Replace the GL renderer with the new rpi-renderer on the Raspberry Pi
backend. This makes Weston on rpi not use EGL or GL anymore, at all.
The weston_plane feature is disabled, since the rpi-renderer does the
same, but better.
Add a command line option to select the output transform. It is not a
weston.ini option for now, since the rpi backend does not read the
configuration file yet. Hopefully that will be done later with some
shared code.
Add the rpi options to 'weston --help' output.
Signed-off-by: Pekka Paalanen <pekka.paalanen@collabora.co.uk>
Dispmanx is the prorietary display API on the Raspberry Pi, which
provides hardware compositing. Every visible surface is assigned a
Dispmanx element, and the hardware or firmware will do all compositing
onto screen. The API supports translation, scaling, flips, discrete
rotations in 90-degree steps, alpha channel on the surfaces, and
full-surface alpha on top.
Previously, Dispmanx capabilities were used via the weston_plane
mechanism, where surfaces were assigned to planes when possible, and
otherwise transparently falling back to GLESv2 compositing. Because we
have no way to use the same memory buffer as a GL texture and a Dispmanx
resource, we had to prepare for both. In the worst case, that means one GL
texture, and two (double-buffered case) Dispmanx resources, all the size
of a whole surface, for all surfaces. This was eating memory fast. To
make things worse (and less slow), the wl_shm buffer was kept around,
since it was copied to either a texture or a resource as needed. This
caused all clients to need two buffers. In a Dispmanx-only renderer, we
can drop the GL texture, and we can release wl_shm buffer immediately
after the first copy, so clients become effectively single-buffered. So
from the worst case of 5 buffers per surface, we go down to 3 or just
2 (single-buffered Dispmanx element, one wl_shm buffer in the client)
buffers per surface.
As this will replace the GL renderer on rpi, we cannot fall back to the
GLESv2 compositing anymore. We lose arbitrary surface rotation, but we
lose also the GL fallback, which caused glitches.
This patch depends on new RaspberryPi firmware. Older firmware may not
render ARGB surfaces correctly, solid color surfaces maybe cause a
performance hit, and the output may completely fail in case the firmware
does not fall back internal off-line compositing properly as needed.
This new rpi-renderer support surface translation and scaling, but not
rotation or transpose (not even in 90-deg steps). In theory, 90-deg step
surface rotation is possible to support. Output transformations are
supported, but flipped variants do not seem to work right.
As a detail, menus and other surfaces that are simply translated with
respect to another surface caused falling back to the GL renderer. The
rpi-renderer handles them directly.
This patch only adds the new renderer, but does not hook it up into use.
Signed-off-by: Pekka Paalanen <pekka.paalanen@collabora.co.uk>
Both GL and pixman renderer (pixman probably only because GL did?)
return the screen capture image as y-flipped, therefore Weston y-flips
it again. However, the future rpi-renderer can produce only right-way-up
(non-flipped) screen captures, and does not need an y-flip.
Add a capability flag for y-flip, which the rpi-renderer will not set,
to get screen captures the right way up.
The wcap recording code needs yet another temporary buffer for the
non-flipped case, since the WCAP format is flipped, and the code
normally overwrites the input image as it compresses it. This becomes
difficult, if the compressor is supposed to flip while processing.
Signed-off-by: Pekka Paalanen <pekka.paalanen@collabora.co.uk>
The upcoming rpi-renderer cannot handle arbitrary rotations. Introduce
Weston capability bits, and add a bit for arbitrary rotation. GL and
Pixman renderers support it.
Shell or any other module must not produce surface transformations with
rotation, if the capability bit is not set. Do not register the surface
rotation binding in desktop shell, if arbitary rotation is not
supported.
Signed-off-by: Pekka Paalanen <pekka.paalanen@collabora.co.uk>
The array of arguments supplied to execv must be NULL terminated. If
unitialized values are used as pointers the exec call may fail with a
EFAULT error ("Bad address").
https://bugs.freedesktop.org/show_bug.cgi?id=64874
If you specify e.g. scale=2 in an output section in weston.ini
we scale all modes by that factor.
We also correctly scale cursor positioning, but ATM there is no
scaling of the cursor sprite itself.
If you specify e.g. scale=2 in weston.ini an output section for the
X11 backend we automatically upscale all normal surfaces by this
amount. Additionally we respect a buffer_scale set on the buffer to
mean that the buffer is already in a scaled form.
This works with both the gl and the pixman renderer. The non-X
backends compile and work, but don't support changing the output
scale (they do downscale as needed due to buffer_scale though).
This also sends the new "scale" and "done" events on wl_output,
making clients aware of the scale.
Rather than storing the shadow_image in the untransformed space
and rotating on copy to hw_buffer we store both on the transformed
space. This means a copy between them is a straight copy, and that
apps supplying correctly transformed surface buffers need not
change them.
We also correctly handle all output transform including the previously
unhandled flipped ones, as well as client supplied buffer_transforms (which
were previously ignored).
We also simplify the actual rendering by just converting any damage
region to output coordinates and set it on a clip and composite
the whole buffer, letting pixman do the rectangle handling. This
means we always do all the transforms, including the surface positioning
as a pixman_image transform. This simplifies the code and sets us up
for handling scaling at a later stage.
The transform looks complicated, but in practice it ends up being
an integer translation almost always, so it will hit the pixman
fastpaths.
The shell_grab_start function sets up a destroy notification on the
shell surface such that when the shell surface is destroyed the pointer
on the grab to the shell surface is set to NULL.
We must therefore check whether the shell surface is NULL and end the
grab if it is.
https://bugs.freedesktop.org/show_bug.cgi?id=64689
This patch fixes a crash with the surface_pong when one of the
seats doesn't have a pointer. This was the case with the RDP compositor
that use a fake seat with no mouse or keyboard.
This patch brings back the user environment from the shell.
In the future, weston-launch could create the Wayland socket earlier, in
which case the user's shell could be used to run Wayland-specific tools
in the new Weston session.
Signed-off-by: Quentin Glidic <sardemff7+git@sardemff7.net>
The only way to create a sub-surface loop by recursive nesting is to
make the main surface (which does not have a role) a sub-surface of any
of its sub-surfaces. All other cases should already be cought.
This change checks for that exact case, and sends a protocol error.
Signed-off-by: Pekka Paalanen <pekka.paalanen@collabora.co.uk>
wl_subsurface.set_desync should apply the cached wl_surface state.
