Foundations Team Updates - Thursday 02 March 2023

The previous status is here: Foundations Team Updates - Thursday 23 February 2023

1 Like


  • make check failed on ppc64el due to building with -O3. Changing to -O2 solved the problem LP: #2008914


  • non x86 architectures were removed. These need to be deleted from the archive. LP #2008913


  • building on am64 fails. Still analyzing


  • The Debian maintainer remove the s390x architecture though the change log says he wanted to keep it. - Currently building in ppa.


  • The package has a build dependency on dropped package nomad. It should be added to the sync-blacklist LP #2009046


  • The package has multiple function/methods that according to the C standard result in undefined behavior and a build failure with -O3 -Wall -Werror LP #2009054, upstream bug report Issue #132
1 Like




  • pulse planning
  • candidate interviews & take-home test reviews
  • 360 prep
1 Like
  • Finished compliance training
  • Merged python-formencode and sync’d sqlobject for compatibility with the latest python-pip
  • Fixed linking in tiledb to avoid tiledb-py FTBFS on riscv64
  • Proposed-migration of pillow: retried some autopkgtests and discovered an issue with python-fabio’s autopkgtests that are passing in lunar-release with the 5.19 kernel, but are often failing in lunar-proposed with the 6.1 kernel. Have alerted kernel team and added migrations hints for sphinx-rtd-theme, pyqt5 and h5py which were blocked by python-fabio’s apparent regression. @bdmurray suggested filing a bug against auto-package-testing, which I did (LP: #2009052).
  • Sponsored botan upload for @xypron
  • Candidate interview
1 Like
  • Fix fakeroot bug with time64 stat struct on 32 bit architectures (Uploaded NMU in Debian, Gianfranco has uploaded the patch in Ubuntu as well).
  • Learn some documents about building snap.
1 Like


+1 maintenance

  • python-omemo:
    • python-omemo build depends on python3-oldmemo and python3-twomemo
    • python-twomemo (build) depends on python3-omemo (>= 1~)
    • python-oldmemo (build) depends on python3-omemo (>= 1~)
    • Drop build dependencies on python3-oldmemo and python3-twomemo and skip tests to break circular dependencies. Rely on autopkgtest instead.
  • python-prometheus-client: jupyter-notebook/6.4.12-2.1 failed with timeout, retried test



  • Compliance training
1 Like


  • Prepared systemd SRU for Focal to fix:
    • LP 1933090 “systemd/245.4-4ubuntu3.6 ADT test failure with linux-hwe-5.11/5.11.0-20.21~20.04.1”
    • LP 2002445 “udev NIC renaming race with mlx5_core driver”
  • Working on systemd SRU for Jammy to fix:
    • LP 2002445 “udev NIC renaming race with mlx5_core driver”
    • LP 2004478 “systemd-networkd’s dhcp4 client ignores local subnet routes”
    • LP 2000880 “systemd-networkd: ActivationPolicy ignored in VLANs”
  • Working on systemd SRU for Kinetic to fix:
    • LP 2002445 “udev NIC renaming race with mlx5_core driver”
    • LP 2004478 “systemd-networkd’s dhcp4 client ignores local subnet routes”
  • Investigating bug reports


  • libio-socket-ssl-perl
    • This seems to be resolved with retries, but still waiting for a couple tests to finish.


  • Annual compliance trainings
1 Like
  • subiquity:
    • PR: 1578 (LP: #2008271, LP: #2008801) - an autoinstall failure had root cause in a keyboard layout name that was not valid, but the install proceeds pretty far before detecting that problem. Do some up-front validation of layout and variant.
    • Investigate autoinstall failure with ubuntu-desktop-installer where autoinstall wasn’t being handed over to subiquity. LP: #2008952 resulted from this analysis, looks like the fix will be in cloud-init.
    • WIP on hfs resize, will put new tool and modified packing in PPA
  • mini-iso-tools:
    • LP: #2008742 - start MIR
    • Working through feedback on ISO build in livecd-rootfs
    • PR: 10 - move build to meson, which makes it really easy to add test coverage reports and a few other nice-to-haves
    • PR: 11 - change ISO to show released images only. Handle the case, like on releases, where a single simplestream JSON may want to result in showing multiple ISOs.
    • WIP on ncurses menu improvements
  • proposed-migration: analyze regressions on packages that trigger with lsb, determined that most if not all of them are failures in the other packages. Retests with various triggers, a few with migration-reference/0
1 Like


