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>
Add the concept of debug key bindings, that are bindings that activate
debug features in the compositor. The bindings are added to a list in
the compositor, but the triggering them is left to the shell.
On the shell side, a global debug key binding is added. When the user
presses mod-shift-space, the shell will invoke the debug bindings based
on the next key press.
This also converts the debug shortcuts for repaint debugging, fan
repaint debugging and the hide overlays shortcut in compositor-drm to
use the new infrastructure.
Add a headless backend and a noop renderer, mainly for testing
purposes. Although no rendering is performed with this backend,
this allow some of the code paths inside Weston and shm clients
to be tested without any windowing system or any need for drm
access.
Culling of the repaint of a surface behind an opaque surface on the
same plane was broken by commit 547149a9 [1]. The idea of that commit
is that the damage obscured by an overlay would remain on the primary
plane damage and be repainted when the overlay moved. However, in the
case the two surfaces are on the same plane, the opaque one is not
obscured, so it ends up being repainted.
This commit adds an opaque field to struct weston_plane, that is built
incrementally when accumulating damage. The opaque region of surfaces
on the same plane are removed from the plane's damage, restoring the
previous culling behavior. But since damage behind opaque region of
other planes is maintained, the bug solved in the mentioned commit is
not regressed.
https://bugs.freedesktop.org/show_bug.cgi?id=56537
Apply wl_surface.frame request only on the next wl_surface.commit
according to the new protocol.
This makes it explicit, which repaint actually triggered the frame
callback, since commit schedules a repaint. Otherwise, something causing
a repaint before a commit could trigger the frame callback too early.
Ensure all demo clients send commit after wl_surface.frame. Note, that
GL apps rely on eglSwapBuffers() sending commit. In toytoolkit, it is
assumed that window_flush() always does a commit.
compositor-wayland assumes renderer->repaint_output does a commit.
Signed-off-by: Pekka Paalanen <ppaalanen@gmail.com>
Make input region double-buffered as specified in the new protocol.
While doing it, get rid of the undef region code, and instead use a
maximum sized real pixman region. This avoids special-casing regions
that might sometimes be undef.
As the input region is now usable by default instead of undef,
weston_surface_update_transform() does not need to reset the input
region anymore.
weston_surface_attach() no longer resets the input region on surface
size change. Therefore, also weston_seat_update_drag_surface() does not
need to reset it.
Update toytoolkit to set input region before calling wl_surface_commit()
or swapBuffers (which does commit).
Signed-off-by: Pekka Paalanen <ppaalanen@gmail.com>
Make wl_surface.set_opaque_region double-buffered as required by the new
protocol. Also, do not reset the opaque region on surface size changes
anymore. Only explicit requests from the client will change the region
now.
In clients, make sure commit happens after setting the opaque region.
Mesa does not need a fix, as it never touches the opaque region.
Signed-off-by: Pekka Paalanen <ppaalanen@gmail.com>
This change depends on the Wayland commit
"protocol: double-buffered state for wl_surface".
Implement double-buffering of damage in the compositor as required by
the new protocol.
Ensure all Weston demo clients call wl_surface_commit() after
wl_surface_damage().
Mesa does not need a fix for this, as the patch adding
wl_surface_commit() call to Mesa already takes care of damage, too;
Mesa commit: "wayland: use wl_surface_commit()"
Signed-off-by: Pekka Paalanen <ppaalanen@gmail.com>
This change depends on the Wayland commit
"protocol: double-buffered state for wl_surface".
Clients are now required to issue wl_surface.commit for the
wl_surface.attach to take effect.
While changing this, change the surface argument to
weston_surface_attach() from wl_surface into weston_surface, for
consistency.
Signed-off-by: Pekka Paalanen <ppaalanen@gmail.com>
This is a more generic fix for the issue solved in 4f521731 where
damage obscured by overlays could be lost in one of the output buffers
due to rapid move of a surface in an overlay plane.
This changes the renderer so it keeps track of the damage in each
buffer. Every time a new frame is drawn, the damage of the frame is
added to all the buffers and the rendered regions are cleared from
the current buffer's damage.
Have only one text_model_factory instead of one per seat.
This commit also introduces destruction of an input method when the
corresponding seat is removed.
Signed-off-by: Jan Arne Petersen <jpetersen@openismus.com>
We rename it flush_damage() as it's the point where we update our rendering
API source (eg, the gles2 texture) according to the accumulated damage,
if necessary.
We move the EGL and GLES2 output repaint code into a new gles2-render.c
file. The eglMakeCurrent, glViewPort, surface loop etc was duplicated
across all backends, but this patch moves it to a new file.
Remove weston_surface::opaque_rect completely.
Instead, set the opaque region in xwayland.
Before this patch, black text in xterm was transparent. Now it is not.
However, this patch fixes only a part of the alpha problem. If you apply
full-surface alpha with super+alt+wheel, the problem reappears. This
problem is still due to bad alpha channel contents on xwayland windows.
Signed-off-by: Pekka Paalanen <ppaalanen@gmail.com>
Draw the borders of all the triangles.
v1: original
v2: add keybinding to enable/disable fan debug (super-alt-space),
cycle colors to make it easier to see individual draws, and
redraw undamaged region to clean up previous frames debug
lines
Signed-off-by: Rob Clark <rob@ti.com>
Remove the weston_surface::blend attribute, which really meant that the
texture produced valid alpha values. This was used to override the opaque
region for RGBX surfaces, which produce undefined values for alpha.
Instead, compile a new shader especially for RGBX surfaces, that
hardcodes the sampled alpha as 1.0.
Before "compositor: optimize/simplify shaders" there was a 'vec4 opaque'
in the shaders, that would cause part of the texture to be forced to
alpha=1.0. Now that is gone, and we need this replacement.
