This is the blending space to monitor space color transform. It needs to
be implemented in the renderers, unless a backend sets
from_blend_to_output_by_backend = true, in which case the backend does
it and the renderer does not.
The intention is that from_blend_to_output_by_backend can be toggled
frame by frame to allow backends to react to dynamic change of output
color profile.
For now, renderers just assert that they don't need to do anything for
output color transform.
Signed-off-by: Pekka Paalanen <pekka.paalanen@collabora.com>
See: https://gitlab.freedesktop.org/wayland/weston/-/issues/467#note_814985
This starts building the framework required for implementing color
management.
The main new interface is struct weston_color_manager. This commit also
adds a no-op color manager implementation, which is used if no other
color manager is loaded. This no-op color manager simply provides
identity color transforms for everything, so that Weston keeps running
exactly like before.
weston_color_manager interface is incomplete and will be extended later.
Colorspace objects are not introduced in this commit. However, when
client content colorspace and output colorspace definitions are
combined, they will produce color transformations from client content to
output blending space and from output blending space to output space.
This commit introduces a placeholder struct for color transforms,
weston_color_transform. Objects of this type are expected to be heavy to
create and store, which is why they are designed to be shared as much as
possible, ideally making their instances unique. As color transform
description is intended to be generic in libweston core, renderers and
backends are expected to derive their own state for each transform
object as necessary. Creating and storing the derived state maybe be
expensive as well, more the reason to re-use these objects as much as
possible. E.g. GL-renderer might upload a 3D LUT into a texture and keep
the texture around. DRM-backend might create a KMS blob for a LUT and
keep that around.
As a color transform depends on both the surface and the output, a
transform object may need to be created for each unique pair of them.
Therefore color transforms are referenced from weston_paint_node. As
paint nodes exist for not just surface+output but surface+view+output
triplets, the code ensures that all paint nodes (having different view)
for the same surface+output have the same color transform state.
As a special case, if weston_color_transform is NULL, it means identity
transform. This short-circuits some checks and memory allocations, but
it does mean we use a separate member on weston_paint_node to know if
the color transform has been initialized or not.
Color transformations are pre-created at the weston_output
paint_node_z_order_list creation step. Currently the z order lists
contain all views globally, which means we populate color transforms we
may never need, e.g. a view is never shown on a particular output.
This problem should get fixed naturally when z order lists are
constructed "pruned" in the future: to contain only those paint nodes
that actually contribute to the output's image.
As nothing actually supports color transforms yet, both renderers and
the DRM-backend assert that they only get identity transforms. This
check has the side-effect that all surface-output pairs actually get a
weston_surface_color_transform_ref even though it points to NULL
weston_color_transform.
This design is inspired by Sebastian Wick's Weston color management
work.
Co-authored-by: Sebastian Wick <sebastian@sebastianwick.net>
Signed-off-by: Pekka Paalanen <pekka.paalanen@collabora.com>
Fixes a definitely lost:
== 56 bytes in 1 blocks are definitely lost in loss record 16 of 45
== at 0x48450F8: malloc (vg_replace_malloc.c:309)
== by 0x4B55E93: wl_event_loop_add_timer (event-loop.c:197)
== by 0x4126CF: weston_compositor_create (in /usr/local/bin/weston)
== by 0x409997: main (in /usr/local/bin/weston)
Signed-off-by: Lujin Wang <luwang@nvidia.com>
Signed-off-by: Pekka Paalanen <pekka.paalanen@collabora.com>
Layers did not have a fini sequence before, which means the compositor
layer list might have stale pointers temporarily when shutting down. A
bigger problem might be having views linger after the destruction of the
layer.
These problems were not observed yet, but if they exist, this patch
should help to find them and then fix them.
The check in weston_compositor_shutdown() is not an assert yet, because
it will trigger until all components call weston_layer_fini() correctly.
Some components do not even have a tear-down function to call it from at
all, like fullscreen-shell.
The same with the check in weston_layer_fini().
Signed-off-by: Pekka Paalanen <pekka.paalanen@collabora.com>
These are all the remaining places that still use the global view_list,
and cannot avoid it. Add a comment to explain why in each.
Now all places that use view_list have been audited for paint node
lists.
Signed-off-by: Pekka Paalanen <pekka.paalanen@collabora.com>
Iterate paint nodes instead of the global view list. Right now this does
not change behavior.
This is a step towards using per-output view lists that can then be
optimized for the output in libweston core.