Otherwise, the sub-surface may be stuck: a commit on the parent surface,
if desynchronized, will not commit the sub-surface because it is
desynchronized, too. A commit on the sub-surface may not happen, if it
is waiting for the frame callback from the previous commit.
Signed-off-by: Pekka Paalanen <ppaalanen@gmail.com>
The subsurface patches changed the center_on_output() behavior a bit.
Instead of using the buffer width and height, it now looks at surface
geometry. In lock_surface_configure() we haven't set up the geometry
when we call center_on_output() so the lock surface would never show
up.
Just the same as it is done in shell.c, if the input method process exits
for any reason, we relaunch it automatically, as it is not possible to
launch a standalone application outside of the weston process.
In v2:
- Proper error message when giving up.
Signed-off-by: Eduardo Lima (Etrunko) <eduardo.lima@intel.com>
This set of changes adds support for searching for a given config file
in the directories listed in $XDG_CONFIG_DIRS if it wasn't found in
$XDG_CONFIG_HOME or ~/.config. This allows packages to install custom
config files in /etc/xdg/weston, for example, thus allowing them to
avoid dealing with home directories.
To avoid a TOCTOU race the config file is actually open()ed during the
search. Its file descriptor is returned and stored in the compositor
for later use when performing subsequent config file parses.
Signed-off-by: Ossama Othman <ossama.othman@intel.com>
The function drm_output_start_repaint_loop() unconditionally issues a
page flip, even if the crtc for that output has not been enabled yet.
That causes the page flip to fail, and drm_output_repaint() is never
called.
Solve this by bypassing the initial page flip if the output needs a
mode set.
This has the caveat of affecting latency predictability for that first
frame and when a "driver" mode fullscreen surface causes a mode set.
However, on both cases the mode set would take an unpredictable amount
of time anyway.
https://bugs.freedesktop.org/show_bug.cgi?id=63812https://bugs.freedesktop.org/show_bug.cgi?id=64183
This allows users to change the assigned display profile in GNOME (using
gnome-control-center) or KDE (using colord-kde) and also allows the profiling
tools to correctly inhibit the calibration state whilst measuring the native
screen response.
The subsurface-server-protocol.h header should not be included
by any headers that are part of the SDK since it is not exported.
Otherwise, SDK consumers will break during compilation.
Move this include from compositor.h to compositor.c.
Fixes https://bugs.freedesktop.org/show_bug.cgi?id=64537
Signed-off-by: U. Artie Eoff <ullysses.a.eoff@intel.com>
We must calculate the bounding box of the surface + subsurfaces set and use
that when maximizing the window or going fullscreen.
Signed-off-by: Pekka Paalanen <ppaalanen@gmail.com>
[pq: changed to weston_surface_get_main_surface(), and used a temporary
variable to clean up the expressions.]
Signed-off-by: Pekka Paalanen <ppaalanen@gmail.com>
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>
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>
Add protocol for sub-surfaces, wl_subcompositor as the global interface,
and wl_subsurface as the per-surface interface extension.
This patch is meant to be reverted, once sub-surfaces are moved into
Wayland core.
Changes in v2:
- Rewrite wl_subcompositor.get_subsurface description, and move mapping
and commit details into wl_subsurface description. Check the wording
in wl_subsurface.set_position description.
- Add wl_subsurface.set_commit_mode request, and document it, with the
commit_mode enum. Add bad_value error code for wl_subsurface.
- Moved the protocol into Weston repository so we can land it upstream
sooner for public exposure. It is to be moved into Wayland core later.
- Add destroy requests to both wl_subcompositor and wl_subsurface, and
document them. Experience has showed, that interfaces should always
have a destructor unless there is a good and future-proof reason to not
have it.
Changes in v3:
- Specify, that wl_subsurface will become inert, if the corresponding
wl_surface is destroyed, instead of requiring a certain destruction
order.
- Replaced wl_subsurface.set_commit_mode with wl_subsurface.set_sync and
wl_subsurface.set_desync. Parent-cached commit mode is now called
synchronized, and independent mode is desynchronized. Removed
commit_mode enum, and bad_value error.
- Added support for nested sub-surfaces.
Signed-off-by: Pekka Paalanen <ppaalanen@gmail.com>
Currently the core input code does surface picking before calling into
the focus callback of the current grab. Not all grabs need to pick a
surface however, so we're doing work we don't have to in those cases.
For example, the shell move and resize grabs don't need to pick and the
default grab in implicit grab mode doesn't either.
With this change, the pointer grab mechanism is now very simple:
the focus callback is called whenever the pointer may have a new focus,
the motion callback is called whenever the pointer moves and
the button callback whenever a button is pressed or released.
This was another complication that we had to have to support the
split between libwayland-server and weston. Different grabs want to send
events relative to different surfaces at different times. The default
grab switches between sending coordinates relative to the 'current' surface,
that is the surface the pointer is currently above, or the 'clicked'
surface, in case of an implicit grab.
The grab focus was set by the grab implementation and the core input code
would transform the pointer position to surface relative coordinates for the
grab focus and store in grab->x/y.
Now we can just let the grab implementation transform the pointer
coordinates itself, leaving the implementation free to transform
according to whichever surface it wants. Or not transform at all if
it doesn't need surface relative coordinates (like the shell move and resize
grabs).
The current surface field was used to track the surface the pointer was
currently over along with pointer position relative to that surface,
regardless of implicit or explicit grabs. The main purpose was to restore
the default grab when another grab terminated. We can now just repick in
that case and avoid keeping that state around, with the destroy listener
overhead that involves.
There was one other use case - we used to optimize out calls to
weston_pointer_set_focus() if the focus didn't actually change. We can
still do that, but we have to do that in the default_grab_focus() handler
and compare against weston_pointer->focus instead.
device_setup_new_drag_surface() and device_release_drag_surface() are both
now fairly small and only called from data_device_start_drag() and
data_device_end_grab() respectively. Folding the two functions in where
they're called from simplifies the code flow a bit.
struct weston_surface is now the only surface type we have (in core, shell.c
has shell_surface, of course). A lot of code gets simpler and we never
have to try to guess whether an API takes a wl_surface or a weston_surface.
We can now update the drag icon position directly from the configure
handler or the grab motion handler, and no longer need
weston_seat_update_drag_surface().
Previously we just got the drag_icon signal and had to figure out what
changed. Now we can directly setup or release the drag icon when the
drag starts and stops.
Also rename input_method_context to wl_input_method_context,
input_panel to wl_input_panel and input_panel_surface to
wl_input_panel_surface.
