Bug#1109275: unblock: mutter/48.4-2

[Simon McVittie]

Package: release.debian.org
Severity: normal
X-Debbugs-Cc: mutter at packages.debian.org
Control: affects -1 + src:mutter
User: release.debian.org at packages.debian.org
Usertags: unblock

I would like this to be considered for Debian 13.0 if possible, but if 
we're too late for that, it can be deferred to 13.1 by converting this 
request into a trixie-pu request.

[ Reason ]

New upstream bugfix release.

[ Impact ]

Contains various targeted bug fixes. In this particular release, most of 
them are for crashes. In Debian, these crashes mainly impact 
gnome-shell, in which they would normally (when using Wayland) be 
immediately fatal to the desktop session, causing data loss.

Another notable change in this release is that visual alerts ("visual 
bell") are now rate-limited to 2 per second. I'm told this is required 
by the European Accessibility Act for products sold in the EU, because 
the default implementation of visual alerts is to flash the entire 
screen, which can trigger photosensitive seizures if done rapidly and 
repeatedly.

[ Tests ]

I'm using the proposed version on my laptop. A functionally equivalent 
version was in experimental for 10 days without regressions reported.

I've confirmed that the visual-bell rate limiting works.

mutter has build-time tests and an autopkgtest, which can't cover 
everything for GUI software but are better than nothing.

[ Risks ]

This is a core component of our default desktop environment.

A Wayland compositor or X11 compositing manager is a complicated thing 
and regressions are possible, but the upstream project is generally good 
about judging risks and fixing regressions promptly. We have updated 
mutter in stable point releases in the past.

The changes are all narrowly-targeted and seem proportionate to the 
issues being fixed.

The only packaging changes on the Debian side are switching between 
branches, to accommodate GNOME 49 development having begun.

The test suite has been somewhat flaky in the past (with spurious 
failures that require a retry). The most recent release in trixie seems 
to have addressed the most common cause of this, but I did see some test 
failures on ppc64el, which were only observed once and disappeared when 
retried. I think they are all race conditions in the test infrastructure 
rather than serious problems with mutter itself (two in mutter's test 
suite and one in python3-dbusmock) and I've reported them at important 
severity. I expect them all to be difficult to reproduce.

[ Checklist ]

  [x] all changes are documented in the d/changelog
  [x] I reviewed all changes and I approve them
  [x] attach debdiff against the package in testing

unblock mutter/48.4-2