Signed-off-by: Pekka Paalanen <pekka.paalanen@collabora.com>
Iterate paint nodes instead of the global view list. Right now this does
not change behavior.
This is a step towards using per-output view lists that can then be
optimized for the output in libweston core.
Signed-off-by: Pekka Paalanen <pekka.paalanen@collabora.com>
This patch creates a per-output paint node list in the same z-order as
the global view_list in weston_compositor.
The next step is to switch output repaints and backends to use the
z-order list instead of view_list.
Having a per-output paint node list for repaints allows including only
those paint nodes that actually contribute to the output image, so that
completely occluded and out-of-screen views can be ignored in libweston
core already.
Signed-off-by: Pekka Paalanen <pekka.paalanen@collabora.com>
This new object is created for every surface-view-output triplet. As
there is always exactly one surface for a view and it does not change
during a view's lifetime, this is really for a view-output pair or a
surface-output pair.
The object is created on-demand as a part of preparing for an output
repaint, so it applies only to surfaces that are going through repaint.
A prerequisite for that is that the surface is mapped, which means it
has a mapped view.
When any one of surface or view gets destroyed or output gets disabled,
all related paint nodes are destroyed.
In future, paint node will be useful for caching surface-output or
view-output pair dependent data:
- damage regions for overlapping outputs
- color transformations
- backend-specific bookkeeping (e.g. DRM KMS plane assigments)
- per-output repaint lists
- surface geometry transformed into output space
Suggested by Daniel Stone in
https://gitlab.freedesktop.org/wayland/weston/-/merge_requests/582#note_899406
PS. The call in weston_view_destroy() to
weston_compositor_build_view_list() might be so that if the view has
sub-surfaces, rebuilding the view list removes those those too and
automagically deletes their views.
Signed-off-by: Pekka Paalanen <pekka.paalanen@collabora.com>
This drops the software presentation clocks that could jump backwards.
See the previous commit "libweston: assert frame times never go
backwards" for the rationale.
Signed-off-by: Pekka Paalanen <pekka.paalanen@collabora.com>
Adding this check was prompted by
https://gitlab.freedesktop.org/wayland/weston/-/merge_requests/609
There is no reason to allow frame times jump backwards, and apparently
we already have code that makes that assumption.
DRM KMS uses CLOCK_MONOTONIC as the vblank and page flip timestamps,
which by definition cannot go backwards. Other backends call
weston_compositor_set_presentation_clock_software().
Frame times are also reported directly to Wayland clients via
presentation-time extension, and clients too will not expect that the
timestamp could go backwards.
So make sure time can never go backwards.
Signed-off-by: Pekka Paalanen <pekka.paalanen@collabora.com>
This adds a heuristic for freeing shader programs that have not been
needed for a while. The intention is to stop Weston accumulating shader
programs indefinitely, especially in the future when color management
will explode the number of possible different shader programs.
Shader programs that have not been used in the past minute are freed,
except always keep the ten most recently used shader programs anyway.
The former rule is to ensure we keep shader programs that are actively
used regardless of how many. The latter rule is to prevent freeing too
many shader programs after Weston has been idle for a long time and then
repaints just a small area. Many of the shader programs could still be
relevant even though not needed in the first repaint after idle.
The numbers ten and one minute in the above are arbitrary and not based
on anything.
These heuristics are simpler to implement than e.g. views taking
references on shader programs. Expiry by time allows shader programs to
survive a while even after their last user is gone, with the hope of
being re-used soon. Tracking actual use instead of references also
adapts to what is actually visible rather than what merely exists.
Keeping the shader list in most recently used order might also make
gl_renderer_get_program() more efficient on average.
last_repaint_start time is used for shader timestamp to avoid calling
clock_gettime() more often. Adding that variable is an ABI break, but
libweston major has already been bumped to 10 since last release.
Signed-off-by: Pekka Paalanen <pekka.paalanen@collabora.com>
Converting a region from global coordinates to output pixel coordinates
will become useful in GL-renderer soon, so move this function to be
shared. It is tricky to reinvent.
Signed-off-by: Pekka Paalanen <pekka.paalanen@collabora.com>
This adds the libweston capability bit for "color operations" which
refers to a renderer's support for operations needed for color
management. GL-renderer will grow the support while Pixman-renderer will
not, which is why the cap is needed.
To make an example use of the cap, this also adds new API:
weston_output_set_renderer_shadow_buffer(). This is a temporary API to
enable future experimental features. The first such feature will be the
renderer internal shadow buffer, the boolean variable for it taken from
Harish Krupo's "weston.ini: introduce use-shadow-fbo in output config".
