On Sat, 24 Jan 2026 at 17:47:54 +0200, Nazarii Kretovych wrote:
I've noticed that the official Git repository of the "mutter" package has a
commit that could be related to the issue. I assume that commit was not
included in the current mutter package Debian Trixie has because it was created
after the commit with tag 48.7.
[1]https://gitlab.gnome.org/GNOME/mutter/-/commits/gnome-48
Commit message:
wayland/dma-buf: Fix direct scanout on drivers without explicit modifiers.
My reading of that commit is that it fixes a less serious regression and
is not directly related to this one. "Direct scanout" is a performance
optimization: while a game, video player or similar app is running in
fullscreen, instead of compositing several windows to produce the final
image, gnome-shell/mutter will arrange for the pixels from the app to
bypass the compositor and be drawn directly to the screen. I believe the
regression fixed by that commit is that fullscreen games didn't trigger
direct scanout as they should have done, so the frame was delivered to
the screen less efficiently (but you would still see the correct
pixels).
Also, that commit says the direct scanout functionality regressed in
mutter 48.2, which means it was already unavailable (on the affected
GPUs) before Debian 13.0, and the upgrade to Debian 13.3 wouldn't have
affected it. In your bug report you said that the versions of
gnome-shell and mutter in Debian 13.2 worked correctly for you, and the
regression only appeared in Debian 13.3.
If you want to try rebuilding mutter with that change cherry-picked,
please do, but I don't think it will do anything to resolve the
regression reported here. I'd prefer it if anyone who can reproduce this
regression could prioritize testing
<https://people.debian.org/~smcv/temp/2026/bug1125941/revert/> first, so
that I can start the process of getting that change into Debian 13.4.
I don't have hardware that is affected by this issue (I do have some
systems with Radeon GPUs but they still seem to be working fine), so I
have to rely on users with affected hardware to confirm whether the
proposed change fixes it.
Thanks,
smcv