Signed-off-by: Jan Arne Petersen <jpetersen@openismus.com>
Add input_panel_surface::set_panel to specify input panel surfaces which
are overlaying the application and are following the input cursor.
Signed-off-by: Jan Arne Petersen <jpetersen@openismus.com>
We used to rely on an ugly hack where the xwayland server would always
report RGB X windows as having ARGB pixels, so that texturing from these
would also sample the undefined alpha. We also relied on Xrender rendering
to RGB X windows to write the alpha channel correctly, so that when we
texture from the RGB X window as an ARGB surface we end up getting the
alpha written by Xrender.
That was obviously all broken. We can instead reparent client windows into
ARGB frame windows. That way we can render the decorations using a
ARGB render pictformat and sample back those alpha values in a well-defined
way. We can also unbreak xwayland and let it report RGB pixel format for
RGB windows. We still need the opaque region or the RGB-only client window
but that's OK.
in drm_fb_create_dumb, the return value of the drmIoctl function call
to map the dumb buffer was never checked, thus the following "if
(ret)" check was invalid as it was checking the previous return value
from the above drmModeAddFB call.
Signed-off-by: Chris Michael <cp.michael@samsung.com>
At the moment we're only extracting interesting strings. We have to be quite
careful parsing the EDID data, as vendors like to do insane things.
The original EDID parsing code was written by me for gnome-color-manager.
Fixes a segfault. Steps to reproduce:
* start weston with the x11 backend
* open a terminal
* click on the icon in the top left corner, choose close
* close the x11 window containing weston
The keycodes received by the FreeRDP server aren't evdev keycodes.
This patch adds the correct convertion to evdev keycodes. After the
patch all keys that are marked as extended in RDP packets become
functionnal (that's the case for the windows key).
Please note that this patch rely on some corrections that have been
pushed on the FreeRDP github tonight.
Most backends relies on gettimeofday(2) for output repaint timestamps
but this is not a requirement. Before this patch repaints coming from
idle_repaint() always used gettimeofday(2) for timestamps. For backends
not using that time source this could cause large jumps between
timestamps.
To fix this, timestamps needs to always come from the backend. This
means that the backend needs to always be responsible of starting the
repaint loop in case that the repaint cannot start immediately.
The drm backend implementation is from the patch by Ander Conselvan de
Oliveira found here:
http://lists.freedesktop.org/archives/wayland-devel/2013-February/007393.html
Signed-off-by: Jonas Ådahl <jadahl@gmail.com>
This API call handles setting the current surface on the wl_pointer and also
maintaining a destroy notification to monitor that surface for destruction.
This is part of the fix for: https://bugzilla.gnome.org/show_bug.cgi?id=696946
This patch is the 6th version of the FreeRDP based compositor.
Changes from last version:
* use pixman_image_get_stride() when appropriate
* always realloc
Reviewed-by: Jason Ekstrand <jason@jlekstrand.net>
This state is used when the user switches the vt. It turns of rendering
and frame events, but doesn't set the DPMS state to off.
As a part of this change, also turn off the idle timer when entering
the SLEEPING or OFFSCREEN states, which fixes
https://bugs.freedesktop.org/show_bug.cgi?id=61910 (rpi backend
untested).
This ensures the popup_grab.initial_up field isn't reset to 0
if the popup was not opened because of a mouse press but because
of moving the mouse with a popup already open. Not doing so will
make the first click outside the client area go ignored.
This patch implements a popup stack. When the first popup is opened
the grab is started, and it is added to a list. Further popups will
be added to this list but the grab won't change. When a popup is
closed it is removed from the list and, if it is now empty, the grab
is ended.
A click outside the client area will send the popup_done event to
all the popups in the list, and the grab will end.
Before this commit, weston-launch returned 0 if weston was killed by a
signal. This makes it hard to automatically test weston by using
weston-launch, as there is no way to know why weston was terminated.
This commit makes weston-launch return 10+N instead, where N is the code
of the signal that terminated weston. 10 was chosen because it allows a
script to distinguish it from the case that weston-launch itself was
killed by a signal (128+N), and does not overlap the standard exit codes
defined in sysexits.h.
Partial fix for https://bugs.freedesktop.org/show_bug.cgi?id=60935. I
can't reproduce the SIGHUP using the fbdev backend.
v3: better commit message.
Surface will preserve its rotation transformation when maximizing, which
will cause inconsistencies (eg. no window shadows drawn)
This patch removes rotation from maximized surface and restore it when
unmaximizing (just like in fullscreen mode)
To remain consisten with client side window, which blocks moving and
resizing of window in maximized state, we should do so on the compositor
side as well.
(until there is some unmaximize notification from server, which will
allow window to start drawing shadows and revert itself to toplevel
state)
Implements surface transform inheritance. A 'parent' pointer is added to
weston_surface::geometry, and is automatically used by
weston_surface_update_transform(). When updating the transform, the
parent transform is updated as needed, too.
shell_map_popup() is converted to use the new
weston_surface_set_transform_parent() function. Now, if we moved the
popup's parent surface while the popup is open, the popup surface will
stick to the parent properly.
Signed-off-by: Pekka Paalanen <ppaalanen@gmail.com>
Instead of directly setting the dirty flag on weston_surface geometry,
use a function for that.
This allows us to hook into geometry dirtying in a following patch.
Also add comments to weston_surface fields, whose modification causes
transform state to become outdated.
Signed-off-by: Pekka Paalanen <ppaalanen@gmail.com>
Remove redundant geometry dirtying from surface_commit() to simplify
further changes.
This code was added in commit 5df8ecac5d
"compositor: Fix partial repaints"
as the fix to:
https://bugs.freedesktop.org/show_bug.cgi?id=56538
The issue fixed by that commit was making the geometry dirty on every
attach, which caused full-surface repaints every time. The bug was
probably introduced by the opaque region changes during implementing
wl_surface.commit. The mentioned commit fixes the opaque handling by
comparing the new and old regions.
However, the commit also introduces additional checks that set
geometry.dirty. In the current code base, this should be unnecessary.
If the pending.sx or pending.sy are not zero, or if the surface size
changes, the configure() hook is responsible for applying the new
values, and so also marking the geometry dirty.
The configure() hook is only called, if there has been a new
wl_surface.attach. Nothing else can change these variables, so this
should be enough.
Signed-off-by: Pekka Paalanen <ppaalanen@gmail.com>
Cc: Ander Conselvan de Oliveira <ander.conselvan.de.oliveira@intel.com>
When an axis (scroll) event results in a key binding function
being executed, eat the scroll event so the underlying window
doesn't receive it.
