We need only check that the region is not empty. If either the input region or
the constraint region have degenerate extents, the intersection from the
previous instruction will set confine_region->data to pixman_region_empty_data.
Fixes: b6423e59
Signed-off-by: Sergio Gómez <sergio.g.delreal@gmail.com>
There's a comment explaining how to hack the DRM-backend in order to
fake that a certain format is not supported by the KMS device. This is
useful in order to test dma-buf feedback implementations using the
simple-dmabuf-feedback client. But with recent changes on the
DRM-backend, this got outdated.
Drop this comment, as everyone interested in this client is probably
familiar enough with the DRM-backend in order to do that.
Also made some adjustments to other comments explaining how this client
works.
Signed-off-by: Leandro Ribeiro <leandro.ribeiro@collabora.com>
We log the reasons why the fb of a certain view was not placed in an
overlay plane and use that for debug purposes. With these reasons we
also decide if the scanout tranche should be included on the dma-buf
feedback or not. For instance:
1. If the reason is the incompatibility between the format/modifier
pair of the fb and those supported by the KMS device, the scanout
tranche is added and feedback is re-sent (so that the client can
re-allocate with parameters that makes it eligible for direct
scanout).
2. If the reason is because we have no overlay planes available, the
scanout tranche is useless. So the scanout tranche is removed and
the feedback re-sent (so that clients can re-allocate with
parameters optimal for the render device).
Also, when we detect that a view is eligible for direct scanout, we
don't even consider sending new feedback, as our interpretation of the
dma-buf feedback spec was that we should avoid bothering clients with
new feedback when they are already hitting direct scanout.
After some discussions and clarifications regarding the spec, we've
realized that Weston should start to also include the scanout tranche
even when the compositor is able to place client's content on overlay
planes. Basically, because this gives a chance for clients to
re-allocate with the proper parameters (not only format/modifier pair,
but also the target_device and the flags) from the scanout tranche. In
this patch we start doing this.
Signed-off-by: Leandro Ribeiro <leandro.ribeiro@collabora.com>
It makes no sense to keep the scanout tranche on the dma-buf feedback if
there are no overlay planes available. So start to remove it.
Signed-off-by: Leandro Ribeiro <leandro.ribeiro@collabora.com>
We're still timing out with our basic xwayland test in CI, so this
enables the XWM logging scope, and enables some further print debugs we
have available in our helper library, in an attempt to further
investigate and determine why we're still timing out.
Signed-off-by: Marius Vlad <marius.vlad@collabora.com>
This will let command-line Git tools re-map my name and e-mail address properly.
I'm using my personal e-mail address and not my Collabora address because I'm
not actively contributing to Wayland anymore and this is mostly for letting
people find me should they dig me up in the project history.
Signed-off-by: Faith Ekstrand <faith@gfxstrand.net>
wayland_output_destroy_shm_buffers() is called immediately before
output_destroy() of the renderer is called. And for the pixman renderer all
renderbuffers must be destroyed before the output can be destroyed.
Also, weston_renderbuffer_unref() is not called when the buffer is released
because buffer->output is now NULL, so the renderbuffer would be leaked.
So just unref the renderbuffer immediately. Set it to NULL to avoid unreffing it
again should wayland_output_destroy_shm_buffers() be called again before the
buffer is released. This can happen during an xdg-shell resize.
Signed-off-by: Michael Olbrich <m.olbrich@pengutronix.de>
This adds a test to ensure that the DRM-backend writeback screenshooter
is working properly.
Signed-off-by: Leandro Ribeiro <leandro.ribeiro@collabora.com>
In commit "drm-backend: add writeback connector screenshooter to
DRM-backend" we were failing the writeback screenshot when the DRM/KMS
driver would take longer than the atomic commit to finish. In this patch
we address such case.
Signed-off-by: Leandro Ribeiro <leandro.ribeiro@collabora.com>
In this patch, we add the writeback connector screenshooter to the
DRM-backend.
This will be useful to create plane composition tests that will run in
our CI, as VKMS already supports writeback connectors.
Signed-off-by: Leandro Ribeiro <leandro.ribeiro@collabora.com>
With this change, we expose the DRM-backend writeback source through the
output capture interface, making it available to clients.
For now we'll always fail writeback screenshots requests, because we
still don't have the writeback screenshooter implementation on the
DRM-backend. We add that in the following commits.
Signed-off-by: Leandro Ribeiro <leandro.ribeiro@collabora.com>
In the following commits we add a writeback screenshooter. For that,
we'll need the formats supported by the writeback connectors. So include
the supported formats in struct drm_writeback.
Signed-off-by: Leandro Ribeiro <leandro.ribeiro@collabora.com>
Allow VNC clients that support the cursor pseudo encoding to render
the cursor themselves. This reduces observable latency of cursor
movement.
