+1 maintenance report

[Olivier Gayot]

Hello,

I was on +1 maintenance this week (July 24th to July 28th) and here is a summary of the work I did:

# python-test-stages

Retried python-test-stages on all arches to unblock tox 4.4 - which has now migrated

# zulucrypt

Analysis:
 Investigated autopkgtest failures in the infrastructure that could not be reproduced locally with qemu.
 The error message made it look scarier that it really was. strace helped to notice open() calls resulting
 in EACCES. It turned out that for some reason, zulucrypt attempts to setuid(SUDO_UID) when the SUDO_UID
 variable is defined. This resulted in errors in the test-suite for virt-servers that use sudo.
Actions:
 * Opened a bug report upstream to get the root cause addressed 
 * Worked around the issue in Ubuntu by un-setting the SUDO_UID variable when running autopkg tests.
   Thanks slyon for sponsoring!
 * Forwarded the delta to Debian.
Status:
 * The package migrated.
 * Waiting for feedback from upstream
 * Bug report https://launchpad.net/bugs/2023614

# python-pytest-trio

Retried autopkgtest with additional triggers to allow it to migrate alongside python-trio. Both packages migrated.

# onednn

Analysis:
 Investigated arm64 build failure with GCC-13. Found an existing bug report upstream and the issue
 was addressed in onednn 3.1.
Actions:
 * Backported the relevant 3.1 changes to onednn 2.7 in Ubuntu. Thanks tsimonq for sponsoring!
 * Forwarded the delta to Debian.
Status:
 * The package migrated.
 * Bug report https://launchpad.net/bugs/2028759

# libprotocol-http2-perl

Analysis:
 Investigated test-suite failure in Ubuntu - that would run fine in Debian
 Lack of verbosity in the test-suite made the error message hard to work with.
 Added more verbose logs and moved error messages to stderr to get them printed when running autopkgtest
 It turned out that the test-suite has tlsv1 hardcoded (which is rejected by OpenSSL seclevel 2 in Ubuntu).
Actions:
 * Submitted a debdiff in Ubuntu
 * Forwarded the delta to Debian
 * Opened a PR upstream - which removes the hardcoded tlsv1 - and ensures more verbose errors
   are displayed.
Status:
 * Waiting for feedback from upstream. I noticed too late an existing PR addressing the issue in
   a slightly different way.
 * Debian picked up our fix and released 1.10-3 - although they are not yet affected. Thanks Gregor
   for the quick reaction.
 * I unsubscribed sponsors. The package should auto-sync and hopefully migrate over the week end.
 * Bug report https://launchpad.net/bugs/2023586

# maxima

* Investigated FTBFS issue on ppc64el. Tried a few things (see below) but didn't manage to obtain a readable
  stack trace from the src/binary-gcl/maxima binary:
  + tried setting DEB_BUILD_OPTIONS="nostrip debug"
  + tried setting CFLAGS=-g
  + tried replacing EXECUTEGCL = $(GCL_NAME) with EXECUTEGCL = $(GCL_NAME) -keep -g
  + tried replacing the /usr/bin/strip binary by a symlink to /bin/true
* Sadly none of it had an effect on the resulting binary - and I didn't have enough time to pursue the investigation.
* Bug report https://launchpad.net/bugs/2024061

# misc

I opened a MP against ubuntu-archive-tools to make retry-autopkgtest-regressions and find-proposed-clusters
automatically default to the current devel series. It uses python3-distro-info internally and falls back to
a hard-coded value (bumped from lunar to mantic) if the distro_info module is not installed
https://code.launchpad.net/~ogayot/ubuntu-archive-tools/+git/ubuntu-archive-tools/+merge/447971

Thanks,
Olivier