There might not be a single plane that is "for" a CRTC, so remove the
API that made it appear as if it did. The existing users only cared if
there was some plane for said CRTC, so replace the getters with API that
just checks the existance at all.
Part-of: <https://gitlab.gnome.org/GNOME/mutter/-/merge_requests/3428>
When there are a set of primary planes, and a set of CRTCs, where each
plane can be used on multiple CRTCs, we need to make sure that when we
mode set and page flip, each CRTC gets assigned an individual plane that
isn't used with any other CRTC.
Do this during the monitor resource assignments that sets up later to be
applied configurations of the mode setting objects, but with the new
hooks into the backend, so that we don't need to teach the monitor
config manager about planes.
Closes: https://gitlab.gnome.org/GNOME/mutter/-/issues/2398
Part-of: <https://gitlab.gnome.org/GNOME/mutter/-/merge_requests/3428>
This means that it doesn't necessarily mean what transform / rotation
the hardware resource gets, e.g. it instead represents the logical
transform related to the configured mode. This allows us to postpone
checking the plane capabilities until later (as rotation capabilities
depends is a plane property), when a plane has been assigned.
This was in practice already handled when configuring the
transform-via-offscreen case, handled when creating the view, and the
mode setting configuration.
Part-of: <https://gitlab.gnome.org/GNOME/mutter/-/merge_requests/3428>
When we're configuring monitors, allow backends to add backend specific
assignments during resource assignment (mapping connectors and CRTCs
etc).
This will later allow the native backend's KMS monitor resources to
assign a primary plane and optionally a cursor plane during
configuration. This will then dictate what plane will be used for
primary plane updates, as well as cursor updates, until reconfigured
again.
Part-of: <https://gitlab.gnome.org/GNOME/mutter/-/merge_requests/3428>
Musl doesn't seem to include this by default so explicitly including it
should fix compilation on Musl.
Tested with Clang 16/17 and GCC 14.
Error:
src/backends/meta-fd-source.c:70:3: error: call to undeclared function 'close'; ISO C99 and later do not support implicit function declarations [-Wimplicit-function-declaration]
close (fd_source->poll_fd.fd);
^
Closes: https://gitlab.gnome.org/GNOME/mutter/-/issues/3078
Part-of: <https://gitlab.gnome.org/GNOME/mutter/-/merge_requests/3455>
Otherwise a tablet in relative mode will never have a tool set and
nothing happens on motion events - meta_wayland_tablet_seat_update()
simply exits early for tablet proximity, button or motion events.
Part-of: <https://gitlab.gnome.org/GNOME/mutter/-/merge_requests/3410>
This is a step in cleaning up the Clutter context management. By making
it a GObject it's easier to add e.g. properties and features that helps
with introspection.
For now, this means the context creation is changed to go via a
"constructor" (clutter_create_context()). This is so that the global
context singleton can be mantained outsid of ClutterContext, until it
can be removed.
Part-of: <https://gitlab.gnome.org/GNOME/mutter/-/merge_requests/2016>
The original purpose of being able to report errors is no longer
relevant, since the Clutter backend is now practically a thin wrapper
around the actual backend, which has already dealt with error reporting.
Thus move this to the regular constructor path.
Part-of: <https://gitlab.gnome.org/GNOME/mutter/-/merge_requests/2016>
Since meta_kms_impl_device_get_sync_file always returns the same
file descriptor referencing the same sync_file, this means the atomic
ioctl doesn't need to wait for any fences to signal. This is fine
because we already waited for the buffer to become idle before applying
the Wayland surface state.
Fixes the atomic commit ioctl spuriously synchronizing to the screen
cast paint (at least with the amdgpu driver), which could result in
the page flip missing its target scanout cycle.
Closes: https://gitlab.gnome.org/GNOME/mutter/-/issues/3148
v2:
* Rename local variable to signaled_sync_file for consistency with new
function name
Part-of: <https://gitlab.gnome.org/GNOME/mutter/-/merge_requests/3389>
It returns a file descriptor which references a signaled sync_file.
v2:
* Change function name and add Doxygen comment to hopefully make its
purpose a bit clearer (Ivan Molodetskikh)
v3: (Jonas Ådahl)
* Create sync_file from scratch via a syncobj, no buffer needed anymore
* Initialize priv->sync_file = 1 and use g_clear_fd in finalize
Part-of: <https://gitlab.gnome.org/GNOME/mutter/-/merge_requests/3389>
While we should ideally have a sensible cursor theme, handle the
case of cursor themes that lack certain cursor names, and fallback
to the 'default' cursor in those cases.
