from August 30 - September 3. I intended to write the report for the
first shift in-between, but got caught up in glibc migration things.
So below are the packages I looked at during both shifts.
delve
=====
There was an autopkgtest failure on arm64. I sync'd 1.7.0-3 from
Debian which skipped the failing test.
ppp/freedombox
==============
ppp was unable to migrate due to an autopkgtest failure in freedombox
and it making some VPN plugins non-installable. I found that the
upload of ppp 2.4.9 started a transition of sorts, and several
packages needed no-change rebuilds in order to pick up the new upper
limit on ppp. Once these were uploaded, ppp was able to migrate.
pcre2/mariadb-10.5
==================
pcre2 was unable to migrate due to an autopkgtest failure in
mariadb-10.5. I found the upload of pcre2 10.37-0ubuntu1 needed a
transition and libpcre2-posix2 needed to be renamed libpcre2-posix3.
I uploaded pcre2 with these changes, and then no-change rebuilds of
hardlink and mariadb-10.5.
curl/diffutils/cod-utils
========================
Both curl and diffutils were blocked by an autopkgtest on ppc64el only
in cod-utils. mwhudson confirmed that the autopkgtest had regressed
in -release so I prepared a force-reset-test hint MP, but after
discussing with bdmurray we decided to add cod-utils/ppc64el to
big_packages instead, and the test succeeded.
openmpi
=======
I merged openmpi from Debian unstable as there were several bug fixes
between 4.1.0-7 and 4.1.0-10.
git-remote-hg
=============
The autopkgtests failed on all architectures. I sync'd 1.0.2.1~ds-3
which added a missing autopkgtest dependency.
python-biopython
================
The autopkgtests were failing due to python3-renderpm being missing.
Seeing there was already a Suggests on python3-renderpm, I upgraded it
to Depends.
ffmpeg/mat2
===========
ffmpeg was unable to migrate due to failing autopkgtests in mat2.
Debian had a new upstream version of mat2 where the expected test
output had been updated for ffmpeg >= 4.4, which I cherry-picked.
mpi4py/openmpi/pmix
===================
One of mpi4py's autopkgtests started to fail after the new version of
pmix had been sync'd. This prevented openmpi from migrating. The
same happened in Debian and I sync'd mpi4py 3.0.3-10 where this test
had been disabled after discussion with upstream.
stylish-haskell/haskell-ghc-lib-parser
======================================
I revisited this one from my last +1 after vorlon's comment in LP:
#1937070. I managed to get haskell-ghc-lib-parser to build on arm64
and armhf by reducing the memory requirement for the build. Now
stylish-haskell's binaries only need to be removed on riscv64 and
s390x then it can migrate.
open-ath9k-htc-firmware
=======================
open-ath9k-htc-firmware FTBFS with binutils 2.36 and gcc 11. I found
a new version in Debian's NEW queue and cherry-picked the patch.
sphinx/listparser
=================
sphinx is unable to migrate due to a failing autopkgtest on armhf. I
found that this test only passed once during the hirsute cycle and has
never passed on any other architecture. I prepared a force-reset-test
hint MP.
intel-ipsec-mb
==============
The upload of intel-ipsec-mb started a transition. I uploaded
no-change rebuilds of stress-ng and dpdk.
libpam-alreadyloggedin
======================
libpam-alreadyloggedin FTBFS with gcc 10.3 when it was imported from
Debian, and now with gcc 11. I found LP: #1935081, tagged it
'update-excuse', tested hellsworth's patch and uploaded.
python-xarray
=============
python-xarray had failing autopkgtests on arm64 and armhf. The new
version in Debian had the same problem and I cherry-picked the patch
which skipped the failing test.
ensmallen
=========
I also revisited this one from my last +1. In the meantime, 2.17.0-2
had been uploaded to Debian which fixed the FTBFS. I sync'd that, but
then had to fix a problem with glibc 2.34. Then there were failing
autopkgtests on arm64 and ppc64el, which were solved by armadillo
1:10.6.2+dfsg-1, so I sync'd that too.
dxvk
====
The latest versions from Debian were failing their autopkgtest on i386
where they passed previously. It turned out that Debian had fixed the
location where files were installed on i386, and this broke the
autopkgtest in Ubuntu where we were incorrectly testing the files from
amd64. I modified the autopkgtest dependencies so that the correct
packages were installed on amd64 and i386. I haven't forwarded this
to Debian yet.
libxml2/ruby-nokogiri
=====================
A previous libxml2 sync caused autopkgtest failures in ruby-nokogiri.
bdmurray sync'd a newer libxml2 from Debian which included a patch
fixing part of the problem, but there were still 3 failing tests.
Debian had a new version of ruby-nokogiri in experimental, but due to
its large number of reverse-dependencies, I decided to cherry-pick the
3 fixes from upstream instead of sync'ing it.
nagios-plugins-contrib
======================
The check_ssl_cert test started to fail due to upstream changes in how
it worked behind a proxy. I found the simplest change was to check
the certificate of launchpad.net instead of www.debian.org during the
test.
--
ubuntu-devel mailing list
ubuntu-devel@lists.ubuntu.com
Modify settings or unsubscribe at: https://lists.ubuntu.com/mailman/listinfo/ubuntu-devel
