Foundations Team Updates - Thursday 11 August 2022

See also Foundations Team Updates - Thursday 04 August 2022

1 Like
  • subiquity
    • PR: 1378 - first fix for guided not respecting the max number of primary partitions, addresses can_be_boot() and a gap
    • PR: 1379 - guided can create logical parts, if the target gap is in an extended partition
    • PR: 1380 - update locale -> layout keyboard mapping with info for ubuntu-desktop-info locales
    • PR: 1382 - add the wait arg to v2 Guided GET, one step closer to obsoleting existing guided methods
    • PR: 1383 - do a few guided storage related refactors
    • PR: 1384 - tweak filesystem object IDs to try to make it more clear that the number on the end is not related to the fs path
    • (no PR yet) LP: #1983036 - root cause issue with ubuntu-desktop-installer manual partitioning not reformatting a partition when it should, write tests and start fix
    • (no PR yet) - add logic to the boot code to be able to project if a partition needs to be created or not. Intended for use with guided as part of overall code that evaluates the possible guided installs to pre-determine if there are enough primary partitions left to be able to install
    • Review PR: 1372, 1381
    • Assist ISO testing
    • Investigate issues with converting a complex preseed file to autoinstall
  • curtin
  • ubuntu-desktop-installer
  • curl vs hddemux vs knot-resolver
    • @jawn-smith did some work to investigate this and found it was due to knot-resolver. That package is failing autopkgtest, but that’s already known. Further control file fix needed to not build on ppc64el (LP: #1983723), done by Debian maintainer. As result, remove the knot-resolver binary packages on s390x and ppc64el, which allowed curl to migrate.
    • However, doing that package removal means that hddemux FTBFS. Only on ppc64el, it had a plan for this problem on s390x. LP: #1984134, in sponsor queue, forwarded to Debian.
2 Likes

Distro

  • apt/bionic SRU verified fine automatically, but manual validation failed, investigation yielded:
    • The new kernel autoremoval code added a new virtual package to the cache to keep track of all kernel packages
    • but if your cache was up-to-date after upgrading apt, you’d get the old cache
    • hence, no kernel was detected
    • need to bump cache minor version
  • Assisted with the 22.04.1 Nvidia offline install issue, providing ideas to solve it from the apt side of things

Infra

  • Did a bit off digging into autopkgtest-cloud why kernel tests suddenly stopped working, noticed accidental override on servers, handed over to @bdmurray at that point.

Orga

  • Filed expenses for DebConf
  • Investigated trains to Oct sprint
  • Hiring take home tests reviews
2 Likes

+1

  • @jawn-smith has been helping with patch review and sponsorship.

  • hfst dep-wait on libfst-dev

    • This package needs libfst-dev (>= 1.7.9) from src:openfst, which is not able to autosync from Debian because of Ubuntu delta. However, this delta was just to fix the build with python3. The openfst in Debian no longer needs python for the build, so we synced the Debian version.

      One follow-on change was needed to avoid excessive memory usage and FTBFS during openfst’s dh_auto_test (LP 1984137).

  • ghdl FTBFS with LLVM 14 (LP 1983908)

    • The package explicitly adds a line for each LLVM version in it’s configure script, hence the history of existing patches to allow newer LLVM versions. After fixing the build with LLVM 14.0, I hit FTBFS on armhf (LP 1984138). After trying a while to fix or work around the single test, I opted to disable to tests on armhf and opened an upstream bug (https://github.com/ghdl/ghdl/issues/2170).
  • libxsmm FTBFS (LP 1984111)

    • This FTBFS with an undefined reference to pthread_yield. Upstream already has fixes for this, so I cherry-picked those.
  • node-grunt-timer FTBFS (LP 1983866)

    • This fails because of a missing node-functional.js dependency. This package was removed from Ubuntu because of https://bugs.debian.org/cgi-bin/bugreport.cgi?bug=1014605, i.e. it never built from source in Debian or Ubuntu. Steve was happy to remove node-grunt-timer from the archive.
  • python-duniterpy breaks silkaj autopkgtest (LP 1984122)

    • The immediate cause is the BlockUID type being changed to BlockID, but the upstream commit that resolves this does not apply cleanly to v0.9.0. The best resolution here is probably to wait for v0.10.0 to be packaged in Debian unstable (currently in experimental).
  • generate-ninja FTBFS (LP 1984224)