Thanks to Scott Moreau for helping me solve this.
Now that we have EXT_buffer_age in mesa, we should stop assuming double
buffering and use the buffer age instead.
Note: this will cause system without the extension to repaint the whole
screen every frame.
Before this change, surface->clip would contain the region of this the
surface that is obscured by opaque containing of other surfaces in any
plane. If a surface was obscured only by a surface in another plane and
the renderer was told to draw that region, the rendering would be
clipped because of that.
This patch changes the clip to be plane specific, so that it contains
only the region of the surface obscured by other surfaces in the same
plane. We also calculate a per plane clip, so that we don't tell the
renderer to draw regions of the primary plane that are obscured by
other planes. This damage remains on the primary plane until the its
clip changes so that they are not obscured anymore.
The core uses this region to clear from the primary plane damage the
area that was repainted. If we add the old buffer damage to that, it
may end up clearing more damage from the primary plane than it was
intended.
The current semantics would implicitly attach the most recently attached
buffer at commit time and send a release event when the buffer enventually
was released. The implicit attach is a little too subtle though and this
patch changes the semantics to always only send release events in response
to an attach event. As a consequence, once a compositor releases a buffer,
it no longer has a reference to it and wl_surfcea.damage is undefined.
Thus, the client side visible change is that damage request must always
be preceeded by a wl_surface.attach request, to ensure there's a valid buffer,
even if that means attaching the same buffer again.
Nothing else calls weston_buffer_post_release and the ref-counting and
sending the release event is essentially what weston_buffer_reference is
all about.
We subtract the screen space opaque region from the transformed surface
region. That's going to fail for when we're scaling up a surface, since
the surface bounding box for the scaled up surface is going to be bigger
than the opaque region. Instead, subtract the surface-space opaque region
from a 0,0 - width,height region and see if that's empty.
We never init this, so we need to copy over the position matrix and then
overwrite the translation entries to make sure we have a valid matrix.
Thanks to Pekka for spotting this (twice).
This way the shell can know when a surface has been unmapped by
checking the value returned by weston_surface_is_mapped(surface).
The configure handlers have now width and height parameters, so
they do not need anymore to check manually the buffer size.
If a surface's buffer is NULL the width and height passed to the
configure are both 0.
Configure is now only called after an attach. The variable
weston_surface.pending.newly_attached is set to 1 on attach, and
after the configure call is reset to 0.
The issue was that touch::down event from the compositor to client apps
would send the previous motion events coordinates and this obviously made
the client do the wrong thing. This happened because we were not waiting
for a SYN event to come from evdev before sending down, motion or up events.
https://bugs.freedesktop.org/show_bug.cgi?id=51909
If there was a fullscreen surface using driver mode when a vt switch is
triggered, but something caused it to be gone when switching back (such
as the client being killed), a call to drm_output_switch_mode() is made
to restore the old mode, and that sets the output's current drm_fb to
NULL, so that the new mode is set drm_output_repaint(). This led to a
crash in vt_func(), because it tried to access output->current for
restoring the old mode.
Fix this by not setting the mode if there's no current fb. Instead,
schedule a repaint so that the mode is set in drm_output_repaint().
https://bugs.freedesktop.org/show_bug.cgi?id=60675
Previously, when coming back from idle the compositor would try to
track if the unlock signal needed to be sent, and the shell would
change the compositor state in order to track when to display or
hide the screensaver.
This patch finishes moving this out of the compositor. With this, the
compositor state should be changed only using the exported functions
weston_compositor_wake() and weston_compositor_sleep(). The unlock
signal will be sent if the compositor wasn't in the ACTIVE state
previously. The lock signal is sent when the compositor becomes idle.
The calls to weston_compositor_wake() in the shell where there to allow
it to trigger the fade in only after the lock surface was configured.
Now the shell has full control of the fade and does not needed to
change the compositor state to do that, so those calls were replaced
with shell_fade() calls.
Previously, the shell would change the idle_time of the compositor to
the screensaver timeout and wake the compositor again, so that after
the timeout the compositor would fade to black and then the shell would
receive the lock signal again and would set the dpms state of the
outputs to off.
Instead, use a screensaver-specific timer for that, and call lock
without the back and forth between the shell and core Weston.
Previously, it was impossible to override the fade in/out behavior of
Weston using a different shell, since this was implemented in core
Weston. This also led to complicated interaction between the shell and
the core when displaying lock surfaces and screensavers.
This patch starts to solve this issue by moving the fade animation out
of the core. On compositor.c, besides deleting the fade code, the idle
handler had to be changed to emit the lock signal, since it was called
from the fade_frame() function before. This causes a slight change of
behavior, since before the fade would happen with the compositor being
active, while now it is already in the idle state. That leads to the
dpms state being set when cancelling the fade with mouse movement, and
in turn, to a slight freeze with drm compositor. This problem will be
fixed in a follow up patch.
On the shell side, the fade was re-implemented in a slightly different
manner. Instead of using a custom frame function, the fade animation
from animation.c is used. The interface for starting the fade was also
changed to take the value of an enum instead of a float alpha value,
in order to improve readability.
The function weston_surface_animation_frame() would schedule a repaint
on all outputs, as weston_surface_schedule_repaint() didn't exist when
it was implemented.
The spring code stops when the current value is withing 0.0002 of the
target. In that case, round the value to 0.0 or 1.0 to enable the use
of fast paths, such as disabling blending in the GL renderer when an
opaque region is set.
Add parameters to weston_fade_run() for setting the initial and target
values for the fade, as well as a parameter to set the spring constant
used for the animation.
Also add the weston_fade_update() function, that allows the animation
to be changed while it is still running.
This will be used to move the fade animation from core Weston into the
shell. These changes are needed to be able to fade out as well as in,
and to be able to reverse the fade in case of user input.
If we get a fullscreen client message before we map the window, we don't
have a shell surface yet. Just dont call set_fullscreen yet, and rely on
xserver_map_shell_surface() to fullscreen it when we map it.
If the window manager doesn't support _NET_WM_STATE_FULLSCREEN, the
fullscreen protocol won't work and we end up waiting indefintely for a
confiure event that doesn't come.
Only try to fullscreen if we have _NET_WM_STATE_FULLSCREEN, and only
wait for map if we're fullscreen.
https://bugs.freedesktop.org/show_bug.cgi?id=61133
When calling shell_map_popup() more than one time on the same shell_surface
the parent transform was getting added more than one time to the transform
list, resulting in an infinite loop when going through the list with
wl_list_for_each in weston_surface_update_transform_enable().