  • added support for downloading packages and installing them in two separate steps: MR 437897 (merged)
  • merged MP 437494 to fix bug in curtin causing apt-config to be triggered unconditionally during curtin’s curthooks stage
  • filed LP: #2008909 to investigate potential issue when apt mirrors directives are conflicting.
  • opened MP 438190 to revert workarounds put in place because of above bug. (no urgency to merge)


  • opened PR 1575 to make subiquity use the new download-then-install options provided by curtin for installing packages. This makes subiquity able to retry the download operation on error ; backed by a FFe: #2008900 that was approved
  • merged PR 1565 to pass debconf-selections as part of curtin’s apt-config and not curtin’s curthooks.
  • fixed wrong source selected during partial autoinstall PR 1567 (merged)
  • made sure the mirror selection does not run apt-get update on host during CI, potentially causing timeouts PR 1574 (merged)
  • sped up integration tests by disabling tracemalloc by default: PR 1576 (merged)
  • fixed crash when adding a tiny disk or partition to a volume group PR 1577 (merged)
  • still working on making mirror testing honor apt settings as expected.


  • asked around for autopkgtest retries of pycirkuit to allow m4 to migrate. Migrated.
1 Like


  • Investigating sudo, curl and lsb proposed migration excuses


  • Fixed a bug where netplan was generating duplicated configuration in few situations, two reported on LP and 3 I’ve found during the investigation. (merged) (PR#329)
  • Fixed a bug that was breaking the dbus integration and improved the error propagation (PR#331)
  • Implemented some integration tests for the dbus integration
  • Code review (PR#332)
1 Like


  • trying to port faketime to 64-bit time_t on 32-bit arches; it’s a mess: symbols have changed, this needs a split-horizon handling, and glibc doesn’t expose the corresponding types or machinery but I’m making progress
  • crypto-configuration: spec not yet signed off but agreed on overall: this only needs tighter specs in some points
1 Like
  • Uploaded xlnx-firmware changes for Xilinx Versal support on lunar for @ethan.hsieh (LP: #2006104) (and prepped for SRU)
  • Spent the rest of the week attempting to stop flash-kernel from handling kernels before their initrd exists (LP: #2007827)
  • Take home test
  • Pi meetings
  • Yup … that’s it
1 Like


The biggest chunk of my week was spent chasing down @jawn-smith’s libarchive vs reprepro regression on s390x. I finally managed to nail down the issue (which had nothing to do with libarchive, incidentally). I opened a MR upstream which has just been merged.


  • Usual retries for glibc 2.37 in -proposed
  • Compliance training
  • Interviews & take-home tests
1 Like


  • Finished backporting Cargo 0.67 and Rust 1.66 to B, F, K and J (LP#2005122).
    • Proposed an SRU for LLVM 15 on J and K (LP#2008755).
  • Currently backporting Cargo 0.68 and Rust 1.67 to B, F, K and J (LP#2007642).


  • Finished proposing MIR for libdigest-md5-file-perl and libswitch-perl (LP#2007279)
1 Like
  • libhtml-parser-perl:
    • worked on blocked proposed migration (3.80-1 to 3.81-1), but regressions solved itself without any input from me.
    • closed LP #1074154, because the release to which the bug was reported for is no longer supported
  • glibc
  • dotnet:
    • I try to catch up what is happening there


