using the same code to remove a device.
This patch also closes properly a mtdev device, bug spotted by Christopher
Michael.
Signed-off-by: Tiago Vignatti <tiago.vignatti@intel.com>
This allows us to move the logic that calls shell->map() or
shell->configure() into shell while allowing it to be overriden for
surfaces that should not be handle by the shell, such as drag icons.
This patch adds a pointer function called configure to weston_surface,
moves the currsent surface_configure() code into shell and implements
a separate configure() for drag surfaces.
surface_attach() does two things: sets up a new buffer as the contents
of the given surface and then calls into shell so it can setup the
position of the surface and map it if necessary. However we do not want
the shell to meddle with some internal surfaces such as drag surfaces.
The intention of this refactoring is to make room for making the part
that calls into shell a virtual function that the compositor can
override for these internal surfaces.
This changes weston_buffer_attach() so it handle all the logic of tying
a buffer to a surface, including unmapping it when the buffer is NULL.
The shell map() vs. configure() logic is then split into a another
function: surface_configure(). In a later commit, this function will be
turned into a function pointer in struct weston_surface.
For shm buffers, es->pitch is set using the stride of the buffer. If
the shell happened to set the surface width to something different than
the buffer width, the contents of the surface would be cropped on the
width during redraw. However, for non-shm surfaces, es->pitch was set to
the surface width. That caused the contents of the buffer to be scaled
on the width when the buffer was wider than the surface.
This makes the behavior on both cases the same: crop on the width and
scale on the height. (which is weird but consistent)
There was a lot of code here to do a lot of work we didn't need to do.
If we damage a surface with a shm buffer attached, all we need to do
is to re-upload the damaged region to the texture. As for drm buffers,
we don't assume anything changes on attach and only update the
regions the client tells us to update in the damage request.
While activating and deactivating rotation mechanism without moving
the pointer, rotation matrix from rotate_grab object is not being
initialised and damage shell surface rotation matrix in
rotate_grab_button handler, making it invertible.
This patch initialise rotate matrix in rotate_binding
and moves surface position check to rotate_grab_motion handler.
How to reproduce: rotate the surface to something like 45 degrees,
resize it drastically, continue to rotate. The surface will jump
some space and the rotation point will not be in the center
of the surface.
Fix is to shift the surface position to match the rotation point
This reverts commit e7ad5cdcd2.
If you ask for setuid install and that fails you didn't get what you
asked for and we shouldn't just silently carry on. If installing weston
somewhere in your home directory and don't want the setuid bit set,
disable that at configure time.
When the compositor is in a repaint cycle, input is processed only once
per frame. However, a call to evdev_input_device_data() would handle at
most 8 events at time. When there was more than 8 events pending for a
given frame, input lag would occur. This was most visible with multi
touch input.
This patch changes the evdev_input_device_data() so that it will handle
all the events available in the fd. In order to do that, the fd is put
in non-blocking mode, so that it is possible to loop on read and stop
on EAGAIN instead of blocking.
Initializing pitch to 1 to avoid xxx/pitch errors
This won't influence the valid texture coordinate calculation, because
in that case buffer_attach will provide the correct value.
This makes the compositor and demo clients work on the current nouveau
nvfx driver. Obviously does not fix any clients that actually want a
depth buffer, but this does allow more people to at least try wayland.
mtdev library translates all multitouch based devices to the slotted evdev
protocol. It provides an uniform interface for Weston, which eases mt
implementation when dealing with a big variety of devices.
Weston on drm now directly depends on such library.
Signed-off-by: Tiago Vignatti <tiago.vignatti@intel.com>
Reviewed-by: Peter Hutterer <peter.hutterer@who-t.net>
Commit f992b2fc removed the saved keyboard focus logic to fix a crash
when the saved surface is destroyed. However, setting keyboard focus to
the first surface on the list ends up trying to set the focus to the
cursor surface most of the time. The end result is a NULL keyboard
focus.
This patch restores the saved keyboard focus logic and fixes the crash
mentioned above using a destroy listener.
Without this change, weston would crash whenever a nil buffer was
passed to input_device_attach() if the cursor sprite was not mapped.
While at it, change the unmapping code to use weston_surface_unmap().