Obviously this patch does not implement the renderer shadow buffer. No
renderer sets WESTON_CAP_COLOR_OPS yet so trying to enable it will fail.
The documentation here is deliberately vague, because the bits needed
for color management will come in trickling for a long time until we can
call it color management in any sense. Until then, the temporary API
shall remain, perhaps poorly named.
Cc: Harish Krupo <harishkrupo@gmail.com>
Signed-off-by: Pekka Paalanen <pekka.paalanen@collabora.com>
Until now we had the test quirks initialization in wet_main(),
just after calling weston_compositor_create(). But there are
some cases that require the quirks during struct weston_compositor
creation time.
Move test quirks initialization to weston_compositor_create()
in order to cover more use cases for the test quirks mechanism.
Signed-off-by: Leandro Ribeiro <leandro.ribeiro@collabora.com>
Until now we had struct wet_testsuite_data as an opaque
struct that should be defined by the testsuite of libweston
users. Instead, keep the data as a void * and document that
users are responsible for defining the data type.
Signed-off-by: Leandro Ribeiro <leandro.ribeiro@collabora.com>
There are some specific cases in which we need Weston to
behave differently when running in the test suite. This
adds a new API to allow the tests to select these behaviors.
For instance, in the DRM backend we plan to add a writeback
connector screenshooter. In case it fails for some
reason, it should fallback to the renderer screenshooter
that all other backends use. But if we add a test to
ensure the correctness of the writeback screenshooter,
we don't want it to fallback to the renderer one, we
want it to fail. With this new API we can choose to
disable the fallback behavior specifically for this test.
Signed-off-by: Leandro Ribeiro <leandro.ribeiro@collabora.com>
The opaque region is used to determine where the views underneath the current
view must be drawn. If the opaque is not clipped, then the area that is part of
the opaque region but not part of the scissor area is not drawn at all.
Signed-off-by: Michael Olbrich <m.olbrich@pengutronix.de>
Subsurfaces inherit the scissor region from the parent surface. Currently
the region is updated at the end of weston_view_update_transform(). As a
result, the old region is used to clip the transform.boundingbox of the
subsurface.
Change the order to update the scissor region after the transform.matrix is
updated but before it is used.
Signed-off-by: Michael Olbrich <m.olbrich@pengutronix.de>
As from commit b7e5f10bf4, weston_view_is_opaque() is called from
debug_scene_graph_cb(), which on its own represents a (different)
scope. By default, we already have a subscriber for the 'log' scope,
which will cause a harmless, yet spurious, message.
Signed-off-by: Marius Vlad <marius.vlad@collabora.com>
compositor_accumulate_damage() is called for each output during repaint.
The DRM backend will only set keep_buffer for the surfaces that are visible on
the current output. So a buffer_ref is released that may still be needed. When
the output that shows the surface is repainted, the buffer_ref is gone and the
surface cannot be put on a plane.
Ignore all surfaces that are not visible on the current output to avoid this.
Signed-off-by: Michael Olbrich <m.olbrich@pengutronix.de>
Currently the debug output for 'drm-backend' can be confusing. In the output of
debug_scene_view_print() views may be listed as 'not opaque' but later, during
plane assignment, other views underneath such a view is reported as 'occluded on
our output'.
This happens because weston_view_is_opaque() has some extra checks to determine
if a view is fully opaque, such as 'is_opaque' provided by the renderer for
formats that have no alpha channel.
Use weston_view_is_opaque() in debug_scene_view_print() as well to get more
accurate results.
Signed-off-by: Michael Olbrich <m.olbrich@pengutronix.de>
All timeline event timestamps are in CLOCK_MONOTONIC already. DRM KMS
timestamps are practically guaranteed to be CLOCK_MONOTONIC too, even though
presentation clock could theoretically be something else. For other backends,
the presentation clock is likely CLOCK_MONOTONIC_RAW due to
weston_compositor_set_presentation_clock_software().
This patch ensures that the recorded vblank timestamp is in CLOCK_MONOTONIC.
Otherwise interpreting the timeline traces might be difficult to do accurately,
since it would be hard to recover the relationship between the presentation
clock and timeline event timestamps.
The time conversion routine is the simplest possible, I don't think we need any
more accurate conversion for timeline purposes. Besides, DRM-backend is the
only backend where the timings actually matter, the other backends are
software-timed anyway.
Since the clock domain of the "vblank" attribute potentially changes, the
attribute is renamed. Wesgr never used this attribute.
