Foundations Team Updates - Thursday 16 March 2023

The previous status is here: https://discourse.ubuntu.com/t/foundations-team-updates-thursday-09-march-2023

1 Like
  • subiquity
    • PR: 1592: docs improvement around password
    • PR: 1594: I did a branch a while back to try to speed up the integration tests, but it was hard to convince myself it was correct. Starting to work on that again. This PR adds the ability to run the TUI test jobs in the background, a requirement for running them in parallel.
    • PR: 1593, PR: 1595, PR: 1596 (LP: #2009797): fix a startup race condition than can produce exception. AttributeError: 'FilesystemController' object has no attribute '_start_task' Also improve some related task synconization usage. Implemented a more agressive version of this as well, in case other similar API bugs are lurking.
    • analyze lzo vs xz compression for the Subiquity snap. lzo is larger but doesn’t seem to help the service start time. Retaining default of xz for now.
    • LP: #1902287 - Update manual test case around raid1. Disks should be manually marked as bootable.
  • probert
    • investigation on why resize tool failures were not logged, fix pending
  • 23.10 planning
  • help 20.04.6 testing
  • greenhouse
1 Like

Rust

  • Currently drafting a proposal to merge cargo and rustc source packages

Distro

  • Finished rustc proposed migration

Ubuntu ROCK

  • Working on validating Chiselled images for C/C++/Rust application deployment
1 Like

Report covering the past two weeks.

proposed-migration

  • freedink: gnulib is awful. package was failing to build on some archs, because in some cases SDL.h recursively includes unistd.h and gnulib provides its own implementation of unistd.h on the path which can’t be included unless config.h is included first. Patched to fix this, uploaded, and QA upload to Debian.
  • btllib: drop depends on libsdsl-dev to a recommends since none of the headers use it. Fixes installability on riscv64. Forwarded to Debian.
  • removed pycuda temporarily from the release pocket because it was blocking boost1.74 migration but with no clear output to explain why. Once boost1.74 migrates, was able to take a closer look and see that libnvidia-compute-515-server conflicts with libcuda* so this would need to be sorted out.
  • removed armhf binary for gnome-subtitles, it was also dropped in Debian and this version is in the next Debian release.
  • merged new node-configurable-http-proxy from Debian to see if it fixes the build-time testsuite failures, since there are some test-related changes there.
  • removed liboqs, it’s not supposed to be released so there’s no reason for us to keep it in -proposed. We are not blocking it from syncing because it may be suitable for release at a later date.
  • removed stylish-haskell from lunar and lunar-proposed, it build-depends on obsolete haskell packages and needs a sourceful update in Debian.
  • singularity-container: not in release, has a new version in unstable so synced that to see if it’s buildable. (It wasn’t.)
  • removed rust-cradle because it depends on an obsolete version of rust-nix Debian bug #1030989
  • removed rust-tree-magic because it build-depends on removed rust-nom-3, and added to extra-removals

Archive

  • Revert mess with NBS report incorrectly telling us to remove linux-image-6.1.0-16-generic.
  • pbdagcon has been uninstallable for a while on riscv64, finally took a look and found there was dependency skew on this arch vs the others due to timing of builds. Pushed a no-change rebuild that resolved it.
  • fixed up priority-mismatches. Now in the clear for lunar.
  • General NBS cleanup, the
    list
    is now looking very good for lunar release
    • best bug report ever about interaction between glibc 2.37 and cyrus-imapd. After much discussion uploaded fixed cyrus-imapd.
  • Handled migration of snakefruit to ubuntu-archive-toolbox. Not completely smooth sailing but everything should be working now. The noticeable impact is that [https://people.canonical.com/~ubuntu-archive/] is now called [https://ubuntu-archive-team.ubuntu.com/]. Main remaining things are:
    • the CGI that rmadison talks to is still on snakefruit because ubuntu-archive-toolbox doesn’t serve http directly, and the web frontend doesn’t run CGI
    • livecd-rootfs talks to archive-team.internal over http as an alias for snakefruit. Again, since ubuntu-archive-toolbox doesn’t serve http directly, this will need to point to ubuntu-archive-team.ubuntu.com instead which means SRUs of livecd-rootfs.

Release

  • Spun images for 20.04.6. Found that the no-change SRU of debian-installer failed to pick up the correct grub binary, so failed the (thorough) SRU test case. Iterated until it was correct.

General

  • sponsored new version of openjdk-20 for @vpa1977 to fix ppc64el ftbfs
  • completed evaluation of the impact of broken 32-bit
    fakeroot
    on the Ubuntu archive. 19k binary
    packages on armhf and i386 analyzed that have built since the December
    coreutils update; after code review found the bug in my script that caused
    this to show no buggy builds, fixed the script and found two buggy builds.
    Fixes uploaded.
1 Like

Out sick Monday, Tuesday, Wednesday.

