Trying to use an underlay will always fail if the output doesn't support
them, so add a quick check here.
Signed-off-by: Derek Foreman <derek.foreman@collabora.com>
There's only one mode where we can skip the renderer, let's base the check
on that instead of checking for an existing scanout fb.
Signed-off-by: Derek Foreman <derek.foreman@collabora.com>
When a paint node is drawing rotated content, it's treated like its axis
aligned bounding box. Even if it only contains fully opaque content,
the parts of the axis aligned bounding box outside of that content
are not opaque.
We need to ensure we don't claim a paint node that isn't axis aligned
is fully opaque, or we'll improperly update regions outside of the
really opaque content.
Fixes 485e1796af
Signed-off-by: Derek Foreman <derek.foreman@collabora.com>
It makes more sense to update the transform than to bail.
These functions are sort of hints - they have to be correct when true, so
the previous code isn't really buggy. But they make more sense at a glance
this way.
Signed-off-by: Derek Foreman <derek.foreman@collabora.com>
The client-buffer test was setting the desired output refresh rate to
the highest possible, posting a new buffer to the screen, waiting for
the frame callback, then requesting a screenshot be taken.
This was not necessary (we already have a mode for tests which only want
screenshots and not a free-running refresh), and also harmful in that it
setting up a potential race.
When gl-renderer gets asked to repaint with nothing to show, it tries to
read back the GL fence status after the dmabuf has signalled. On drivers
with the threaded context enabled, the GL fence would not be readable,
even if the attached dmabuf was.
The easy fix to this is to just not free-run refresh.
Signed-off-by: Daniel Stone <daniels@collabora.com>
As weston_surface_attach_solid() can change quite a lot about a surface,
we need to mark it as dirty.
Signed-off-by: Daniel Stone <daniels@collabora.com>
transform_damage() returns an allocated set of quads; if a surface had
both opaque and blended regions, we were overwriting the
previously-allocated set of quads for the blended region.
Luckily, transform_damage() doesn't need to be called twice anyway, so
we can fix this by only calling it once in the first case.
Signed-off-by: Daniel Stone <daniels@collabora.com>
No behavior changes; this is a follow-up of "drm: handle client buffer
destroyed while writeback scheduled".
That commit protects against clients disconnecting while a writeback job
is scheduled, which could otherwise lead to crashes if the buffer is
destroyed before the wb job completes. However, output capture provides
the same functionality: it listens to the client buffer destroy event to
retire the capture task.
Instead of listening to the wl_buffer destroy event, simply listen to
the capture task destroy event. GL renderer already follows this
pattern, and now DRM aligns with it.
See also:
weston_capture_task_buffer_destroy_handler()
weston_capture_task_add_destroy_listener()
Signed-off-by: Leandro Ribeiro <leandro.ribeiro@collabora.com>
No behavior change, just a refactor to make it more clear that the state
is freed after the capture task is retired.
Signed-off-by: Leandro Ribeiro <leandro.ribeiro@collabora.com>
YUV420_8BIT and YUV420_10BIT are special DRM formats, which exist to
allow for NV12/P010-alike formats having combined storage for luma &
chroma, rather than split planes.
This is notably used to support AFBC compression for YUV buffers, as
seen with at least Hantro codec engines and Mali GPUs.
Signed-off-by: Daniel Stone <daniels@collabora.com>
Add a way for weston-color to disable color-management, so we have a
simple single-pixel-buffer test.
Signed-off-by: Derek Foreman <derek.foreman@collabora.com>
Before "drm: make writeback format negotiation more robust", pulling a
writeback capture task while another writeback was in progress could
lead to a crash.
That commit avoids the crash, but it relies on
drm_output_find_compatible_writeback() to fail if a writeback task is
already in progress, as the majority of hardware probably support a
single writeback connector compatible with the CRTC.
Although unlikely, hardware may support more than one writeback
compatible with the CRTC. That would break our code, as our writeback
implementation does not support simultaneous writeback tasks per output.
This adds an explicit check and retires the writeback task if there's
already another writeback in progress.
Signed-off-by: Leandro Ribeiro <leandro.ribeiro@collabora.com>
Currently drm_output_get_writeback_formats() returns the formats
supported by a single writeback connector compatible with output->crtc.
This list is used to populate struct weston_output_capture_source_info
through weston_output_update_capture_info().
Later, when pulling writeback capture tasks, we call
drm_output_get_writeback_formats() again. However, as
drm_output_find_compatible_writeback() skips writeback connectors in
use, the returned format list may now differ.
Also, when selecting a writeback connector we implicitly rely on
drm_output_find_compatible_writeback() returning the same connector as
before, without verifying that the chosen connector supports the format
of the buffer provided by the client.