Function weston_load_image() was deleted in f02a649a but the wayland
backend was not adapted to the new interface. This probably went
unoticed because the prototype for the missing function was not deleted
from compositor.h so the backend would compile without warnings.
The configure default is to setuid root the weston compositor.
However, if installing as non-root (say, to your prefix in homedir),
the install fails anyway, even if you didn't need setuid to run weston
in your configuration.
On one hand, getopt (in particular the -o suboption syntax) sucks on the
server side, and on the client side we would like to avoid the glib
dependency. We can roll out own option parser and solve both problems
and save a few lines of code total.
Udev provides a convenient helper. Use it instead of working with the
property-list directly.
Signed-off-by: David Herrmann <dh.herrmann@googlemail.com>
If we have multiple video devices on the system (card0, card1), we should
ignore hotplug events for cards that we do not use. This avoids calling
update_outputs() if the event was not generated by our device so we avoid
refreshing the DRM information if it didn't change.
Signed-off-by: David Herrmann <dh.herrmann@googlemail.com>
When creating outputs in the drm compositor, if allocating crtcs
fails, then free the drm resources. Also, if the base output list is
empty, free drm resources
If map is called with a surface of type none it will call
weston_surface_assign_output, even though the surface will
not be mapped.
This change was lost with the changes for using wl_layers.
This can happen for instance running a toytoolkit application, launching
the popup menu and then clicking on a surface that does not belong to
the client that create the popup surface.
When a client with fullscreen surface displayed was aborted by Ctrl-C, the
black surface still be there. Destroy the black surface in
destroy_shell_surface().
When we're repainting, there's no point in polling for input events.
We just read input events once before each repaint and send out events
as needed. The input events come with an accurate timestamp, so this
doesn't affect the timing information and client should always look at
the event timestamps if they're trying to determine pointer motion
speed or double click speed. If we go idle (stop repainting) we add the
input devices back into the primary main loop and wait for the next event.
This avoids waking up the compositor separately (one or more times per
frame) to handle input events. We also avoid updating cursor position
and other compositor state after the client has rendered its new
frame, reducing lag between what the client renders and the pointer
position.
This can happen for instance if the client that started the drag
crashes. Weston would crash because of the invalid surface pointed by
device->drag_surface.
Fix this by reseting the drag surface to nil on a destroy listener.
The surface data structure is now a list of list of surfaces. The core
compositor defines the fade and cursor layer, and it's up to the shell to
provide more layers for the various surface types it implements.
We can now clip the surface bounding box against the previous frame
opaque clip, and then just union the result (visible damage) into
compositor->damage immediately.
Undo fullscreen in shell_unset_fullscreen(), do all the stacking order
in shell_stack_fullscreen(), and configure black surface, method in
shell_configure_fullscreen().
Signed-off-by: Alex Wu <zhiwen.wu@linux.intel.com>
Signed-off-by: Juan Zhao <juan.j.zhao@linux.intel.com>
libbacklight is 300 lines of code in one .c file, and we're relying on
udev changes that aren't yet upstream. For now, let's just keep a
copy in weston and if the Xorg DDX drivers start using libbacklight and
it becomes more widely available, we'll make it an external dependency.
DPMS kicks in only when wscreensaver is launched, in the moment that shell
call lock() for the second time. Backlight control internals are managed by
libbacklight:
http://cgit.freedesktop.org/~vignatti/libbacklight/
Signed-off-by: Tiago Vignatti <tiago.vignatti@intel.com>
This allows each output back end to optimize drawing using overlay planes
and cursors (yet to be integrated). If a surface is assigned to a
plane, the back end should clear its damage field so that the later
repaint code won't look at it.
Ideally, we would want to use <modifier>+Scroll binding but that will have
to wait for axis events. For now we just use keybindings. Zoom in/out with
Super+Up/Down.
We just set the input region to the bounding box of the window frame
and set the opaque region to be the opaque rectangle inside the window
if the child widget is opaque.
If client send set_fullscreen/set_transient request before the first attach,
compositor has no chance to map the surface due to "if (es->output == NULL)".
You can pull it from git://gitorious.org/wayland-for-krh/weston.git map-bug
We never want to update the transform and then damage below. Damage
below is always used to trigger a repaint where the surface used to be
so we need to record the damage before updating the transform.