This commit removes the old transform before adding it again.
Because seat_get_keyboard is called after initial
weston_surface_activate, wl_data_device_set_keyboard_focus
fails to send data offer for newly connected client due to
wl_seat.focus_resource being NULL.
This patch calls wl_data_device_set_keyboard_focus
in seat_get_keyboard, so it can send data offer for
newly created client (when wl_keyboard.resource_list
and wl_seat.focus_resource are properly set up).
https://bugs.freedesktop.org/show_bug.cgi?id=60617
We always call enable_udev_monitor and add_devices together and always
disable_udev_monitor and remove_devices together. Let's just have one
entry point for enable and one for disable.
We used to have a bit of naming trouble when the protocol object was called
wl_input_device and the individual evdev devices were call evdev_device.
And we didn't have a drm_seat. Now that we've fixed all that, it's clear
that the drm_seat is all about udev discovery and hotplug of evdev devices,
so let's call it udev_seat instead.
Otherwise we'll kill whatever other display sever we're switching back to.
The tricky thing here is that we never explicitly set drm master in the
startup path, we get that implicitly from being the first to open the
drm device. Even so, we need to drop it before switching VTs.
This patch installs the three header files that define the compositor
plugin interface as well as a pkg-config file. This allows
building weston plugins outside the weston tree. We currently don't make
any guarantees about the plugin API/ABI except that within a stable
branch we won't break it.
libunwind has a dwarf parser and automatically queries the dlinfo
for location of dlopened modules. The resulting backtrace is much
better and includes stack frames in dynamically loaded modules.
krh: Originally submitted for Xorg, adapted for weston:
http://lists.x.org/archives/xorg-devel/2013-February/035493.html
Note this require libunwind at least 1.1 to get the pkg-config files.
Signed-off-by: Marcin Slusarz <marcin.slusarz@gmail.com>
Add an extra cursor_position, which also allows to change the anchor
(for slections). Change the index type to int to allow setting it before
the beginning of a commited string.
The cursor should not be moved as a direct repsonse to this event but
atomically on the next commit_string event.
Signed-off-by: Jan Arne Petersen <jpetersen@openismus.com>
Allows to show/hide the input panel (virtual keyboard) more independent
of focus (some applications might to require additionaly click on a
focused entry to show the input panel).
Signed-off-by: Jan Arne Petersen <jpetersen@openismus.com>
Allows for atomic state changes. Updated surrounding text, content type
and micro focus is taken into account all at once at commit.
Signed-off-by: Jan Arne Petersen <jpetersen@openismus.com>
Move the input_panel interface from desktop-shell to input-method (since
it is not really tied to desktop-shell).
Add an input_panel_surface interface like wl_shell_surface to make it
easier to extend it. Also add a parameter to the set_toplevel request to
be able to specify where to show an input panel surface on the screen.
Signed-off-by: Jan Arne Petersen <jpetersen@openismus.com>
Add a request to be called by the client when the word currently being
composed is tapped by the user. Input methods often use this information
to offer more word suggestions to the user.
Signed-off-by: Jan Arne Petersen <jpetersen@openismus.com>
Support content types in text protocol. Content is defined by a hint
bitmask and a purpose field.
Signed-off-by: Jan Arne Petersen <jpetersen@openismus.com>
Also add a separate preedit-cursor event and add a commit argument to
preedit-string to allow to support commit on reset. Fix editor and
keyboard example to adapt to the protocol changes.
Signed-off-by: Jan Arne Petersen <jpetersen@openismus.com>
The lock hook in desktop-shell only changes the DPMS state the second
time it is called, because during the first time it launches the
screensaver and wakes the compositor again when the screensaver surface
is configured. However, if the screensaver fails to launch, the output
is left in an enabled state, even thought there's no content being
displayed on the screen.
Fix this by disabling the outputs when the screensaver dies if the
shell is still locked.
https://bugs.freedesktop.org/show_bug.cgi?id=60084
Some window managers (kwin at least) don't give us our final fullscreen
size before map_notify. Currently we stop waiting for configire_notify
once we get mapped and in case of kwin that means we don't receive our
fullscreen size configure_notify. With this patch, if we don't get a
configure_notify before map_notify, we just wait for the first one after
map_notify and hope that's our size.
https://bugs.freedesktop.org/show_bug.cgi?id=60608
We were setting up the opaque region to be the window size and then
intersecting with the opaque rectange that's always contained in the window
rectangle. Just initialize to the opaque rectangle.
This must be the last regression from the surface commit changes. We
need to set the pending.input region, otherwise the region will be
overwritten on commit.
https://bugs.freedesktop.org/show_bug.cgi?id=60610
When a surface in deemed unresponsive and we trigger the busy cursor, it's
no longer possible to rotate the surface. This is clearly unacceptable.
This patch adds handling of right clicks to the busy grab so that we can
continue to rotate a surface even if the client is sleeping.
https://bugs.freedesktop.org/show_bug.cgi?id=60609
If our surface has width and height set to the same dimensions as the output
then we can bypassing the scale factor calculation and addition of the
transformation.
The use case that led to this optimisation is the playback of video using
gstreamer-vaapi with the "scale" method. The video is the same dimensions as
the output (1080p.)
When fading in, if a repaint was triggered after a call to
weston_compositor_fade() but before the first call to fade_frame(),
the fade surface wouldn't be drawn because its alpha channel wasn't
initialized properly.
We can now handle fullscreen X windows. X clients request to go fullscreen
buy sending a _NET_WM_STATE client message to the root window. When that
happens we call into the shell interface and asks the shell to make the
surface fullscreen. The shell will then resize the window, which causes
the X wm to configure the X window appropriately.
Make sure we ignore configure requests from fullscreened clients and send out
the synthetic configure notify as required in that case.
Finally, inspect _NET_WM_STATE before mapping so we can handle initial
fullscreen correctly.
Add a frame buffer backend using pixman to render to fbdev.
This has been tested against nouveaufb but nothing else. Much of the code
came straight from the rpi backend (and copyright has been attributed
accordingly).
The behaviour of this backend on less modern frame buffers has yet to be
tested.
The refresh rate is calculated from the frame buffer's metadata. Every frame
is finished in synchrony with the refresh rate.
Frame buffer devices are currently specified on the command line (or using
the default of /dev/fb0); udev could be used in future to enumerate them.
pixman is used for compositing, and a suitable pixman format is built from
the frame buffer's metadata. This doesn't support the full range of
frame buffer formats, but does support varying BPPs of RGBA and ARGB. That
should be enough for now.