  • completed compliance trainings
1 Like


  • Modified autopkgtest to install efibootmgr on arm64 for bionic to workaround an unapproved grub2 SRU.
  • Created an MP adding the new autopkgtest-cloud-workers to the mojo/service-bundle. This was manually done during the ceph disaster recovery.
  • Additionally modified the n-worker settings in the service-bundle to reflect the current situation in production.
  • On Sunday investigated a reported issue with the armhf queue being stuck. A couple of workers were down due to a full /srv partition, it holds the lxc containers, but the root cause was different than what we normally see. @paride investigated it further on Monday and discovered that it was due to a defunct process in the autopkgtest for meson which was somehow filling the partition.
  • Added meson/armhf/lunar to the never_run packages due to @paride’s investigation above.
  • Killed a bunch of defunct meson tests runs on the lxd-armhf workers.
  • Again updated the mojo/service-bundle for autopkgtest-cloud to reflect the new lxd-armhf workers which were reprovisioned as a part of troubleshooting the meson failure.
  • Added jsurf-alggeo to big_packages for arm64 and amd64.
  • Reviewed and merged an MP modifying t
    he files created by autopkgtest-cloud so that the summary file is included in result.tar and nothing is dropped from artifacts.tar.gz
  • Built the autopkgtest-cloud-worker charm with the above change and updated the staging autopkgtest environment to use charm revision
    148 from edge.
  • Brought down an autopkgtest-cloud-worker unit which was accidentally deployed when using mojo run in production.


  • Update meta-release and raspi imaging files in bzr and on changelogs.u.c for the 22.04.2 point release.
  • Disabled auto-sync as a part of Debian Import / Feature Freeze.
  • Sent Lunar Feature Freeze announcement email.
  • Added a hint for boost1.81 as it is deny listed but @jawn-smith said it passed.


  • Compliance training
  • Take home test reviews
1 Like
  • apt snapshot integration work
  • apt upstream licensing and 2.6.0 merge merging preparation ahead of debian hard freeze
  • compliance training
  • minor bug triage
  • some scope evaluation for 23.10 roadmap
1 Like


  • removed rinutils, had a Debian bug filed but removal from Ubuntu hadn’t been requested
  • removed linux-restricted-modules-oem-5.17, for an old kernel and would never migrate
  • removed minizinc from the release pocket because it FTBFS. The new version was stuck in -proposed because of a bug in boost, but since the old version is broken, removing it lets the new version in without ppc64el.
  • no-change rebuild of cgreen which had a dependency on old libbinutils.
  • found that repowerd still builds ok in Debian, so retried the build in Ubuntu (even though the last failure was only 3 weeks old). Sussed that the failure was LTO-related; uploaded change that disabled LTO and uploaded the workaround, forwarding to Debian.
  • node-object-inspect's autopkgtests fail because they require a newer version of node-es-abstract than is in the release pocket. As that package has been stuck in -proposed since May, rather than just testing node-object-inspect with --all-proposed, I looked at unblocking
    node-es-abstract by retriggering tests for its dependency, node-deep-equal, with --all-proposed. This took quite a while but eventually passed, letting us retrigger node-object-inspect alone, and it migrated.
  • removed s3ql which had been removed before but a newer version synced which is still dep-wait.
  • worked through boost1.74 blockers:
    • most just required autopkgtests retriggering
    • gromacs on ppc64el has been stuck for a while and blocked votca. Resolved by forcing the build with gcc-11, which is the toolchain used for the last successful build. No bug filed on gcc for this, will try to follow up today.


  • landed livecd-rootfs changes to get xubuntu-minimal and edubuntu images building properly
  • dumped and requeued all autopkgtests on arm64 because there were too many there that britney wasn’t actually waiting for the results of


  • working with @vpa1977 to get ca-certificates-java dependency loop resolved
  • reviewed spec around managing ciphers for crypto stacks
  • biweekly meeting for grub/secureboot: checked path for 20.04.6 and UC22
  • confirmed Ubuntu is affected by fakeroot breakage on armhf and i386; waiting for fixed fakeroot to analyze what we might need to rebuild
1 Like



  • jtreg7 package reviewed and accepted into proposed thanks to @vorlon. The discussion about inclusion in Debian is happening here.
  • openjdk-20 FTBS for ppcel64 in 20+26 build, preparing 20+36 upload, currently being built



1 Like