This change depens on the Wayland core commit:
"protocol: remove absolute coordinates from pointer".
Remove the absolute coordinates from pointer motion and pointer_focus
events.
Signed-off-by: Pekka Paalanen <ppaalanen@gmail.com>
Needed for implementing drag'n'drop icons. When a drag starts, the
compositor will position the top-left corner of the client supplied
icon surface at the cursor hotspot. On the first attach to that
surface, the client may want to reposition it but shell->map did not
take sx and sy parameters.
This changes shell->map interface to take sx and sy parameters and
change dekstop shell implementation to update the position of a
surface of type none according to those parameters. Since a surface
of type none won't actually be mapped, the effect of this change is
only visible for surfaces that are made visible by the compositor.
Signed-off-by: Ander Conselvan de Oliveira <ander.conselvan.de.oliveira@intel.com>
Surfaces like drag'n'drop icons shouldn't receive events as a normal
surface but are still created by the client so add a way for the
compositor to enable or disable the picking of a client surface.
Signed-off-by: Ander Conselvan de Oliveira <ander.conselvan.de.oliveira@intel.com>
The condition to return from surface_attach with a null buffer involves
es->output being non-null. However if a surface was just created this
field would be null and an attach of a null buffer would cause the
compositor to crash.
The other crash happened if surface_attach was called twice with a
null buffer after a valid buffer was attached to the surface. Since
es->buffer was not being set to NULL, surface_attach() would call
wl_list_remove(&es->buffer_destroy_listener.link) twice for the same
surface.
Signed-off-by: Ander Conselvan de Oliveira <ander.conselvan.de.oliveira@intel.com>
I know it's hard to figure out what the shells or backends will be
using as internal API at this point but we can try to minimize the
amount of WL_EXPORT being used anyway.
[pq: redone due to my earlier changes]
Signed-off-by: Pekka Paalanen <ppaalanen@gmail.com>
This clean-up seems alright, but someone with better knowledge has to
doublecheck this function. I guess there's a lot of space for clean-up
there.
[pq: looks ok, redone since did not apply after my changes]
Signed-off-by: Pekka Paalanen <ppaalanen@gmail.com>
Setting a window type is a non-visual operation, it is not supposed to
affect the rendering immediately. Therefore it does not need to apply
surface damage.
The proper surface damage is applied on an attach request following a
window type change.
Signed-off-by: Pekka Paalanen <ppaalanen@gmail.com>
As weston_surface_update_transform() automatically applies before and
after damage on surface geometry change, we don't need to explicitly add
the same damage in motion_notify() for the cursor surface.
We still need the side-effect, that is scheduling a repaint.
Signed-off-by: Pekka Paalanen <ppaalanen@gmail.com>
Round off fractions from non-transformed surface position coordinates.
Transformed surface moved by a client may have non-integer position
coordinates. That is required to prevent drifting on continuous
resizing.
We can round the position to integers when the surface is not (anymore)
transformed.
This change may fix a rounding inconsistency in the opaque region setup,
where the rectangle is initialised from the coordinates without explicit
rounding operation.
Signed-off-by: Pekka Paalanen <ppaalanen@gmail.com>
In the past, weston_surface_configure() was used to change the geometry,
apply damage, and assign an output.
Remove all calls to weston_surface_configure() that do not change the
surface geometry. Add damage calls where needed to keep the wanted
configure side-effects.
Signed-off-by: Pekka Paalanen <ppaalanen@gmail.com>
Move the call to assign an output from weston_surface_configure() to
weston_surface_update_transform().
As update_transform takes new geometry into use, it should also reassign
the output for the surface, but only if an output was already assigned.
Add explicit assing output calls to where we relied on
weston_surface_configure() unconditionally assigning the output.
Signed-off-by: Pekka Paalanen <ppaalanen@gmail.com>
Move the surface opaque region setup from weston_surface_configure() to
weston_surface_update_transform(), so we have less reason to call
update_transform from configure. Opaque region depends on geometry,
after all.
Also move the opaque field from weston_surface to
weston_surface::transform to make this obvious.