Signed-off-by: Philipp Zabel <philipp.zabel@gmail.com>
The state does not own the fd. This is usually not a problem, because the
in_fence_fd of the state is assigned during drm_assign_planes() and then
immediately used in drm_repaint_flush(). It cannot be closed in-between.
However, in the fallback path in drm_output_start_repaint_loop(), the state is
duplicated. At this point in time, the in_fence_fd may be invalid because it was
replaced in a new commit of the corresponding surface.
The plane state was already committed to the kernel when it is copied, so the
fence is no longer needed. So just clear it.
Signed-off-by: Michael Olbrich <m.olbrich@pengutronix.de>
With multiple DRM devices, the state for one device may be empty during
repaint_flush(). This can happen for example if an output of one device triggers
the repaint and there are no screens attached to the other device and therefore
no active outputs.
The atomic commit will actually fail because the commit contains the
DRM_MODE_PAGE_FLIP_EVENT flag but no CRTCs.
Avoid this by skipping the commit entirly. There is nothing to to anyways.
Signed-off-by: Michael Olbrich <m.olbrich@pengutronix.de>
NVIDIA is more pedantic than Mesa and correctly complains that `1.0f`
is not valid syntax in the OpenGL ES Shading Language version 1.00.
And we are indeed using SL version 1.00 by virtue of using an ES 2.0
context.
So use the syntax compatible with the context we've created.
Signed-off-by: Daniel van Vugt <daniel.van.vugt@canonical.com>
The surface of the focus_view may be a subsurface. So get the corresponding main
surface first.
Without this get_ivi_shell_surface(), will cast committed_private (which is a
weston_subsurface object) to ivi_shell_surface. Weston crashes shortly
afterwards.
Signed-off-by: Michael Olbrich <m.olbrich@pengutronix.de>
With some displays connect, disconnect, connect events can happen is a very
short amount of time. When this happens, the output global may already be
destroyed when a client tries to bind it. As a result, the client is
disconnected with a protocol error. See [1] for more details on the general
problem.
To mitigate this problem call wl_global_remove() first and call
wl_global_destroy() several seconds later. This is inspired by the
implementation for the same problem in wlroots.
[1] https://gitlab.freedesktop.org/wayland/wayland/-/issues/10
Signed-off-by: Michael Olbrich <m.olbrich@pengutronix.de>
Input panels are used for complex text composition for CJK alphabets and for
onscreen keyboards. Support for this is already implemented in libweston and
the desktop shell.
This adds extends the IVI shell to add support for input panels as well. The
low-level parts are implemented in the IVI shell. The positioning of the input
panels is delegated to the controller.
Support for input panels and the relevant protocols is only enabled if the
controller attaches a listener to the new signals.
Signed-off-by: Michael Olbrich <m.olbrich@pengutronix.de>
This makes it possible for the controller to distinguish different surface
types. For IVI and desktop surfaces, this is not all that useful, but it will be
needed for a new surface type: input panels.
Signed-off-by: Michael Olbrich <m.olbrich@pengutronix.de>
shell_surface_send_configure() should only be called for ivi surfaces. Other (or
no) *configure() calls are needed for other types of surfaces.
Make sure that get_ivi_shell_surface() returns NULL if the surface is not an ivi
surface. And don't ignore wrong surfaces but assert instead to make it obvious
that new surface types need special handling in ivi_layout_surface_set_size().
Signed-off-by: Michael Olbrich <m.olbrich@pengutronix.de>
Since the logic of pointer constraints assumes a valid view throughout, add a
signal to disable constraints when its current view is unmapped by Weston.
The assumption that a previously unmapped view is valid already leads to the
constraints code crashing. This can happen when attaching a NULL buffer to the
surface and commiting, which effectively unmaps the view with the side effect of
clearing the surface's input region, which is then assumed valid inside
maybe_warp_confined_pointer().
Fixes: #721
Signed-off-by: Sergio Gómez <sergio.g.delreal@gmail.com>
Currently, the surface destroy listener in pointer constraints is redundant,
since surface destruction already handles pointer constraints destruction (see
libweston/compositor.c:weston_surface_unref()).
Signed-off-by: Sergio Gómez <sergio.g.delreal@gmail.com>
This makes flight recorder creation faster by using wider store
instructions (depending on the memset() implementation).
Signed-off-by: Loïc Molinari <loic.molinari@gmail.com>
This is a flag used to track whether the position has changed, not
whether the position is set.
Signed-off-by: Derek Foreman <derek.foreman@collabora.com>
This patch introduces a small library wrapper around XCB to be used in
Xwayland tests.
It's being designed such that we do not advance without accounting for
all X11 events when changing the window state. It adds a fence that
waits for all events to be processed, and only after all the events have
been accounted for, to proceed further, resuming execution of the
tests.
This works by keeping a tentative_state list for the client and a
window state that gets applied when the event we waited for has been
received.
This is useful in test clients, which could verify at the end after
receiving all events that the correct state has been applied. Acts as a
way to verify that the we never get or have a different state than the
one we expect.