The following are not currently supported:
• FOURCC
• Non-packed formats (interleaved, planes, etc.)
• Non-true-colour formats (monochrome, greyscale, etc.)
• Big-endian formats (with component MSBs on the right)
• Non-RGBA and non-ARGB formats
Signed-off-by: Philip Withnall <philip@tecnocode.co.uk>
The drm planes (sprites) only support translation and scaling. Now that
we have matrix.type, we can just look there to see if the transform is
compatible with kms.
Introduce several matrix transform types and track type for matrix.
Could be usefull for activating some fastpath that depends on some
transform type.
Signed-off-by: Vasily Khoruzhick <anarsoul@gmail.com>
Before, cairo-util.h would combine pixman and cairo includes. X11 and
Wayland compositors uses this to load an image as a pixman_image_t but are
forced to include cairo headers. Clients use load_cairo_surface to
load images as cairo_surface_t's, but are forced to include pixman.h.
We move the load_image pixman prototype to its own header, so compositors
can get at the pixman prototype without including cairo.h and clients
can include the cairo based function without including pixman.h.
The X11 backend uses a shadow buffer to be able to support transformed
outputs. However, this belongs in the renderer, since otherwise this
code would have to be copied into every backend that uses the pixman
renderer and supports transformed outputs.
The Android backend has been fairly unused, since we do not have
projects actively using it. It gets basic build testing as part of the
normal build, but runtime testing it takes a considerable effort, and so
I have not done that in a long time. The code is slowly rotting, and
with new emerging backends it starts to be a burden, since it cannot use
udev, but needs evdev.
Therefore to ease Weston going forward, remove the Android backend. If
someone misses it, it can still be found in the stable 1.0 branch, and
of course resurrected from git history if someone wants to put in the
effort and maintain it.
Signed-off-by: Pekka Paalanen <ppaalanen@gmail.com>
The list became stale probably when the lock_layer was introduced. Now
one less (ab)user of weston_surface::link.
Also add a comment on screensaver_configure(), that it is (and has been)
broken for pre-started screensavers.
Signed-off-by: Pekka Paalanen <ppaalanen@gmail.com>
To be able to have a fullscreen surface on one output, and interact with
surfaces on another output, don't lower the fullscreen layer on
activate.
Signed-off-by: Jonas Ådahl <jadahl@gmail.com>
The input region of the black surface placed under the fullscreen shell
surface did not have a specified input region. Because the initial input
region of a surface is infinity, no other surface on any other output
could get any focus.
Signed-off-by: Jonas Ådahl <jadahl@gmail.com>
In particular if the hash table lookup fails and returns NULL then that value
would be passed into weston_wm_window_schedule_repaint which does not accept a
NULL value.
weston_compositor::read_format is in Pixman values now, so comparing to
a GL value does not work. Compare to the right value.
This fix affects only the log output of the GL renderer.
Signed-off-by: Pekka Paalanen <ppaalanen@gmail.com>
When --use-shm is passed to weston and x11 backend is active,
it will use SHM surfaces with pixman renderer instead of EGL
Signed-off-by: Vasily Khoruzhick <anarsoul@gmail.com>
This renderer could be used when there's no graphic accelerator available,
for example in (future) KMS and fbdev backends.
Signed-off-by: Vasily Khoruzhick <anarsoul@gmail.com>
These formats are used by pixman renderer for framebuffer surface, without
this patch screenshoter produces empty image.
Signed-off-by: Vasily Khoruzhick <anarsoul@gmail.com>
Call drm_output_init_egl() instead of duplicating the gbm surface and
gl renderer state initialization code.
Note that this makes error handling a bit worse. Before, if we failed
to allocate a gbm surface we could still recover. Failing the renderer
state creation would lead to inconsisten state. Now we end up in
inconsistent state on both cases.
At this point, we reallocated the gbm surface, but we don't have an
fb with the right size to use. If we're going to a larger mode, the fb
would be too small and the mode set would fail. Besides, the repaint
logic will already do a mode set if necessary, so rely on that instead.
When an output suffers a mode switch, it is possible that a pointer
was inside the old output area but falls outside of it with the new
size. In that case, move the cursor to the output's bottom-right
corner. Otherwise, there's a crash in clip_pointer_motion().
After a mode switch, the output region and transformation matrix need
to be updated. The call to weston_output_move() would do the former but
not the latter, but calling that when the output remains in the same
coordinate doesn't make much sense. Instead, update this state and the
transformation matrix in weston_output_mode_switch().
We want to make sure that the matrix symbols are exported from weston and
that modules get them from there. To do that, we pull matrix.[ch] out of
libshared and back into weston. calibrator now also links to matrix.[ch]
and we add a IN_WESTON define to enable the WL_EXPORT macro when compiled
inside weston.
Trying to create a ARGB framebuffer for scanout results in EINVAL when
trying to queue the pageflip. This patch overrides the format we pass
to addfb2 in case of primary buffers like we do for sprites.
Since we always have to inspect and override the format, don't try to
look up the format in drm_fb_get_from_bo(). Instead return format from
drm_output_check_scanout_format().
Rename drm_surface_format_supported() to drm_output_check_sprite_format()
and make it follow the same convention.
We started scanning out ARGB surfaces in commit e920941032.
Since the call to drmModeSetPlane() fails with EINVAL if the supplied
fb has an unsupported format (which is the case of ARGB8888), the fb
format needs to be overridden.
This avoids calling weston_surface_set_position twice on the same surface. The
second call has no effect in many cases because sx and sy
are usually zero on this path.
This change now means that any sx/sy values passed into ::attach will be
ignored on the first attach for popup surfaces. This similar to the behaviour
for other surface types.
At this point path must point to an allocated string since otherwise the
asprintf that makes the allocation would have failed and we would have
returned earlier.
Make overlays work when the client uses a buffer with the same
transformation as the output.
In order to calculate the destination rectangle, the same logic in
weston_surface_to_buffer_float() is needed, but with the output
dimensions instead. For that reason, this patch generalizes this
function into weston_transformed_{coord,rect} and moves it to util.c.
The surface functions are then implemented using those.
A client can reliably avoid allocating a second buffer per surface, if
the compositor sends the wl_buffer.release event before the frame
callback. To enable clients' single-buffering, release the wl_buffer
early if possible. Otherwise clients will double-buffer.
Releasing early is not possible, if the backend needs the buffer for
migrating a surface to or from a non-primary weston_plane. In that case,
a new buffer must arrive, before the old can be released. Backends will
indicate this by setting weston_surface:keep_buffer to 1 in
assign_planes().