Signed-off-by: Pekka Paalanen <ppaalanen@gmail.com>
Remove redundant weston_surface_update_transform() calls from within
output repaint paths, and add a comment that we need to rely on
surface->geometry.dirty == 0 within the repaint sub-functions.
Now that weston_surface_update_transform() does damage as needed, and
weston_output_repaint() explicitly calls update_transform, we can reduce
the updates in rotate_grab_motion() to simply scheduling a repaint. This
will guarantee that the change in rotation ends up on screen ASAP.
Signed-off-by: Pekka Paalanen <ppaalanen@gmail.com>
weston_surface_update_transform() is similar to
weston_surface_configure() in that it changes the surface region on
screen. Unlike configure, update_transform forgets to deal damage at
all, yet it is the only place where we can do damage_below() as needed.
Add a damage_below call to deal damage for the old surface region only
when needed. This uses the cached state from surface->transform to get
the old region.
Add a damage call to deal damage for the new surface region, after
updating the cached state.
Add a repaint call, since we have changed the scene and should render it
out.
This change fixes the rotation not updating the screen properly.
Signed-off-by: Pekka Paalanen <ppaalanen@gmail.com>
The non-transformed case looked a little odd, calling
weston_surface_to_global(), since it already tests for transform.enabled
and simply uses width, height for the box.
Streamline it, by open-coding weston_surface_to_global(), and avoiding
another call into weston_surface_update_transform(). This way it does
not look suspicious to me.
Signed-off-by: Pekka Paalanen <ppaalanen@gmail.com>
In surface_compute_bbox(), call surface_to_global_float() instead of
weston_surface_to_global(). This avoids the recursion:
weston_surface_update_transform()
weston_surface_update_transform_enable()
surface_compute_bbox()
weston_surface_to_global()
weston_surface_update_transform()
which might be non-obvious when reading the code.
Computing the min and max coordinates in floats, we can have a tight
rounding margin by using floor() and ceil().
Signed-off-by: Pekka Paalanen <ppaalanen@gmail.com>
Split two helper functions out of weston_surface_update_transform() to:
- make the code clearer
- update the bounding box properly even if transformation fails
- unify the return point
Also add a comment on what matrix.d[12] is.
Signed-off-by: Pekka Paalanen <ppaalanen@gmail.com>
Extract the core into a function that does not call
weston_surface_update_transform() or schedule repaint.
Signed-off-by: Pekka Paalanen <ppaalanen@gmail.com>
Rotating a surface should not force a full display repaint, so remove
that.
This change exposes a bug: weston_surface_update_transform() does not
apply damage, but it does change surface geometry. While you rotate a
surface, repaints do not work.
Signed-off-by: Pekka Paalanen <ppaalanen@gmail.com>
Transform a menu popup the same as its parent surface.
The parent's transformation is snapshotted at the popup map() time, and
does not follow further parent motion.
Signed-off-by: Pekka Paalanen <ppaalanen@gmail.com>
Computing the real maximal opaque screen aligned rectangle of a
transformed surface is hard. Let's drop the opaque optimisation
instead.
Signed-off-by: Pekka Paalanen <ppaalanen@gmail.com>
When a transformed (rotated) surface is continuously resized from its
top-left corner, its location will drift. This is due to accumulating
rounding errors in transforming an offset from surface-local to global
coordinates in surface_attach().
Diminish the drift down to unobservable level by changing the
weston_surface global position from integer to float.
The offset transformation is now done without rounding. To preserve the
precision, wl_shell::configure() interface must use floats, and so does
weston_surface_configure(), too.
The con of this patch is that it adds inconsistency to the surface
position coordinates: sometimes they are floats, sometimes integers.
Commit a0d6dc4f26 lost one line of code in
the refactoring, and so did not reset the surface damage on repaint
anymore. This causes damage to only accumulate, leading to a full
display redraw every cycle and hiding damage tracking issues.
Put the damage clear back.
Signed-off-by: Pekka Paalanen <ppaalanen@gmail.com>
The uniform location variables should be signed, according to the OpenGL
ES 2 specification. Moreover, GL_NONE, i.e. 0, is not an invalid nor
special location; it is actually used as a valid uniform location.
Change struct weston_shader uniform members to signed.
Stop using 0 for identifying a non-existing uniform, use -1 instead.