With it, this converts test-xwayland to using libxcb (together with
xcb-cursor-dev) rather than using Xlib, and with it it removes any Xlib
dependency we might have in the tests.
This only adds support for map/unmap/create/destroy/property notify.
A follow-up would be to expand this library to track window movement
and resizing.
Signed-off-by: Marius Vlad <marius.vlad@collabora.com>
In future code the window link will end up in a state where it may or may
not be on the unpaired_window_list and we'll want to go from that state
to one where it's definitely not on the list.
Initting the list after removal (in these two places) allows us to
unconditionally remove it later.
Signed-off-by: Derek Foreman <derek.foreman@collabora.com>
Add the --additional-devices parameter to Weston to add secondary drm devices
that will only be used as outputs, but not for rendering.
We can only fail the repaint for the entire backend, but not for single
devices. Thus, if one of the devices fail, we have to fail the repaint for the
entire backend.
Signed-off-by: Michael Tretter <m.tretter@pengutronix.de>
Additional devices don't have a gbm device. Therefore, we cannot create gbm bos
for the cursor.
If the output device differs from the gbm device, fall back to the allocation of
a dumb buffer for the cursor on the output device. Update the cursor sprite with
a memcpy to the already mapped dumb buffer that belongs to the current cursor.
Signed-off-by: Michael Tretter <m.tretter@pengutronix.de>
If the GBM bo was allocated on a different device than the device that is used
for the fb, we have to import the fd first and update the handle.
Use drmPrimeFDToHandle directly instead of using a gbm device for the scanout
device, since a gbm device would require a gbm implementation, which is often
not available for devices that only support scanout.
Signed-off-by: Michael Tretter <m.tretter@pengutronix.de>
If we are using multiple GPUs and are not able to use modifiers to ensure that
the formats are compatible, we have to use linear buffers for the transfer.
Signed-off-by: Michael Tretter <m.tretter@pengutronix.de>
The hash table implementation is useful for other modules as well. Move it from
xwayland to the shared code.
Signed-off-by: Michael Tretter <m.tretter@pengutronix.de>
Weston uses a cached drm_fb when a view is shown multiple times. If the view is
shown on multiple outputs backed by different DRM devices, Weston returns the
cached drm_fb for the first device that was used for the import. This causes a
failure when adding the fb to the other device.
Use a list of all drm_fbs to cache the buf_fb per device, and check for the
device before reusing a drm_fb.
Signed-off-by: Michael Tretter <m.tretter@pengutronix.de>
The faked z position must be created for each device. Therefore, the device
itself must be passed to the function. If only the backend is passed, the faked
z position would be only created for the primary device.
Signed-off-by: Michael Tretter <m.tretter@pengutronix.de>
Now that struct weston_renderbuffer is refcounted, hold a reference for
renderbuffers on the pixman_output_state::renderbuffer_list. This allows
backends to destroy the renderer output state and release renderbuffer
references in any order without running into an assert().
To avoid breaking resizing, We also have to drop the renderbuffer list
during pixman_renderer_resize_output(). The backends have to create new
renderbuffers afterwards.
Signed-off-by: Philipp Zabel <p.zabel@pengutronix.de>
A config event with width == 0 or height == 0 from the shell is a hint
to the client to choose its own dimensions. Since X11 clients don't
support such hints we make a best guess by trying to use the last saved
dimensions or, as a fallback, the current dimensions.
This hint is mainly used by libweston/desktop shells when transitioning
to a normal state from maximized, fullscreen or after a resize [1].
Without support for this hint the aforementioned transition causes
xwayland surfaces to be configured to a 1x1 size.
To be able to use the last saved dimensions with xwayland surface, the
shell needs to first set the maximized/fullscreen state and only then
set the new size, which is currently the case for desktop-shell.
Otherwise, if the new size is set first, then the last saved dimensions
will be set to the fullscreen/maximized values and won't be useful when
restoring to a normal window size.
[1] Recently we've introduced ba82af938a
"desktop-shell: do not forget to reset pending config size after
resizes". As we were not handling the 0x0 size hint, resizing X
applications started to fail. This patch fixes that.
Signed-off-by: Alexandros Frantzis <alexandros.frantzis@collabora.com>
Signed-off-by: Leandro Ribeiro <leandro.ribeiro@collabora.com>
Co-authored-by: Leandro Ribeiro <leandro.ribeiro@collabora.com>
Align the function name and arguments of vnc_convert_damage() with
weston_region_global_to_output(). It does not support rotation and
stores the result in a pixman_region16_t, but otherwise it serves the
same purpose.
Signed-off-by: Philipp Zabel <philipp.zabel@gmail.com>
I also snuck in a trivial change to drag_surface_configure at the same
time to avoid yet another micro patch.
Signed-off-by: Derek Foreman <derek.foreman@collabora.com>