A proper buffer reference in the backends would be better than the
keep_buffer flag, but that would require a per-surface backend private.
The rpi and DRM backends are updated to set keep_buffer, other backends
do not support planes, so do not have to set it.
Signed-off-by: Pekka Paalanen <ppaalanen@gmail.com>
Instead of relying on the compositor core to keep the wl_buffer around
and unreleased, take a new reference to it in gl-renderer. This makes
sure in the future, that the gl-renderer always has the buffer at hand,
client misbehaviour excluded.
The reference is taken in the attach callback, and released in the
flush_damage callback after copy to texture, or when the next attach
callback with a different buffer occurs.
If the surface is not on the primary plane, the buffer is not released
in flush_damage. This ensures, that the buffer stays valid in case the
surface migrates to the primary plane later.
Signed-off-by: Pekka Paalanen <ppaalanen@gmail.com>
The wl_buffer reference counting API has been inconsistent. You would
manually increment the refcount and register a destroy listener, as
opposed to calling weston_buffer_post_release(), which internally
decremented the refcount, and then removing a list item.
Replace both cases with a single function:
weston_buffer_reference(weston_buffer_reference *ref, wl_buffer *buffer)
Buffer is assigned to ref->buffer, while taking care of all the refcounting
and release posting. You take a reference by passing a non-NULL buffer, and
release a reference by passing NULL as buffer. The function uses an
internal wl_buffer destroy listener, so the pointer gets reset on
destruction automatically.
This is inspired by the pipe_resource_reference() of Mesa, and modified
by krh's suggestion to add struct weston_buffer_reference.
Additionally, when a surface gets destroyed, the associated wl_buffer
will send a release event. Often the buffer is already destroyed on
client side, so the event will be discarded by libwayland-client.
Compositor-drm.c is converted to use weston_buffer_reference.
Signed-off-by: Pekka Paalanen <ppaalanen@gmail.com>
The protocol does not require us to flush_damage() on wl_buffer
destruction. In fact, by the time the server receives this request, the
client may have already clobbered the buffer's storage, so we could be
reading undefined data. Instead, just forget about the buffer. The
protocol already says, that a client must not destroy a buffer that is
being read by the server, or the window contents become undefined.
The practical reason for this change is that the following commit can
consolidate wl_buffer destruction listener handlers.
Signed-off-by: Pekka Paalanen <ppaalanen@gmail.com>
If a client called wl_surface.attach with the same wl_buffer as
previously, the compositor would mistakenly send a release on that
buffer. This will cause problems only when clients start to properly use
the wl_buffer.release event.
Do not send wl_buffer.release if the same buffer is attached again.
Signed-off-by: Pekka Paalanen <ppaalanen@gmail.com>
Do not remove the input panel layer from the layer list when the
shell is locked in hide_input_panels().
https://bugs.freedesktop.org/show_bug.cgi?id=56543
Signed-off-by: Jan Arne Petersen <jpetersen@openismus.com>
When redrawing surfaces, use_shader() checks if the desired shader is
already in use to avoid a call to glUseProgram(). However, once the
debug binding is activated, that same check would prevent the usage of
the recompiled shaders until something cause a different shader to be
passed to use_shader().
It's pure WM-related function, so use the same 'weston_wm_' prefix that others
in the same file are already using.
Signed-off-by: Tiago Vignatti <tiago.vignatti@intel.com>
It was already being used in most of launcher.c and window-manager.c, so let's
make a standard everywhere now.
Signed-off-by: Tiago Vignatti <tiago.vignatti@intel.com>
Fix the following build warnings, and the build failures due to the
warning fixes:
CC libshared_cairo_la-image-loader.lo
image-loader.c:369:1: warning: no previous prototype for 'load_image'
CC x11_backend_la-compositor-x11.lo
compositor-x11.c: In function 'x11_output_set_icon':
compositor-x11.c:396:2: warning: implicit declaration of function 'load_image'
compositor-x11.c:396:8: warning: assignment makes pointer from integer without a cast
CC wayland_backend_la-compositor-wayland.lo
compositor-wayland.c: In function 'create_border':
compositor-wayland.c:97:2: warning: implicit declaration of function 'load_image'
compositor-wayland.c:97:8: warning: assignment makes pointer from integer without a cast
Signed-off-by: Pekka Paalanen <ppaalanen@gmail.com>
The implementation of buffer transformation didn't handle transformed
shm buffers properly. The partial texture upload was broken since the
damage is in surface coordinates that don't necessarily match the
buffer's coordinates. It also wouldn't handle the buffer stride
properly, resulting in incorrect rendering if it didn't match the
buffer's width.
The logic used for converting texture coordinates was generalized and
moved out of the renderer, since this conversion may be useful in other
places, such as the backends.
We were pulling in cairo and the image loading libraries through libshared.
Split out libshared into a core libshared and a libshared-cairo that
pulls in the extra libraries.
Since surface.commit was introduced, opqaue regions are stored in a pending
variable that isn't used until surface.commit. Xwayland uses the surface opaque
region as a way to tell weston what region of the surface should be opaque.
However when this pending opaque region was introduced, xwm was not updated
and so we have the 'black = transparent' problem again. This patch fixes the
problem by having xwm use the pending opaque regions.
Allow an input method to forward (unfiltered) key and modifier events
from the hardware keyboard to the client.
Signed-off-by: Jan Arne Petersen <jpetersen@openismus.com>
Rename the key event in text_model to keysym and add serial, time and
modifiers arguments. Add a modifiers_map event to transfer an array of
0-terminated modifier names, so that a mapping of modifiers to the
modifier bit mask is possible.
Signed-off-by: Jan Arne Petersen <jpetersen@openismus.com>
Implement the wl_surface.set_buffer_transform request. This includes
tracking the double-buffered buffer transformation parameter and making
the gl renderer able to handle transformed buffers.
Rename print_egl_error_state() to gl_renderer_print_egl_error_state()
and exports it.
Remove the copy of that function from the rpi backend, and call
the exported function instead.
Signed-off-by: Pekka Paalanen <ppaalanen@gmail.com>
Backends may move surfaces to different planes, in which case damage is
generated in the primary plane. This damage is usually passed to the
renderer, but in some cases the backend may decide to not render
anything (that's the case when drm compositor scans out a client
buffer). In that case the damage on the primary plane would be
discarded, leading to artifacts later.
This patch makes the backend's responsibility to clear the damage on
the primary plane, so that unrendered damage is kept for as long as
necessary.