Make drm_output_get_writeback_formats() return the union of formats
supported by all writeback connectors compatible with output->crtc. This
makes the returned format list deterministic, regardless of whether a
writeback connector is currently in use. Although most hardware probably
supports a single writeback compatible with the CRTC, this is a good
change as it makes the code more generic and robust.
Also, add a new format param to drm_output_find_compatible_writeback(),
so now the the selected writeback can be validated against the requested
format.
The main benefit of this patch (and the reason why I wrote) is enabling
us to fix an issue when a writeback task is already in progress and
additional ones are requested:
1. weston_output_pull_capture_task() depends on the writeback format
list
2. if a writeback is already in progress,
drm_output_get_writeback_formats() returns NULL (assuming there's a
single writeback connector).
3. weston_output_pull_capture_task() crashes Weston, as the list of
writeback formats we pass does not match the one stored in struct
weston_output_capture_source_info.
With the format list now deterministic, we'll be able to safely pull the
capture task and retire it. The next commit implements this behavior.
Signed-off-by: Leandro Ribeiro <leandro.ribeiro@collabora.com>
This function does more than just checking if it should wait for
completion: it completes the screenshot if possible. So rename to avoid
confusion.
This also adds documentation to the function.
Signed-off-by: Leandro Ribeiro <leandro.ribeiro@collabora.com>
Currently when a client buffer gets destroyed, the output capture task
gets destroyed with weston_capture_task_buffer_destroy_handler().
The problem is that we may have a writeback task scheduled, so
wb_state->ct would be pointing to a ct that has already been retired
and destroyed.
In this commit we start handling this case.
Signed-off-by: Leandro Ribeiro <leandro.ribeiro@collabora.com>
Otherwise we would end up checking the output of the GL renderer, not
verifying that we set the DRM properties correctly.
Coincidentally this also seems to work around CI flakiness of the test.
Signed-off-by: Robert Mader <robert.mader@collabora.com>
On drivers without explicit modifier support we can't use
eglQueryDmaBufModifiersEXT() to check the correct texture target. We
already hardcode GL_TEXTURE_EXTERNAL_OES for various YUV/YCbCr formats -
let's assume it applies to all of them.
Signed-off-by: Robert Mader <robert.mader@collabora.com>
Perhaps this would make things a bit more obvious to newcomers not
being familiar with historical 2D bitmap hardware sprite.
Signed-off-by: Marius Vlad <marius.vlad@collabora.com>
This disables all of the hardware planes not just overlay (primary and
underlays as well). Change also the debug message.
Signed-off-by: Marius Vlad <marius.vlad@collabora.com>
The intention is that you should be able to replace these with the
pipewire-backend that you can load together with the DRM-backend. The
functionality should be equivalent, but the libweston software
architecture becomes more maintainable for upstream. Also the
pipewire-backend is not tied to the DRM-backend, and can work with any
other backend and even alone.
Once the remoting and pipewire plugins are gone, we can remove the
virtual output API from DRM-backend.
Signed-off-by: Pekka Paalanen <pekka.paalanen@collabora.com>
This was the original video recorder completely in software and with a
custom file format. It is no longer useful to keep. It used the old way
of getting screenshots: hooking to weston_output::frame_signal and
calling weston_renderer::read_pixels(). This old way stalled Weston
until the GPU work was complete, and supported only wl_shm buffers.
Nowadays video recording should be arranged with the pipewire-backend
which supports dmabuf delivery.
Signed-off-by: Pekka Paalanen <pekka.paalanen@collabora.com>
Unused.
This was using the old way of getting screenshots: hooking to
weston_output::frame_signal and calling weston_renderer::read_pixels().
This old way stalled Weston until the GPU work was complete, and
supported only wl_shm buffers.
At this time there is no libweston API for frontends/plugins to ask for
a screenshot, but there is a protocol interface in
weston-output-capture.xml.
Signed-off-by: Pekka Paalanen <pekka.paalanen@collabora.com>
No functional change, just move the comments in the else branch.
Added with 5429302e, ("backend-drm: add KMS plane's formats to
per-surface dma-buf feedback").
Signed-off-by: Marius Vlad <marius.vlad@collabora.com>
Meson's 'coverage-html' target defaults to using the Perl script LCOV.
LCOV has problems with newer GCCs like 14 in Debian Trixie, leading to
many consistency errors about function end lines. Gcovr OTOH runs just
fine.
Create a new target
$ meson compile gcovr-report
that creates a coverage report in HTML and Cobertura using gcovr.