Furthermore, as the spec says a) glGetUniformLocation() will return -1
for non-active/existing uniforms, and b) glUniform*() function will
simply ignore all calls with location -1, we can simplify the code. We
don't have to avoid locating uniforms that don't exist, and we don't
need to test for them in weston_surface_draw() either.
Remove the micro-optimisation that avoids setting 'alpha' uniform if it
has not changed, in the name of simplification.
Unify shader creation by dropping init_solid_shader(), and calling
weston_shader_init() instead. The downside is that we compile the vertex
shader twice at startup now.
Signed-off-by: Pekka Paalanen <ppaalanen@gmail.com>
The x, y offsets in attach request are in surface-local coordinates, as
that is the only coordinate system the clients know about. The offset
must be transformed to global coordinate system to be applied properly.
This approximately fixes the top-left resizing of transformed surfaces.
However, it suffers from drift due to accumulating rounding errors in
continuous resizing.
Signed-off-by: Pekka Paalanen <ppaalanen@gmail.com>
This fixes the resize pointer motion vs. surface size mismatch for
right/bottom direction resizes. Top/left resizes need further fixes in
surface motion.
Additionally there is some clean-up in weston_surface_resize() to
eliminate a failure path, and fixing the Weston resize binding's resize
direction heuristic to follow transformations.
Signed-off-by: Pekka Paalanen <ppaalanen@gmail.com>
In the stack of transformations, change the rotation to be applied
to the surface before the absolute positioning. Doing so avoids having
to undo and redo the absolute positioning, and we can simply use the
surface center in local coordinates as the origin.
This fixes the surface move. Before, the surface moved along the surface
local axis, but the user expects it to move along the global axis with
the pointer.
Signed-off-by: Pekka Paalanen <ppaalanen@gmail.com>
Change weston_surface_damage*() functions to use the full surface
bounding box or call surface_compute_bbox() to find the bounding box for
an arbitrary rectangle.
This should fix all rendering artifacts for non-opaque (i.e. ARGB)
transformed surfaces.
Signed-off-by: Pekka Paalanen <ppaalanen@gmail.com>
Use the bounding box to compute an approximation of which output
contains most of the surface.
Move the region32 init outside the loop, and fini it, too.
Signed-off-by: Pekka Paalanen <ppaalanen@gmail.com>
If we ever have transformed cursor surfaces, we would better use the
bounding box to check if it is on the given output.
Signed-off-by: Pekka Paalanen <ppaalanen@gmail.com>
Use the proper bounding box in clipping the surface repaint area. Fixes
excessive clipping for transformed surfaces.
Also don't leak the region32 on early return.
Signed-off-by: Pekka Paalanen <ppaalanen@gmail.com>
weston_surface::transform.boundingbox depends on width and height, and
therefore geometry.dirty flag, so move width and height into geometry.
Fix all users and check that the dirty flag is set.
Signed-off-by: Pekka Paalanen <ppaalanen@gmail.com>
Compute a surface bounding box, especially for transformed surfaces, for
which one cannot simply use x,y,width,height.
The bounding box depends on width and height, so these are now under the
geometry.dirty flag.
Signed-off-by: Pekka Paalanen <ppaalanen@gmail.com>
Now that we can insert a transformation before the surface position
translation, we can drop geometry.x,y from the zoom transformation. That
was just undoing and redoing the position translation.
Signed-off-by: Pekka Paalanen <ppaalanen@gmail.com>
Not sure this check belongs here, but as the position checks are here
too, I added this. Just so we don't forget.
Signed-off-by: Pekka Paalanen <ppaalanen@gmail.com>
weston_surface::transform.position depends on x,y, and therefore the
dirty flag, so move x and y into geometry.
Also add the missing dirty flags.
Signed-off-by: Pekka Paalanen <ppaalanen@gmail.com>
Put the surface translation (absolute position) into the surface
transformations list. This allows to set additional transformations
before and after the global translation.
Having the translation cached, changing the surface x,y now requires to
set the geometry.dirty flag.