    • The gn binary is DWARFv5, but dh_dwz is not fully compatible with DWARFv5. Setting the -gdwarf-4 build flag forces DWARFv4 and makes dh_dwz happy.
  • libtree FTBFS on some arches (LP 1984008)

    • The immediate cause of this is that one of the test cases unconditionally compiles binaries with -m32 and -m64 flags, which do not exist on all arches. Upstream has an open issue tracking this, and due to the complexity it seems better to let it be sorted out there. Opened an update-excuse bug with this info.
  • deepin-album FTBFS with LTO enabled (LP 1984124)

    • Disabling LTO fixes the build.
  • faketime FTBFS (LP 1984446)

    • This package had a test snippet that used a format specifier for time_t that was not portable to 32 bit arches. I fixed the print statement and the patch was merged upstream.

Other

2 Likes

Distro

Netplan

  • ip leases PR#284, review comments & merge
  • VXLAN review comments, PR#288
  • netplan status --all discussion & spec comments (Michael, Steve, Amy/UX)
  • netplan 0.105 release prep
  • drafting a netplan status prototype

Misc

  • sprint travel planning & request
2 Likes
  • Short week, national holiday and travel
  • FTBFS reports
  • Proposed migration:
    • Still working on [MIR] lib*-perl for lintian 2.115 (LP: #1980662)
    • Retried all regressed autopkgtests (795 in total), then triggered migration-reference/0 tests for those that still failed (491 in total). Presumably, the difference (304) is the number that passed.
  • Since my last update, I sync’d the following TIL universe packages: python-ase, joblib, toil, deal.ii, crossfire, libgetdata, pacparser, csound, wreport and catch
2 Likes

ubuntu-image

+1 maintenance

  • Working on repackaging new upstream olive-editor.
    • Uploaded patch for opencv.
    • Uploaded patch for openimageio.
    • Packaged new upstream opencolorio.
    • Packaged new upstream olive-editor but haven’t uploaded yet because of a linker error.

Go 1.19 transition

  • Did another set of bulk rebuilds with the changed dh-golang.
  • Wrote some scripts to find packages that FTBFS in one PPA but not another to indentify the number of new failures.
  • Synced golang-defaults 1.19 from Debian.
  • Uploaded new version of gitbrute to resolve new FTBFS.

Distro

  • Sponsored a few different packages for @enr0n to support his +1 maintenance.
    • ghdl.
    • openfst.
    • deepin-album reviewed but waiting to upload until python3-stdlib issue is fixed.
    • libxsmm reviewed but waiting to upload until python3-stdlib issue is fixed.
1 Like

subiquity / curtin

  • reworked and merged MP 427671 in curtin and PR 1372 in subiquity to invoke curtin install multiple times. Log files are now contained in a single known location.
  • opened and merged PR 1381 to make Subiquity execute hook scripts just before executing late-commands. This gives the desktop installer an opportunity to copy over files to the target system and run additional commands that have no builtin support in Subiquity.
  • ran into a scenario where using an autoinstall file generated from a previous install crashed the new install. Will rerun the test-case and open a LP bug with the logs.
  • working again on the implementation of Ubuntu Pro in Subiquity.

distro

  • zlib migration: tested @Frank’s patches for samtools-legacy to get libbio-samtools-perl fixed with new zlib. This requires a no-change rebuild of libbio-samtools-perl since it statically links against libbam-dev. Thanks to @slyon for sponsoring the work!
1 Like

The Ubuntu QA team’s status can be found at Ubuntu QA Team Weekly Status Update

1 Like

Distro

Other

  • Completed all six in-house training sessions.
1 Like

22.04.1 point release

Testing snapcraft 7.1.1 on RISC-V

Automating testing SPDK, DPDK

  • works fine now on x86
  • RISC-V test automation still in progress

Demo video on behalf of StarFive

1 Like

Distro

  • Figured out the tracker-extract crash issue thanks to bdrung, SRU uploaded, waiting for review (and librarian to come back online). LP: #1983859 (cc @bdrung)
  • Investigated an issue with the Desktop installer not installing nvidia drivers. LP: #1984171 (cc @sil2100)

glibc

  • Finalized a candidate package for 2.36
  • Investigated build failures for said package, eventually narrowed down to gcc 12 by @mwhudson

Netplan

Rust

  • Still working on the security aspect of the upcoming http-parser MIR
2 Likes