The 'grey rectangle' fallback is still left, in case we even fail
to load a 'default' cursor.
Part-of: <https://gitlab.gnome.org/GNOME/mutter/-/merge_requests/3295>
When a stream is destroyed by a consumer, mutter won't be able to
recognize that.
For mutter, the stream just paused, but did not disconnect, because the
connection state of a PipeWire stream only represents, whether the
respective PipeWire context is connected to PipeWire.
In addition to that, it may be the case, that the stream consumer just
recreates the stream.
So even if mutter would be able to know, when the stream consumer
destroyed a stream, but not the whole screencast or remote-desktop
session, then mutter would not know, whether the stream will be resumed
eventually or not.
So, add an explicit API call to the screencast interface to stop a
stream.
For virtual streams, this also means, that the respective virtual
monitor is destroyed.
Closes: https://gitlab.gnome.org/GNOME/mutter/-/issues/2889
Part-of: <https://gitlab.gnome.org/GNOME/mutter/-/merge_requests/3307>
When a virtual stream is destroyed, its respective virtual monitor is
destroyed too. When the virtual monitor is destroyed, mutter reloads
the monitor manager.
However, at this point, the virtual stream is not completely destroyed
yet. The viewport of the virtual monitor still exists at this point and
when the monitor manager reloads, it will try to fetch the logical
monitor of the now destroyed virtual monitor, which will fail and thus
gnome-shell will run into a segfault.
Fix this situation by reloading the monitor manager in an idle callback.
When the monitor manager reloads, the virtual monitor is completely
gone, since the viewport of the virtual monitor is destroyed after the
virtual monitor itself.
Closes: https://gitlab.gnome.org/GNOME/mutter/-/issues/2864
Part-of: <https://gitlab.gnome.org/GNOME/mutter/-/merge_requests/3307>
There is no way to set any gamma luts, or do anything other color
management related. Eventually we'll probably want to, but that requires
bringing color management plumbing to PipeWire.
Doing this is also needed when running a headless session, as when
headless, polkit doesn't let us create colord devices without explicit
user permission, meaning we'll spam the session with useless dialogs
each time a session is started.
Part-of: <https://gitlab.gnome.org/GNOME/mutter/-/merge_requests/3423>
`count_mode_setting_devices` was incorrect in both name and in function.
What it was actually doing was counting GPUs that had been registered with
the backend so far (during the `init_gpus` loop). What it was intended to
do was to count the number of `MetaRenderDeviceEglStream` instances, which
is the thing we're limited to only one of. So `count_mode_setting_devices`
would return zero whenever the first GPU initialized happened to be a
`MetaRenderDeviceEglStream`, which would in turn prevent
`MetaRenderDeviceEglStream` from successfully initializing. Seems it only
ever worked in the case of a hybrid system where the first GPU initialized
was GBM-based.
Now we count `MetaRenderDeviceEglStream` instances (zero or one) externally.
This allows initialization to succeed when it happens to be the first (or
only) GPU. And so `MUTTER_DEBUG_FORCE_EGL_STREAM=1` now works.
Part-of: <https://gitlab.gnome.org/GNOME/mutter/-/merge_requests/2905>
For scanout on a secondary GPU, for the time being try only formats
which are guaranteed to be renderable with GLES3, which notably excludes
10 bpc formats without alpha channel.
v2:
* Use separate format array for 10 bpc formats without alpha.
Part-of: <https://gitlab.gnome.org/GNOME/mutter/-/merge_requests/3139>
If the EGL_KHR_no_config_context extension is supported, use it to
choose a format per onscreen which is compatible with the scanout CRTC
and the GL rendering API used.
Suggested by Jonas Ådahl.
v2:
* Drop code which checked for GLES3 renderability. Makes no sense for
various reasons, in particular that EGLconfigs are about EGLSurfaces,
whereas secondary GPU contexts use an FBO for blitting.
* Use error parameter directly for meta_renderer_native_choose_gbm_format
call (Jonas Ådahl)
Part-of: <https://gitlab.gnome.org/GNOME/mutter/-/merge_requests/3139>
Preparation for the following commits, no functional change intended.
v2:
* Pass through MetaEgl pointer
v3:
* Make it return gboolean (Robert Mader)
v4:
* Add debug logging and corresponding purpose parameter
v5:
* Fix excessive function parameter indentation (Jonas Ådahl)
Part-of: <https://gitlab.gnome.org/GNOME/mutter/-/merge_requests/3139>
In profilers with a timeline or flame graph views it is a very common
scenario that a span name must be displayed in an area too short to fit
it. In this case, profilers may implement automatic shortening to show
the most important part of the span name in the available area. This
makes it easier to tell what's going on without having to zoom all the
way in.
