+1 maintenance report

[Simon Chopin]

Hi,

This is a short report of what I've been up to this week as part of my
+1 rotation. Oddly enough, I seemed to have picked time-consuming issues
to work on, which makes for a short status.

=======================================================================

Items requiring follow up:

* git-annex needs fresh eyes, it's decidedly non trivial
* rust-pangocairo and rust-graphene-sys will require a manual sync once
  processed on the Debian side

=======================================================================

### git-annex

Picked it up from bdrung's shift last week. I banged my head against it
for a day and a half with a little bit of progress, but no solve in
sight, see https://bugs.launchpad.net/ubuntu/+source/ghc/+bug/2019992
for more details. It might turn out to be an actual linker issue, or at
the very least a GHC one.

### matplotlib

This was more productive. The package was synced from Debian, then
dropped because it was holding up the python3.11 transition, then
re-synced, at which point the autopkgtests were still failing in Ubuntu
but passing in Debian.

It turns out to have been caused by our pkgbinarymangler PNG
optimization pass that changed the color encoding of the icons,
triggering a latent bug in the tk backend. There was an upstream fix for
it.

The fixed package has been uploaded to Ubuntu.

See https://bugs.launchpad.net/ubuntu/+source/matplotlib/+bug/2020666

### dh-cargo

The package had been stuck in -proposed for a while after my merge
during the Prague sprint, partly due to the sheer volume of
reverse-dependencies. I simply retried in bulk the remaining failing
tests, both against dh-cargo and migration-reference, the package
migrated shortly after that.

### r-base

The new upstream version was uploaded by mistake to Debian unstable
despite the freeze, and was subsequently autosynced. That triggered
quite a few autopkgtest regressions due to a change upstream on policy
regarding calling native symbols.

I simply documented the issue and moved on, as the R ecosystem is fairly
self-contained and we can reasonably expect Debian to sort out the issue
soon after their release.

See https://bugs.launchpad.net/ubuntu/+source/r-base/+bug/2020799

### rust-pangocairo / rust-gio / rust-gir-format-check

A bunch of GTK-related Rust packages were synced from Debian
experimental, presumably as dependencies for new versions of GNOME
applications and/or libraries.

rust-pangocairo and rust-gio autopkgtests are failing in the same way.
It turns out to be due to upstream unknowingly using a feature of the
dependency gir-format-check that's only available from the version 0.1.2
onwards, while Debian and Ubuntu only ship 0.1.1.

I've prepared the 0.1.3 for an upload in Debian and am currently
awaiting sponsorship within the Rust team for it. I didn't upload
directly to Ubuntu as I'd rather introduce as few deltas as possible in
that ecosystem, rather relying on syncs.

See https://bugs.launchpad.net/ubuntu/+source/rust-gio/+bug/2020880

### rust-graphene-sys

Also a part of the Rust GTK ecosystem. The tests were failing on armhf
and s390x, due to new tests checking alignment assumptions and those
assumptions being incorrect for armhf and s390x.

See https://bugs.launchpad.net/ubuntu/+source/rust-graphene-sys/+bug/2020902

### visual-excuses

To find on which package to work on, I used mclemenceau's visual-excuses
tool. Of course, I stumbled upon and fixed some small issues :)

https://github.com/mclemenceau/visual-excuses/pull/4

Cheers,
Simon