  • prepared gnu-efi 3.0.15 backport to kinetic, test rebuilt reverse dependencies in main locally
  • test rebuilt fwupd-efi against older gnu-efi in jammy (it builds too)
  • uploaded fwupd/bionic SRU to remove EFI binaries
  • prepared and submitted for signing fwupd-efi/bionic
1 Like

Short week, only 3 days.

Distro

Netplan

Misc

  • take-home-test reviews
  • candidate interviews
  • Development Plan presentation
1 Like

Strange week as I’ve been moving into a new house… and then back out, and then in again.

ubuntu-image

Other

  • 360s
  • Interviews
  • Take home test reviews
  • e2fsprogs proposed migration (work in progress)
1 Like

glibc

  • glibc 2.37-0ubuntu1 finally migrated out of lunar-proposed!
  • "You’ll never believe what happens when you write snprintf(buf, INT_MAX, "foo")!"…
    Well, stuff breaks. LP bug and upstream thread. Investigating this took a good long while but was worth it.
  • Since those two hurdles have been cleared, I also uploaded 2.37-0ubuntu2, because the autopkgtest queues were too empty for my taste.

Misc

  • Small batch of apport reviews
  • Take-home test reviews
  • 360 reviews
  • Netplan config spec review
1 Like
1 Like

was off last week

subiquity

  • opened PR 1597 to move from core20 to core22. This turned out to be a real headache because of a regression introduced in jammy when running pip3 install --prefix. https://discuss.python.org/t/18240 had some pointers about what’s going on but no clear path forward. I managed to work around the issue and did multiple successful installs with core22-based subiquity.
    Also filed a FFe for it.
  • before leaving for vacation last week, started to work on the SSH screen to decouple the key import operation from the form submission. The screen shows a list of imported keys and the users gets the opportunity to import multiple keys or remove specific ones.
  • starting to look at bug fixing

misc

  • 360 reviews
  • post vacation catchup
  • spec review
1 Like

Distro

  • Proposed migration:
    • zip (just timed out because slow network)
  • FTBFS
    • golang-golang-x-tools, because a new test is added in golang 1.20.2 (wait sponsor to sync #2011511)
    • git-sizer, FTBFS before golang 1.19 (wait sponsor to sync #2011561)
    • pt-websocket, FTBFS before golang 1.19, removed both in Debian and Ubuntu.

Snap

  • Get familiar with Go snap. Trying to make it build on snapcraft 7.x.
  • Slowly catching up the issue list on Docker snap.
1 Like

MIR inetutils-telnet

dotnet

Documentation

ubuntu-mir

ubuntu-maintainers-handbook

Bug-Control

Other

  • 20.04.6 image testing
  • Canonical Security Training (EMEA)

Massive Thanks to Graham Inggs (ginggs), Lukas Märdian (slyon) and Miriam España Acebal (mirespace)!

1 Like

systemd

  • Worked on spec for ManagedOOMPreference integration
  • Re-did systemd Focal SRU since there was a security update

Other

1 Like

Distro

Apport

1 Like

autopkgtest

  • Submitted, merged an MP for autopkgtest reverting a commit which was used to work around systemd’s oom-killer killing all the processes in the testbed. The work around should no longer be necessary with the version of systemd in the -release pocket.
  • Updated the autopkgtest code base on both the autopkgtest-cloud-worker units.
  • Removed an autopkgtest-cloud-worker unit with less resources available than the other two.
  • Modified the archive mirror used by test runners in bos01 and bos02 to work around an issue with those DCs accessing ftpmaster.internal.
  • Submitted a MP which allows the autopkgtest-cloud-worker units to use a data center specific mirror file when creating an adt image. This will allow us to more easily workaround failures connecting to ftpmaster.
  • Wrote a discourse post which will be used to post information about the status of the autopkgtest service.
  • Updated the above as some services improved.
  • Worked with IS regarding being unable to use nova or openstack commands in bos01.
  • Bumped the number of amd64 workers due to the queue processing slowly and the fact that we used to be running a lot more workers.

Error Tracker

  • Modified and deployed updated daisy code which no longer accepts crash reports about snaps to save database space as there was an IS alert about a Cassandra node filling up.
  • Modified and deployed updated daisy code which no longer accepts crash reports from Bionic as that release is going EoL soon and again the database is filling up.

Ubuntu

  • Investigate failures to boot with the 6.1 kernel.

Misc

  • Caught up after being on holiday
  • Take home test reviews
  • Met with IS regarding some servers which are in 3FP and need migrating
  • Met with IS regarding the Error Tracker database
  • Administrative work
1 Like

jtreg7:

  • A discussion about inclusion in Debian
  • Submitted a draft packaging (in review).

openjdk:
Preparing for April security update:

1 Like