To test: launch simple-shm, and use the super+alt+mousewheel combination
to make it transparent. It should not show a light cross over the window.
Signed-off-by: Pekka Paalanen <ppaalanen@gmail.com>
Re-work how the shaders and emitted vertices work. Rather than always
rendering clip-rect sized quads and doing transformation in tex coords
(and requiring the corresponding clipping in frag shader), instead
emit transformed vertices, clipped wrt. dirty region, and use simpler
frag shaders. Also, split the rendering, so blended surfaces with an
opaque region have the opaque region drawn with blend disabled. The
result is considerably fewer pixels drawn with blend enabled, and much
fewer cycles in the frag shader.
This requires having some more complex logic to figure out the vertices
of the shape which forms the intersection of the clip rect and the
transformed surface. Which has perhaps got a few bugs or missing cases,
still (visual glitches in some cases) but at this point more or less is
starting to work. I think it is at least far enough along to get some
initial review.
The result, on small SoC GPU (omap4/pandaboard) on 1920x1080 display,
for simple stuff like moving windows around, I get 60fps (before 30fps
or less), and pushing YUV buffers for hw decoded 1080p video goes from
~6fps to 30fps, with no drop in framerate for transformed/rotated video
surface.
v1: original
v2: check that perpendicular intersect vertex falls within bounds of
transformed surface
v3: update w/ comments and fixes from Pekka Paalanen
v4: fix for full surface alpha from Pekka Paalanen, fix compositor-
wayland build
Signed-off-by: Rob Clark <rob@ti.com>
In cases where the GPU can natively handle certain YUV formats,
eglQueryWaylandBufferWL() can return the value EGL_TEXTURE_EXTERNAL_WL
and the compositor will treat the buffer as a single egl-image-external.
See:
http://www.khronos.org/registry/gles/extensions/OES/OES_EGL_image_external.txt
v1: original
v2: rename EGL_TEXTURE_EXTERNAL_OES -> EGL_TEXTURE_EXTERNAL_WL and query
for the extension
v3: fix build without updated mesa headers, if EGL_TEXTURE_EXTERNAL_WL
#define is missing from older mesa headers.
v4: resend without missing parts
Signed-off-by: Rob Clark <rob@ti.com>
This patch allows rotation and mirroring outputs for x11 and drm backends.
A new 'transform' key can be set in the [output] section. From the protocol:
"The flipped values correspond to an initial flip around a vertical axis
followed by rotation."
The transform key can be one of the following 8 strings:
normal
90
180
270
flipped
flipped-90
flipped-180
flipped-270
Add a wl_seat argument to the activate and deactivate requests of
text_method.
On activation a text_model gets assigned to the input_method of the
wl_seat specified in the activate request.
In cases where we know the surface bounding box doesn't change in the
next frame, we can limit redraws to only the outputs the surface is
currently on. We could do even better by forcing the transform
update so we know where the surface will be in the next frame, but
this is a much simpler first step.
When the entire output is transformed or we're capturing the output
for screenshot or video, disable all output specific overlays
(drm planes, hw cursors etc) and move all surfaces into the primary
plane.
When we analyze and accumulate damage prior to repainting, we need to
accumulate damage per plane, so that whatever damage a surface
contributes is accumulated in the plane that it's assigned to. Before,
we would always accumulate damge in the primary plane, which caused
repaints in the primary plane whenever a surface in a sprite or
framebuffer was damaged. Eliminating this repaint is a big win for
cases where we pageflip to a client surface or use a sprite overlay.
This also prepares for fixing the missing cursor updates, since we
now track damage to the cursor surface in a dedicated sprite plane.
In the wl_seat conversion, struct wl_touch got fields for the focused
surface and the client resource for the input device being focused.
However, the conversion was incomplete: the old fields
weston_seat::touch_focus* we still used by the event dispatching code,
but the new code never set them. Therefore no touch events were ever
sent.
From weston_seat, remove the fields touch_focus, touch_focus_listener,
touch_focus_resource, and touch_focus_resource_listener. They are
replaced by the corresponding fields and listeners from struct
wl_touch.
While doing this, fix touch_set_focus().
If touch_set_focus() was called first with surface A, and then with
surface B, without being called with NULL in between, it would corrupt
the destroy_signal list. It was equivalent of calling wl_signal_add()
for different signal sources with the same listener without removing in
between.
Now, touch_set_focus() first removes focus and listeners, and then
attempts to assign focus if requested. If the target client has not
subscribed for touch events, the touch focus will now be NULL.
Before this patch, the touch focus was left to the previous surface.
NOTE: this patch depends on the patch "server: add lose_touch_focus()"
for Wayland.
Signed-off-by: Pekka Paalanen <ppaalanen@gmail.com>
When we hit a segv, it's often the case that we might crash again in
the attempt to clean up. Instead we introduce a minimal restore callback
in the backend abstraction, that shuts down as simply as possible. Then
we can call that from the segv handler, and then to aid debugging, we
raise SIGTRAP in the segv handler. This lets us run gdb on weston from
a different vt, and if we tell gdb
(gdb) handle SIGSEGV nostop
gdb won't stop when the segv happens but let weston clean up and switch vt,
and then stop when SIGTRAP is raised.
It's also possible to just let gdb catch the segv, and then use sysrq+k
followed by manual vt switch to get back.
Pekka Paalanen
Ander Conselvan de Oliveira
Kristian Høgsberg
John Kåre Alsaker
Tiago Vignatti
Philipp Brüschweiler
Rob Clark
Scott Moreau
Jan Arne Petersen
Rob Bradford
Gwenole Beauchesne