Signed-off-by: Pekka Paalanen <ppaalanen@gmail.com>
Previously, if a surface was transformed, it repainted as a whole,
regardless of the computed repaint region. As damage regions determine
repaint regions and whether a surface is considered for drawing at all,
this lead to disappearing surfaces if all surfaces were considered
transformed. Also transparent transformed surfaces were redrawn without
the surfaces below being properly redrawn, leading to alpha-saturation.
Fix that by making texture_region() use the proper global-to-surface
coordinate transformation for texture coordinates. This makes it
possible to call texture_region() also for transformed surfaces.
As texture coordinates may now lie outside the valid texture image, the
fragment shader is modified to check the fragment texture coordinates.
The special path texture_transformed_surface() is no longer used and is
removed.
This change fixes many of the rendering artifacts related to transformed
surfaces.
Signed-off-by: Pekka Paalanen <ppaalanen@gmail.com>
For unifying the coordinate system handling, introduce functions for
converting explicitly between the global and the surface local
coordinate systems.
Use these functions in the input path, replacing
weston_surface_transform().
In the draw path, rewrite transform_vertex() to take in surface local
coordinates.
As shell now uses the new functions, the rotation origin is properly
placed in the middle of the surface.
Signed-off-by: Pekka Paalanen <ppaalanen@gmail.com>
Add the key binding Super+Alt+MouseLeftButton to start rotating a
surface by dragging. The rotation is removed, when the drag is near the
rotation origin.
Rotated surface are a stress test for input event coordinate
transformations, damage region tracking, draw transformations, and
window move and resize orientation.
Signed-off-by: Pekka Paalanen <ppaalanen@gmail.com>
When converting input coordinates from global to surface-local system,
apply the full inverse surface transformation instead of just
translation.
Move weston_surface_update_transform() implementation realier in the
file, no changes.
Signed-off-by: Pekka Paalanen <ppaalanen@gmail.com>
The compositor will likely do an order of magnitude less matrix
inversions than point transformations with an inverse, hence we do not
really need the optimised path for single-shot invert-and-transform.
Expose only the computing of the explicit inverse matrix in the API.
However, the matrix inversion tests need access to the internal
functions. Designate a unit test build by #defining UNIT_TEST, and
export the internal functions in that case.
Signed-off-by: Pekka Paalanen <ppaalanen@gmail.com>
Add a new directory tests/ for unit test applications. This directory
will be built only if --enable-tests is given to ./configure.
Add matrix-test application. It excercises especially the
weston_matrix_invert() and weston_matrix_inverse_transform() functions.
It has one test for correctness and precision, and other tests for
measuring the speed of various matrix operations.
For the record, the correctness test prints:
a random matrix:
1.112418e-02 2.628150e+00 8.205844e+02 -1.147526e-04
4.943677e-04 -1.117819e-04 -9.158849e-06 3.678122e-02
7.915063e-03 -3.093254e-04 -4.376583e+02 3.424706e-02
-2.504038e+02 2.481788e+03 -7.545445e+01 1.752909e-03
The matrix multiplied by its inverse, error:
0.000000e+00 -0.000000e+00 -0.000000e+00 -0.000000e+00
0.000000e+00 0.000000e+00 0.000000e+00 0.000000e+00
-0.000000e+00 -0.000000e+00 0.000000e+00 -0.000000e+00
0.000000e+00 0.000000e+00 0.000000e+00 0.000000e+00
max abs error: 0, original determinant 11595.2
Running a test loop for 10 seconds...
test fail, det: -0.00464805, error sup: inf
test fail, det: -0.0424053, error sup: 1.30787e-06
test fail, det: 5.15191, error sup: 1.15956e-06
tests: 6791767 ok, 1 not invertible but ok, 3 failed.
Total: 6791771 iterations.
These results are expected with the current precision thresholds in
src/matrix.c and tests/matrix-test.c. The random number generator is
seeded with a constant, so the random numbers should be the same on
every run. Machine speed and scheduling affect how many iterations are
run.
Signed-off-by: Pekka Paalanen <ppaalanen@gmail.com>
Implement 4x4 matrix inversion based on LU-decomposition with partial
pivoting.
Instead of simply computing the inverse matrix explicitly, introduce the
type struct weston_inverse_matrix for storing the LU-decomposition and
the permutation from pivoting. Using doubles, this struct has greater
precision than struct weston_matrix.
