+1 maintenance report, July 10th July 18th

[Andreas Hasenack]

I had many high priority interrups in the original week of the +1
maintenance shift. Also had a patch pilot afternoon, and the full-day
SRU shift, so that didn't leave many days for +1 maint. I tried to
compensate by doing something in the weekend and this week until
today.

# fenics-dolfinx
ftbfs on amd64
doxygen is core dumping
Filed: https://bugs.launchpad.net/ubuntu/+source/doxygen/+bug/2026834
and https://bugs.debian.org/cgi-bin/bugreport.cgi?bug=1040864
TL;DR: upstream cairo changed how PDFs are produced, and that broke
doxygen. New upstream doxygen 1.9.7 has a series of fixes, but Ubuntu
and Debian are still at 1.9.4.
I have a WIP branch on top of 1.9.4, but there are a ton of patches
needed. It looks risky. I explained as such in the debian bug and the
maintainer agreed to update to 1.9.7, and that he would take care of
it. I try to refrain from going ahead of debian because of the orig
tarball generation, which, due to compression, ends up having a
different hash each time.

We could still go ahead of debian if this update takes too long. I
have a WIP branch that needs some polishing:
https://code.launchpad.net/~ahasenack/ubuntu/+source/doxygen/+git/doxygen/+ref/mantic-update-doxygen

# unicorn ftbfs
Fix: https://yhbt.net/unicorn-public/20230620104620.54277-1-bofh@yhbt.net/T/#u
Filed: https://bugs.launchpad.net/ubuntu/+source/unicorn/+bug/2027865
https://bugs.debian.org/cgi-bin/bugreport.cgi?bug=1041222

Uploaded to ubuntu, and proposed PR in Salsa. Eventually debian merged
the PR and uploaded, and I synced it into mantic. It's currently in
migration.

# rspamd segfaulting on armhf only
This took quite a while to troubleshoot. I don't have a clear cause
for the crash, but came up with some alternatives.
https://bugs.launchpad.net/ubuntu/+source/rspamd/+bug/2027879
I opted for disabling the symbolic-functions linker flag that Ubuntu
has, only for the armhf build. The four alternatives I had are shown
in the PR at https://code.launchpad.net/~ahasenack/ubuntu/+source/rspamd/+git/rspamd/+merge/446998
I also was never able to get a backtrace with symbols in the frame
that crashed, which is the statically linked jemalloc_pic library. And
upstream didn't show much interested in this crash case
(https://github.com/rspamd/rspamd/issues/4545).

It's in migration, and now I see that riscv64 is failing to build,
even though my change shouldn't have affected it (I have it ifdef'ed
on armhf).

Yep, that was it, sorry about this poor shift.