On the first frame with zoom activated, the spring used for animation
will have a current value of zero. The translation calculated with that
value will be invalid (not a number). Using this value later leads to
having an invalid output matrix, so nothing is composited in the first
zoomed frame.
This problem is most evident when a sprite plane is being used. In that
case, enabling the zoom will cause the surface to be moved back to the
primary plane, but because of the bug described above, this surface
would not actually be rendered causing a quick flicker.
Move fields current_buffer and buffer_damage out of weston_output into
gl_output_state, since they are actually specific to the renderer.
Also bring back the previous_damage field so that the screenshooter
can get the damage for the previous frame in a renderer independent
way.
disable_planes should only be incremented when zoom.active actually
toggles. Otherwise the counter will be incremented too many times,
and planes will no longer get used.
Signed-off-by: Ville Syrjälä <ville.syrjala@linux.intel.com>
Signed-off-by: Ander Conselvan de Oliveira <ander.conselvan.de.oliveira@intel.com>
This moves the surface color state into gles2-renderer. To do this it
adds two new weston_renderer functions. create_surface to be able to
create per-surface renderer state, and surface_set_color to set the
color of a surface and changes it to a color surface.
This moves the EGLConfig, EGLContext and EGLDisplay fields into
gles2-renderer. It also moves EGLDisplay creation and EGLConfig
selection into gles2-renderer.
This introduces callbacks for output creation and destruction for the
gles2-renderer. This enables the gles2-renderer to have per-output
state. EGL surface creation is now done by the output_create callback
and the EGL surface is stored in the new per-output gles2-renderer
state. On the first output_create call, the gles2-renderer will setup
it's GL context. This is because EGL requires a EGL surface to be able
to use the GL context.
This makes drm_fb_get_from_bo() use drmModeAddFB2() if possible so that
drm_output_prepare_overlay_surface() can use this instead of keeping
track of the fbs and buffers itself.
Let the compositor generic code decide what to do when the buffer goes
away. We still have a valid reference do the bo, so we can still show
the client contents until something else triggers a repaint.
If the sprite is disabled and we're not enabling it on the next frame,
nothing is done in the vblank handler, so there's no need to ask for a
vblank event.
The old implementation didn't work because we set the minimum and maximum
sizes so that the WM can't resize us. That makes the fullscreen protocol
not work. Additionally we were requesting fullscreen after mapping, which
requires the more complicated (and potentially flickery) client message
approach.
Now we just set the _NET_WM_STATE before mapping and avoid setting
the size hints in case of fullscreen. That's all good, but the problem
is that we now have to wait for configure notify before we know
what size our output will be. For now we just block and pull events from
X until we get the size.
Ideally we would treat the map as an output hotplug event and just add the
output at that point, but we can't start up with no outputs present.
That may be worth fixing, but for now, the block-on-map is fine.
Dispmanx elements are like hardware overlays. Assign one weston_surface
to each overlay created, and the VideoCore will composite it on screen.
The maximum number of elements is configurable via the command line.
Specifying zero will disable the overlays (planes/elements) altogether,
and use only GLESv2 compositing.
You need an up-to-date Raspberry Pi firmware for:
- vc_dispmanx_resource_create(), that will also take stride. Otherwise
surfaces ending up in elements may show up as corrupted.
- off-line compositing support. The on-line compositing of elements
cannot handle too many elements. Look for the comments around
DEFAULT_MAX_PLANES in the code.
Elements must be double-buffered to avoid tearing. Therefore two buffers
(Dispmanx resources) are allocated for each element. A command line
option is added to allow single-buffering instead to save memory, with
the risk of tearing.
The page flip timer is replaced with the Dispmanx update completion
callback. The callback is executed in a separate thread, therefore a
pipe is set up to integrate properly with Weston core.
If not disabled, usually all surfaces are assigned into planes, and
nothing is composited in GLESv2. Planes do not support surface
transformations though, so compositing will automatically switch the
necessary surfaces to GLESv2 compositing as needed. Switching between
GLESv2 and elements may cause transient visual glitches and jerks.
Signed-off-by: Pekka Paalanen <ppaalanen@gmail.com>
Add a new backend for the Raspberry Pi.
This backend uses the DispmanX API to initialise the display, and create
an EGLSurface, so that GLESv2 rendering is shown on the "framebuffer".
No X server is involved. All compositing happens through GLESv2.
The created EGLSurface is specifically configured as buffer content
preserving, otherwise Weston wouuld show only the latest damage and
everything else was black. This may be sub-optimal, since we are not
alternating between two buffers, like the DRM backend is, and content
preserving may imply a fullscreen copy on each frame.
Page flips are not properly hooked up yet. The display update will
block, and we use a timer to call weston_output_finish_frame(), just
like the x11 backend does.
This backend handles the VT and tty just like the DRM backend does.
While VT switching works in theory, the display output seems to be
frozen while switched away from Weston. You can still switch back.
Seats and connectors cannot be explicitly specified, and multiple seats
are not expected.
Udev is used to find the input devices. Input devices are opened
directly, weston-launch is not supported at this time. You may need to
confirm that your pi user has access to input device nodes.
The Raspberry Pi backend is built by default. It can be build-tested
without the Raspberry Pi headers and libraries, because we provide stubs
in rpi-bcm-stubs.h, but such resulting binary is non-functional. If
using stubs, the backend is built but not installed.
VT and tty handling, and udev related code are pretty much copied from
the DRM backend, hence the copyrights. The rpi-bcm-stubs.h code is
copied from the headers on Raspberry Pi, including their copyright
notice, and modified.
Signed-off-by: Pekka Paalanen <ppaalanen@gmail.com>
When a surface is on a non-primary plane (overlay), we do not need to
keep the GL texture up-to-date, since we are not using it. Avoid calling
glTex(Sub)Image2D in that case, and accumulate the texture damage
separately.
This is especially useful for backends, that can put wl_shm buffers into
overlays.
The empty damage check has to be moved from surface_accumulate_damage()
into gles2_renderer_flush_damage(), because it really needs to check the
accumulated damage, not only the current damage. Otherwise, if a surface
migrates from a plane to the primary plane, and does not have new
damage, the texture would not be updated even for accumulated damage.
Signed-off-by: Pekka Paalanen <ppaalanen@gmail.com>
Instead of hardcoding drm-backend.so as the default if environment
presents neither Wayland nor X11, have a ./configure option to change
it. It still defaults to drm-backend.so, if not given.
Signed-off-by: Pekka Paalanen <ppaalanen@gmail.com>