The current trace span names in Mutter don't really follow any system
and cannot really be shortened automatically.
The Tracy profiler shortens with C++ in mind. Consider an example C++
name:
SomeNamespace::SomeClass::some_method(args)
The method name is the most important part, and the arguments with the
class name will be cut if necessary in the order of importance.
This logic makes sence for other languages too, like Rust. I can see it
being implemented in other profilers like Sysprof, since it's generally
useful.
Hence, this commit adjusts our trace names to look like C++ and arrange
the parts of the name in the respective order of importance.
Part-of: <https://gitlab.gnome.org/GNOME/mutter/-/merge_requests/3402>
Now that the monitor screencast records to DMA-BUF buffers immediately
(since bc2f1145d8), and we know which phase of the paint rountines we
are (since last commit), we have the opportunity to bring back the
blitting technique.
Bring back blitting. This time, instead of simply failing if the blit
fails, add a fallback path that does a stage paint if something goes
wrong. Unlike the previous implementation of blitting, this one only
blits the current view - it does not blit all views that intersect
with the screencasted monitor.
Embedded cursors should still be fine because hardware cursor is
inhibited while embedded cursor screencasts are running.
Part-of: <https://gitlab.gnome.org/GNOME/mutter/-/merge_requests/3406>
Track where we are in terms of the paint cycle. Do this through an
enumeration that is passed through the paint vfuncs of screencast
sources.
Right now, this information is not used by any one of the sources,
but next patch will use it to prevent blitting when detached from
the paint cycle.
Part-of: <https://gitlab.gnome.org/GNOME/mutter/-/merge_requests/3406>
With the existing ClutterInputMode terminology (inherited from XI2),
hardware devices may be "physical" (i.e. attached to a logical device),
or "floating" (i.e. detached from all logical devices).
In the native backend, tablet devices are closer to "floating" than
"physical", since they do not emit events relative to the ClutterSeat
pointer logical device, nor drive the MetaCursorTracker sprite. This
is in contrast to X11 where all tablet devices drive the Virtual
Core Pointer by default, along with every other pointing device.
Change this mode in the Wayland backend to be more coherent. The
existing checks on the ClutterInputMode along Mutter seem appropriate
for handling these as floating devices, since they mainly care about
logical vs non-logical.
Part-of: <https://gitlab.gnome.org/GNOME/mutter/-/merge_requests/3393>
Scoped traces are less error prone, and they can still be ended
prematurely if needed (this commit makes that work). The only case this
doesn't support is starting a trace inside a scope but ending outside,
but this is pretty unusual, plus we have anchored traces for a limited
variation of that.
Part-of: <https://gitlab.gnome.org/GNOME/mutter/-/merge_requests/3396>
Allow only specific files to use those deprecated APIs making
it easier to find where deprecated APIs are still in use
and avoid introducing new usages without being noticed
Part-of: <https://gitlab.gnome.org/GNOME/mutter/-/merge_requests/3400>
Group all the three config files from clutter/cogl/meta into one
and also remove unnused configurations and replace duplicated ones
This also fixes Cogl usage of HAS_X11/HAS_XLIB to match the expected
build options
Part-of: <https://gitlab.gnome.org/GNOME/mutter/-/merge_requests/3368>
Sticky keys configuration changes reset the pressed modifier state mask,
even though the XKB state might already match with the expected new
state. In those cases we can avoid the XKB state mask update completely.
This also fixes a crash at initialization with sticky keys toggled on,
since configuring the device a11y settings will trigger a XKB state
mask merely reassuring the initial state with no modifiers pressed,
while the connection between the ClutterSeat and the impl object has
not been set up yet. This crash was introduced by commit 00bb4190b
("backends/native: Drop device_native->seat_impl field").
Part-of: <https://gitlab.gnome.org/GNOME/mutter/-/merge_requests/3392>
This struct contains the pressed/latched/locked set of modifiers applying
to the event, and may be filled in by backends generating those events.
Other places where we forward modified key events, state may be normally
obtained from the original event.
Since this constructor is used in a variety of places, this commit
updates them all in one go.
Part-of: <https://gitlab.gnome.org/GNOME/mutter/-/merge_requests/3369>
- Make Texture a parent GObject class and move the vtable funcs as vfuncs
instead of an interface as we would like to have dispose free the TextureLoader.
- Make the various texture sub-types inherit from it.