If you need only few (less than 5, presumably) multiplications with the
inverse matrix, is it cheaper to use weston_inverse_matrix, and not
compute the inverse matrix explicitly into a weston_matrix.
Signed-off-by: Pekka Paalanen <ppaalanen@gmail.com>
Remove the inverse matrix member from struct weston_transform. It is
easier (and probably faster, too) to create and store only forward
transformation matrices in a list, multiply them once, and then invert
the final matrix, rather than creating both forward and inverse
matrices, and multiplying both.
Add a stub for the 4x4 matrix inversion function.
Signed-off-by: Pekka Paalanen <ppaalanen@gmail.com>
Add comments explaining the matrix storage and multiplication, so that
no-one else needs to decipher them again.
Signed-off-by: Pekka Paalanen <ppaalanen@gmail.com>
Having at most one transformation object attached to a surface is not
enough anymore. If we have a surface that needs to be scaled to
fullscreen, and then we have the zoom animation, we already need two
transformations combined.
Implement support for multiple transformations by adding a transformation
list. The final transformation is the ordered composite of those in the
list. To avoid traversing the list every single time, add a dirty flag,
and cache the final transformation.
The existing transformation users (only zoom) are converted.
Note: surface drawing should honour all kinds of transformations, but
not damage region code nor input event translating code take
transformations into account, AFAICT. Therefore anything but translation
will probably behave badly until they are fixed.
Cc: Juan Zhao <juan.j.zhao@linux.intel.com>
Signed-off-by: Pekka Paalanen <ppaalanen@gmail.com>
This removes more special cases from weston_output_repaint() and we
avoid creating and destroying the surface for each animation frame.
We gain another special case in weston_compositor_top(), but that's
less of a problem, and we'll fix that later.
We've trimmed down the actual repaint loop to just iterating through the
surface list and calling weston_surface_draw(), so we push that to the
backend without too much code duplication.
This was supposed to draw black borders around a fullscreen surface that
was smaller than the output. We don't want to special case that in the
repaint loop, but may use a different shader or such. And we want the
surface to have an opaque region that covers the output so that that
will eliminate overdraw of lower surfaces.
This reverts commit fc6ccb868f.
We still need root permissions for drmDrop/SetMaster. Without
integration with ConsoleKit or systemd we also don't have access
to /dev/dri/cardX in the case where we open a new VT.
==30224== Conditional jump or move depends on uninitialised value(s)
==30224== at 0x40EE3A0: evdev_flush_motion (evdev.c:284)
==30224== by 0x40EE6DC: evdev_input_device_data (evdev.c:352)
==30224== by 0x4034710: wl_event_source_fd_dispatch (event-loop.c:76)
==30224== by 0x4035171: wl_event_loop_dispatch (event-loop.c:462)
==30224== by 0x4032F76: wl_display_run (wayland-server.c:785)
==30224== by 0x8050972: main (compositor.c:2183)
Signed-off-by: Tiago Vignatti <tiago.vignatti@intel.com>
I could crash Weston by trying to open another menu from a panel while
one menu from it was already showing.
Program received signal SIGSEGV, Segmentation fault.
0x00007ffff40a9872 in popup_grab_focus (grab=0x761968, time=4130706528, surface=0x0, x=-227, y=15) at shell.c:440
440 if (surface->resource.client == client) {
(gdb) bt
0 0x00007ffff40a9872 in popup_grab_focus (grab=0x761968, time=4130706528, surface=0x0, x=-227, y=15) at shell.c:440
1 0x0000000000406977 in weston_device_repick (device=0x70b4e0, time=4130706528) at compositor.c:360
2 0x0000000000406a36 in weston_compositor_repick (compositor=0x619960) at compositor.c:382
3 0x0000000000406ac8 in destroy_surface (resource=0x6fc6f0) at compositor.c:397
4 0x00007ffff7bd33d8 in destroy_resource (element=0x6fc6f0, data=0x7fffffffd9fc) at wayland-server.c:355
5 0x00007ffff7bd8d98 in for_each_helper (entries=0x757808, func=0x7ffff7bd332c <destroy_resource>, data=0x7fffffffd9fc)
at wayland-util.c:264
6 0x00007ffff7bd8dd4 in wl_map_for_each (map=0x757808, func=0x7ffff7bd332c <destroy_resource>, data=0x7fffffffd9fc)
at wayland-util.c:270
7 0x00007ffff7bd34dc in wl_client_destroy (client=0x7577d0) at wayland-server.c:385
8 0x00007ffff7bd2e36 in wl_client_connection_data (fd=17, mask=1, data=0x7577d0) at wayland-server.c:187
9 0x00007ffff7bd5bde in wl_event_source_fd_dispatch (source=0x74cda0, ep=0x7fffffffdae0) at event-loop.c:76
10 0x00007ffff7bd665b in wl_event_loop_dispatch (loop=0x618900, timeout=-1) at event-loop.c:462
11 0x00007ffff7bd42a9 in wl_display_run (display=0x6188b0) at wayland-server.c:785
12 0x000000000040b1e1 in main (argc=1, argv=0x7fffffffdef8) at compositor.c:2182
Modify popup_grab_focus() to deal with a NULL surface.
