The second patch needs an edit because part of it doesn't apply to the
actual source (4.5.0 doesn't contain 427793aa88fb57 which applied the
@flaky markers to the test). But quodlibet builds again on my machine
with this second patch.

Le Fri, 5 May 2023 15:31:42 +0200,
Alice BRENON <alice.brenon@ens-lyon.fr> a écrit :

> This problem seems to have been reported and fixed upstream:
> 
> https://github.com/quodlibet/quodlibet/commit/5f55431a28509fd4f4f7b40dc246f3d34fa8549e
> 
> Since no newer release has been out, I guess we can backport the
> commit itself. I've tried with the attached patch. The operon tests
> still fail due to some weird utime magic, a problem also known and
> discussed at:
> 
> https://github.com/quodlibet/quodlibet/pull/4053/commits/06a32b319f065550efe0d2a9ff10ca6bdc32b893
> 
> So I guess we could backport that as well. Does anyone have a better
> fix ?