Signed-off-by: Pekka Paalanen <pekka.paalanen@collabora.com>
I was doing
$ mseon setup --wipe
$ meson test
and hit
../../git/weston/tests/constraints-test.c:31:10:
fatal error: pointer-constraints-unstable-v1-client-protocol.h:
No such file or directory
31 | #include "pointer-constraints-unstable-v1-client-protocol.h"
Signed-off-by: Pekka Paalanen <pekka.paalanen@collabora.com>
In function ‘encode_PAM_comment_line’,
inlined from ‘write_PAM_image_rgba’ at ../../git/weston/tests/surface-screenshot-test.c:85:9,
inlined from ‘trigger_binding’ at ../../git/weston/tests/surface-screenshot-test.c:202:8:
../../git/weston/tests/surface-screenshot-test.c:44:22: error: ‘desc’ may be used uninitialized [-Werror=maybe-uninitialized]
44 | size_t len = strlen(comment);
Fixes: d40af215a3
Signed-off-by: Pekka Paalanen <pekka.paalanen@collabora.com>
Since glibc version 2.43, bsearch may return const void * instead of
void * when the input array is const:
"For ISO C23, the functions bsearch, memchr, strchr, strpbrk, strrchr,
strstr, wcschr, wcspbrk, wcsrchr, wcsstr and wmemchr that return
pointers into their input arrays now have definitions as macros that
return a pointer to a const-qualified type when the input argument is a
pointer to a const-qualified type."
So change variable that receives the return value from bsearch to const,
as the input array has the const qualifier. This fixes a "discards const
qualifier from pointer" error.
Signed-off-by: Leandro Ribeiro <leandro.ribeiro@collabora.com>
Before, these pieces of code were allocating, formatting and freeing the
bit flags string regardless of whether debug logging was used or not.
Now, the bit flags string is formatted only when debug logging is
active, and it does not involve malloc+free.
This should improve performance a bit.
Signed-off-by: Pekka Paalanen <pekka.paalanen@collabora.com>
The only difference is a small one in the debug print, otherwise this is
completely identical.
Makes drm_pending_state_apply_atomic() slightly more maintainable.
Signed-off-by: Pekka Paalanen <pekka.paalanen@collabora.com>
This should cut the cost of debug_scene_view_print_buffer() in half on
ARM A55 CPU. Debug printing is quite expensive on such platform.
Signed-off-by: Pekka Paalanen <pekka.paalanen@collabora.com>
Two lines open-coded in two places was not much a problem, but I'm going
to add a new member to weston_buffer that needs freeing, and I want to
do it in just one place.
Signed-off-by: Pekka Paalanen <pekka.paalanen@collabora.com>
Use fputs() as much as possible. My theory is that since fprintf() needs
to scan through the format string to find any formatting codes, it must
be less efficient than fputs() that does not scan.
Signed-off-by: Pekka Paalanen <pekka.paalanen@collabora.com>
This is useful for cases where we already have an open stream which we
can pass straight in and use it when printing node information.
Signed-off-by: Marius Vlad <marius.vlad@collabora.com>
Being able to print the scene-graph straight into a FILE removes one
temporary memory allocation that used to be mandatory. That memory
allocation is now gone from the DRM-backend debug log. It has moved into
the scene-graph log scope. In the case of
weston_log_subscription_printf() it shouldn't matter, because it is only
used when new subscribers appear.
Signed-off-by: Pekka Paalanen <pekka.paalanen@collabora.com>
The old implementation malloc'd a temporary buffer to hold the formatted
string, flushed it out to subscribers, and freed the buffer. On every
single call.
Forwarding the formatted output to the input stream instead avoids the
malloc. It is flushed explicitly so that interleaving messages through
multiple scopes continues to work. Color-lcms module uses that.
Also fix reference to the non-existing function
weston_debug_stream_write().
Signed-off-by: Pekka Paalanen <pekka.paalanen@collabora.com>
Printing directly into an stdio stream and hooking up to the flush (the
write callback) avoids having to allocate+free a temporary buffer on
every printing call. The FILE uses a permanent buffer for storing the
text, and once it fills up or a newline is detected, it is flushed out.
It is fine to mix the new and the old APIs, since the old API flushes
the stream.
The buffer size of 3 kB is just a guess.
The man-page for fopencookie() recommends setting _FILE_OFFSET_BITS to
64. Even though this patch does not strictly need it (we don't implement
seek or take the address of fopencookie()), I think it's good to follow
anyway.
Signed-off-by: Pekka Paalanen <pekka.paalanen@collabora.com>
When building without any of the x11 and wayland we still to have some
unused variables/functions.
Signed-off-by: Marius Vlad <marius.vlad@collabora.com>
Reviewed-by: Erico Nunes <nunes.erico@gmail.com>