Signed-off-by: Pekka Paalanen <pekka.paalanen@collabora.com>
If a surface has subsurfaces then the surface itself is in the subsurface
list. To avoid printing it again there is a check to skip the child view,
if it is the same as the current view.
However, this fails when a surface with subsurfaces has two (or more) views:
The check to skip the parent fails for the other view and the two views are
printed again and again until a stack overflow occurs.
So instead check if the parent view of the subsurface view is the current
view. This way, any view that does not belong to a real subsurface is
skipped.
As a side effect, this ensures that each view of the subsurfaces is only
printed once at the correct place in the hierarchy.
Signed-off-by: Michael Olbrich <m.olbrich@pengutronix.de>
Like physical size, subpixel arrangement, etc, transform advises of a
physical transform of a head, if present.
This commit adds the transform member and setter to weston_head, however
it is currently unused.
Signed-off-by: Lucas Stach <l.stach@pengutronix.de>
[daniels: Extracted from one of Lucas's patches.]
Signed-off-by: Daniel Stone <daniels@collabora.com>
It was discovered in issue #99 that the implementations of the 90 and 270
degree rotations were actually the inverse of what the Wayland specification
spelled out. This patch fixes the libweston implementation to follow the
specification.
As a result, the behaviour of the the weston.ini transform key also changes. To
force all users to re-think their configuration, the transform key values are
also changed. Since Weston and libweston change their behaviour, the handling
of clients' buffer transform changes too.
All the functions had their 90/270 cases simply swapped, probably due to
confusion of whether WL_OUTPUT_TRANSFORM_* refers to rotating the monitor or
the content.
Hint: a key to understanding weston_matrix_rotate_xy(m, c, s) is that the
rotation matrix is formed as
c -s
s c
that is, it's column-major. This fooled me at first.
Fixing window.c fixes weston-terminal and weston-transformed.
In simple-damage, window_get_transformed_ball() is fixed to follow the proper
transform definitions, but the fix to the viewport path in redraw() is purely
mechanical. The viewport path looks broken to me in the presence of any
transform, but it is not this patch's job to fix it.
Screen-share fix just repeats the general code fix pattern, I did not even try
to understand that bit.
https://gitlab.freedesktop.org/wayland/weston/issues/99
Signed-off-by: Pekka Paalanen <pekka.paalanen@collabora.com>
Clarifies which direction the transformation happens. All exported function
need documentation.
Signed-off-by: Pekka Paalanen <pekka.paalanen@collabora.com>
The only reason why we have both weston_compositor_tear_down() and
weston_compositor_destroy() is that the only we had to destroy
the log context was keeping weston_compositor alive and calling
weston_log_ctx_compositor_destroy().
After commit "weston-log: replace weston_log_ctx_compositor_destroy()
by weston_log_ctx_destroy()", it's not necessary to keep a zombie
weston_compositor just to be able to call
weston_log_ctx_compositor_destroy().
Fold weston_compositor_tear_down() into weston_compositor_destroy(),
as this split is useless now.
Signed-off-by: Leandro Ribeiro <leandrohr@riseup.net>
The function weston_log_ctx_compositor_setup() is being called only inside
weston_compositor_create() and it is so tiny that the code gets easier to
follow if it gets folded in weston_compositor_create().
Signed-off-by: Leandro Ribeiro <leandrohr@riseup.net>
Commit "weston-log: add function to avoid direct
access to compositor members in non-core code" added the
function weston_compositor_add_log_scope mainly to allow
libweston users to avoid direct accessing core structs, as
weston_compositor.
Replace weston_log_context_add_log_scope usage by
weston_compositor_add_log_scope.
Signed-off-by: Leandro Ribeiro <leandrohr@riseup.net>
weston_environment_get_fd was declared in weston-launch and implemented
in compositor.c. Since the function is not used elsewhere in the code,
it is replaced by a static function in launcher-weston-launch.c
Signed-off-by: Guillaume Champagne <champagne.guillaume.c@gmail.com>
There's a function named weston_compositor_log_scope_destroy()
but it doesn't take a struct weston_compositor argument.
Rename it to weston_log_scope_destroy(), as the argument is a
struct weston_log_scope.
Signed-off-by: Leandro Ribeiro <leandrohr@riseup.net>
There's a function named weston_compositor_add_log_scope()
but it doesn't take a struct weston_compositor argument.
Rename it to weston_log_ctx_add_log_scope(), as
the log_scope is being added to a log_context.
Also, bump libweston_major to 9.
