Pipewire 0.3.52 via commit a1f33a99 introduced a change that affects how
long pipewiresrc holds onto the pw_buffers it dequeued. Before that
change the pw_buffer was held until the end of the videoconvert element
at the beginning of the pipeline. After that change the pw_buffer was
held onto until the filesink at the end of the pipeline. This was
starving MetaScreenCastStreamSrc of pw_buffers to record new frames
into, resulting in the majority of frames being missed, especially in
situations in which the encoder was taking longer.
Pipewire 0.3.57 via commit 1ea1d525 will allow queuing the pw_buffer
early again via the `always-copy` option. This however is only a
workaround until a proper solution is found in either pipewire or
gstreamer that does not depend on copying the buffer contents and
instead queues the pw_buffer again after videoconvert as prior to
a1f33a99.
Since it is not clear yet when a proper solution will arrive, this makes
use of `always-copy` as a workaround for now and should be reverted once
it is no longer needed.
Related: https://gitlab.gnome.org/GNOME/gnome-shell/-/issues/5585
Related: https://gitlab.freedesktop.org/pipewire/pipewire/-/issues/2461
Related: https://gitlab.freedesktop.org/gstreamer/gstreamer/-/issues/283
Part-of: <https://gitlab.gnome.org/GNOME/gnome-shell/-/merge_requests/2436>
The object the SignalTracker belongs to is stored in a map managed by
the SignalManager which keeps a reference to that object. This map is
never destroyed nor is any entry ever removed. This leads to all objects
that ever had SignalTrackers used on them being kept alive even after
all references outside of the SignalTracker are long gone. This then
also extends to other objects which are leaked indirectly through
reference chains from these objects.
And if some of those objects are GObjects, this will prevent them from
being finalized, leaking further resources. A StWidget for example will
not release its shadow textures.
Fix this by destroying the SignalTracker and removing it from the
SignalManager once the last signal it was tracking has been untracked.
A WeakMap could have been used as well, but we need the Map to be
iterable in some of the following changes.
Closes: https://gitlab.gnome.org/GNOME/gnome-shell/-/issues/5807
Closes: https://gitlab.gnome.org/GNOME/gnome-shell/-/issues/5796
Part-of: <https://gitlab.gnome.org/GNOME/gnome-shell/-/merge_requests/2466>
After the next commit, when some classes, such as PopupMenuManager try
to disconnect via a destroy handler, the SignalTracker might have
already been destroyed, so trying to get it from the SignalManager will
cause it to create a new one, which will then try to connect to the
destroy signal of the already destroyed object.
This could for example be triggered by changing backgrounds.
Fix this by not doing anything in disconnectObject if there is no
SignalTracker for that object.
Part-of: <https://gitlab.gnome.org/GNOME/gnome-shell/-/merge_requests/2466>
This reverts commit 085102be74.
We need the SignalManager map top be iterable at shutdown for some of
the following changes. A WeakMap is not iterable. This revert changes it
back to a regular Map, which re-introduces the leaks caused by this.
Those will be fixed differently by the two followup commits.
Part-of: <https://gitlab.gnome.org/GNOME/gnome-shell/-/merge_requests/2466>
Commit 9168f6055e marked the parameter as nullable, which is in
conflict with the precondition check. But given that NULL is
valid for the underlying ClutterText (and GtkLabel accepts it
as well), there's no strong argument for disallowing NULL, so
remove the precondition check.
Part-of: <https://gitlab.gnome.org/GNOME/gnome-shell/-/merge_requests/2465>