Signed-off-by: Pekka Paalanen <ppaalanen@gmail.com>
When we destroy a surface, we damage the surface below so that the area
exposed by the disappearing surface will be repainted. However, if that
surface also is destroyed, the damage information is lost and we fail to
repaint that area.
This commit introduces weston_surface_flush_damage(), which flushes the
surface damage the the surface below when a surface is destroyed. When
multiple surfaces are destroyed at the same time, the damage now accumulates
and sinks down through the surface stack as it should.
gbm_bo_create_from_egl_image() should catch this based on the
GBM_BO_USE_CURSOR_64X64 flag. It currently doesn't so we end up with
a cursor buffer with invalid stride.
If we don't cancel the repaint, we end up pointlessly redrawing the output.
What's worse is that pageflipping to the new buffer eventually fails and
we miss the finish_frame callback, leaving the compositor stuck when we
re-enter the vt.
Normally the repaint will trigger a pageflip, which flips back to our
fb, but that doesn't work if the kms output has been turned off or
set to a different mode.
If the desktop-shell client goes away for any reason, respawn it. To
avoid harmful looping, limit the respawning to 5 times within 30
seconds, and then give up.
Signed-off-by: Pekka Paalanen <ppaalanen@gmail.com>
This is typically when launcing from a pty such as an X (or Wayland)
terminal or from an ssh session. Opening a new vt typically requires root
priviledges, so weston must be setuid root or laucnhed as root for this
to work.
We can only set up once we've acquired the VT and we shouldn't release the
VT until we've cleaned up. Before we would release the VT first, and then
race to drop drm master as X tried to get drm master. Which would kill X.
Part of the point of KD_GRAPHICS mode is that the kernel doesn't try
to restore the VT contents when we switch bach, but leaves that
to the user mode process. This avoids ugly flicker of text mode contents
before the compositor takes over.
When we later upload the cursor image with glTexImage2D(), that will unbind
the image just fine. No need for a NULL upload. Except that that doesn't
currently happen with mesa drivers, but the NULL upload is redundant either
way.
This lands the basic behavior of the popup surface type, but there are still
a number of details to be worked out. Mainly there's a hardcoded timeout
to handle the case of releasing the popup button outside any of the
client windows, which triggers popup_end if it happens after the timeout.
Maybe we just need to add that as an argument, or we could add a new event
that fires in this case to let the client decide whether it ends the popup
or not.
Usually there should be at least one input device, when Weston starts
up, or is reactivated by a VT switch. Add a nice warning, in case there
are no input devices.
This is to give a clue to users who happen to try Weston on DRM, and
do not get any response.
Add also a message to another failure case, that may lead to missing
input devices.
Signed-off-by: Pekka Paalanen <ppaalanen@gmail.com>
Weston initialises to faded-out state, which means all outputs are just
black, even if they render fine.
Previously the fade-in was triggered probably by some random input
event, since Weston was not guaranteed to wake up. Especially with the
drm backend, and the usual problem of not having permissions to input
devices, Weston would not fade in at all and appear broken.
Force Weston to fade in right after initialisation. This show that at
least rendering works, if it does not respond to any input.
Signed-off-by: Pekka Paalanen <ppaalanen@gmail.com>