- Make all the sub-types constructors return a CoglTexture instead of their respective
specific type. As most of the times, the used functions accept a CoglTexture,
like all the GTK widgets constructors returning GtkWidget.
- Fix up the basics of gi-docgen for all these types.
- Remove CoglPrimitiveTexture as it is useless: It is just a texture underhood.
- Remove CoglMetaTexture: for the exact same reason as above.
- Switch various memory management functions to use g_ variant instead of the cogl_ one
Note we would still want to get rid of the _cogl_texture_init which is something
for the next commit
Part-of: <https://gitlab.gnome.org/GNOME/mutter/-/merge_requests/3193>
Create a fake monitor region right of the right-most monitor and if a
horizontal barrier extends into that region, fail the barrier. Barriers
are aligned on the top/left edge of the pixel so the most natural
barrier of (e.g. 0-1024) is also wrong - it's one pixel into the next
monitor.
Check this for nonexisting screens on the right too to avoid clients
suddenly failing when multiple monitors are present.
Part-of: <https://gitlab.gnome.org/GNOME/mutter/-/merge_requests/3319>
Assuming two 1920x1080 screens next to each other: a horizontal barrier
starting at 1920 going east is always outside the left screen.
Assuming two 1920x1080 screens on top of each other: a vertical barrier
starting at 1080 going south is always outside the top screen.
Part-of: <https://gitlab.gnome.org/GNOME/mutter/-/merge_requests/3319>
When running headless, only the invalid modifiers are advertised.
That breaks with the NVIDIA proprietary driver which then rejects the
buffers created with the invalid modifier, and that kills Xwayland,
meaning that running Xwayland on top of a mutter based compositor
headless is not possible.
The reason the modifiers are not sent is because AddFb2 is not supported
when running headless.
Other compositors (weston, wlroots) would still send the modifiers even
without AddFb2, and Xwayland works fine on those compositors when
running headless.
Remove the requirement for AddFb2 to send the modifiers, so that
Xwayland can work fine on top of mutter headless with the NVIDIA
proprietary driver.
Closes: https://gitlab.gnome.org/GNOME/mutter/-/issues/3060
Part-of: <https://gitlab.gnome.org/GNOME/mutter/-/merge_requests/3320>
Certain kernel drivers can take an unreasonably long time to
complete mode setting operations. That excessive CPU time is charged
to the process's rlimits which can lead to the process getting killed
if the thread is a real-time thread.
This commit inhibits real-time scheduling around mode setting
commits, since those commits are the ones currently presenting as
excessively slow.
Closes: https://gitlab.gnome.org/GNOME/mutter/-/issues/3037
Part-of: <https://gitlab.gnome.org/GNOME/mutter/-/merge_requests/3324>
At the moment if a thread is made real-time there's no going back,
it stays real-time for the duration of its life.
That's suboptimal because real-time threads are expected by RTKit to
have an rlimit on their CPU time and certain GPU drivers in the kernel
can exceed that CPU time during certain operations like DPMS off.
This commit adds two new ref counted functions:
meta_thread_{un,}inhibit_realtime_in_impl
that allow turning a thread real-time or normally scheduled. At the same
time, this commit stores the RTKit proxy as private data on the thread
so that it can be reused by the above apis.
A subsequent commit will use the new APIs.
Part-of: <https://gitlab.gnome.org/GNOME/mutter/-/merge_requests/3324>
Most of the code writes "real-time" as "realtime" not "real_time".
The only exception is one function `request_real_time_scheduling`.
This commit changes that function for consistency.
Part-of: <https://gitlab.gnome.org/GNOME/mutter/-/merge_requests/3324>
If we queued a mode set, but didn't end up compositing all frames, we'll
have pending mode sets in a hash table waiting to be applied. If we
before all monitors again try to reconfigure things we should drop the
old pending mode sets and start fresh.
We already do this when we're doing so when generating views, but when
just unsetting modes, we didn't, so fix that.
Related: https://bugzilla.redhat.com/show_bug.cgi?id=2242612
Part-of: <https://gitlab.gnome.org/GNOME/mutter/-/merge_requests/3318>
We had a function called "reset_modes()" on MetaRendererNative, but what
it expected to do was to unset all modes on all CRTCs. Despite this, it
had code to unset modes on unconfigured CRTCs, probably because it was
used for multiple things in the past.
Make this a bit easier to follow by renaming the function
"unset_modes()" and fold the function doing the unsetting into the
function itself.
Part-of: <https://gitlab.gnome.org/GNOME/mutter/-/merge_requests/3318>