Signed-off-by: Leandro Ribeiro <leandrohr@riseup.net>
The member previous_damage from struct weston_output is no longer necessary.
First, stop calling init, fini and copying output_damage to it. Then remove
it from struct weston_output.
Signed-off-by: Leandro Ribeiro <leandrohr@riseup.net>
The emission of frame_signal has to happen before a flip, otherwise
glReadPixels() could read an old frame or even worse an uninitialized buffer.
So move frame_signal emission back to renderers.
This reverts commit 2619bfe420.
Signed-off-by: Leandro Ribeiro <leandrohr@riseup.net>
This will make possible to users that are listening to frame_signal to get
previous_damage from the data parameter instead of using
output->previous_damage.
Signed-off-by: Leandro Ribeiro <leandrohr@riseup.net>
In order to remove duplication and make the code easier to follow, move
frame_signal emission from renderers to weston_output_repaint(). This should
have no observable effect.
Signed-off-by: Leandro Ribeiro <leandrohr@riseup.net>
This is necessary for the test harness to be able to execute the compositor
multiple times in the same process. As we never unload opened modules, the
first compositor iteration will leave them all loaded and following compositor
iterations will then have them already loaded.
Signed-off-by: Pekka Paalanen <pekka.paalanen@collabora.com>
This introduces a new convention of checking through the compositor destroy
listener if the plugin is already initialized. If the plugin is already
initialized, then the plugin entry function succeeds as a no-op. This makes it
safe to load the same plugin multiple times in a running compositor.
Currently module loading functions return failure if a plugin is already
loaded, but that will change in the future. Therefore we need this other method
of ensuring we do not double-initialize a plugin which would lead to list
corruptions the very least.
All plugins are converted to use the new helper, except:
- those that do not have a destroy listener already, and
- hmi-controller which does the same open-coded as the common code pattern
did not fit there.
Plugins should always have a compositor destroy listener registered since they
very least allocate a struct to hold their data. Hence omissions are
highlighted in code.
Backends do not need this because weston_compositor_load_backend() already
protects against double-init. GL-renderer does not export a standard module
init function so cannot be initialized the usual way and therefore is not
vulnerable to double-init.
Signed-off-by: Pekka Paalanen <pekka.paalanen@collabora.com>
Adds a new callback 'can_scanout_dmabuf' in weston_backend, which
can be set by the back-end do determine if the buffer supplied can be
imported directly by KMS.
This patch adds a wrapper over it, 'weston_compositor_dmabuf_can_scanout'
which is called before importing the dmabuf in the GPU if the
direct_display dmabuf is being set. If that's true and the check
failed, we refuse to create a wl_buffer.
This patch avoids importing in the GPU.
Signed-off-by: Marius Vlad <marius.vlad@collabora.com>
Due to an error in driving GitLab, this commit erroneously contained the
entirety of !267 (zpos support in the KMS backend) squashed into one
single commit, pushed into master.
In order to keep the history clean, this is being reverted; a rebased
version of !267 with the clear individual commits which were already
present will be applied in its place.
This reverts commit 95e3b0deae.
timeline subscription
When subscribing over the command line to the 'timeline' scope we hit
the situation where we could emit a timeline message but without the
weston_output object being (fully) enabled. The timeline subscription
object requires to install its own callback on the 'destroy_signal' but
at that time, the 'destroy_signal' is not initialized.
This moves 'destroy_signal' initialization before timeline has a chance
to emit a timeline subscription message for that weston_output.
While at it, move also 'frame_signal' initialization before any function
call to keep them nicely organized.
Signed-off-by: Marius Vlad <marius.vlad@collabora.com>
We notify the timeline of the fact that the object suffered
modifications through the 'set_label' function. Remove the old
refresh variable.
Signed-off-by: Marius Vlad <marius.vlad@collabora.com>
With the timeline scope being created it is time to convert TL_POINT()
to use the timeline scope through the compositor instance.
This patch removes the global variable allowing to run the new timeline
code.
Signed-off-by: Marius Vlad <marius.vlad@collabora.com>
With it this removes the parts responsible for creating the file,
timeline_log class, removes the debug key binding when creating the
compositor instace, keeping only what can be re-used.
Signed-off-by: Marius Vlad <marius.vlad@collabora.com>
Pekka Paalanen
Lujin Wang
Maxime Roussin-Bélanger
Leandro Ribeiro
Michael Olbrich
Marius Vlad
Lucas Stach
Leandro Ribeiro
Guillaume Champagne
Daniel Stone
Marius Vlad