Foundations Team Updates - Thursday 18 May 2023

The previous status is here: Foundations Team Updates - Thursday 11 May 2023

1 Like

Short week due to public holidays.

Distro

Netplan

Misc

  • take-home test reviews
1 Like

Short week due to public holidays.

Distro

  • Went through all of the openssl tickets on launchpad and either closed, planned to close, or marked as Invalid, around 75% of them and started getting a better pictures of some recurring topics to work on
  • Sent my application for bugcontrol so that I could mark tickets as WONTFIX
  • Worked on a batch packages for analysis of the impact of time_t changes for armhf
  • Started a discussion on which openssl version to use since some are LTS (we’re likely going to stick to LTS ones)
  • Did some tzdata reviews
1 Like

Dotnet

  • Still working on bootstrapping a new dotnet8-preview package on Mantic
    • I’ve been getting errors on building the non-bootstrap package, which involved a bit of investigating and interacting with upstream in order to figure those out. This is in progress.
1 Like
  • Drafting specification for Go development tools
  • Continue reviewing dhcpcd5 package.
    • Staged my changes at https://launchpad.net/~zhsj/+archive/ubuntu/dhcpcd5
      • Remove Conflicts/Replaces dhcp-client
      • Drop incompatible ntpd integration
      • Fix autopkgtests
    • The program fails to run on ppc64el.
      Bad syscall, maybe related to its seccomp profile. Still investigating.
1 Like
1 Like
  • Curtin
    • Looking at supporting older Python versions in prep for a 23.1 Curtin release. Most of the changes should be merged to the curtin main branch and a few will be localized to the 23.1 release branch, which will be the last release of curtin to support python 3.5.
    • Fixed VM setup scripts to make it trivial to setup a test VM for the integration tests.
    • Various fixes for unittest and integration test reported problems to get tests running on Xenial/Bionic/Focal/Jammy hosts.
    • Working on a quirk to detect availability of non-512 byte sector size support and to enable the use, and run the matching tests, appropriately.
    • Fix several issues around older verions of filesystems tools.
    • Fix problems where sysfs should be reporting things and isn’t.
    • Adding a debug flag on how the integration tests are run to make it easier to set breakpoints and make modifications to the version of curtin being run in the integration test.
  • Setup and wrote first post for a blog focused on Autoinstall and other installer topics.
  • Greenhouse
  • Out sick 1 day
1 Like

systemd

Distro

1 Like
  • Addressed more comments on the cargo and rustc source package merging proposal
  • Finished the initial merge of the cargo and rustc source package
1 Like

Short week, with another swap day after the engineering sprint.

Archive

  • Did some work to restore select versions of cloud kernels that are current on cloud k8s images and let the existing standard daemonset images work again
  • NEW queue processing of Debian syncs
  • Misc package removals of ftbfs-in-proposed packages

Release

  • Did what I could to reduce the number of libreoffice autopkgtests running in parallel, as this was crushing infrastructure due to unreasonably large (when uncompressed) test output
  • Hinted through more packages blocked by sbuild/amd64 due to postfix failing
  • Worked on getting launchpad-buildd change landed that lets us use devtmpfs in livefs containers, so that losetup -P works
    • Landed a patch to livecd-rootfs to make the buildd images not call grub-install --removable which for some reason fails in this environment (but we shouldnt be using the –removable` flag anyway, dropped a while ago for cloud images).
  • Worked with @mwhudson to land long-pending livecd-rootfs change to refactor how we install tasks to not depend on the archive metadata. Did test image builds in a ppa across all flavors, then played whack-a-mole with build failures on ubuntucinnamon, ubuntu-unity, edubuntu, related to the new code doing a better job of enforcing seeds
  • Worked on migrating ancientminister to cdimage-internal. Now waiting on IS to schedule a coordinated cutover of the web frontends.

Distro

1 Like

Netplan

  • Continued the work on the error handling improvements PR#334
  • Prepared the release of Netplan 0.106.1 (pending release)
  • New autopkgtests for Network Manager and Netplan MR#443011
  • Currently working on improvements in the keyfile reader to fully support tunnels so Network Manager will not emit nm-devices for supported tunnel modes.
1 Like
  • Worked on migration of netplan network-manager overrides (LP: #2020110)
  • Worked on libcamera-apps packaging (LP: #2020101); some lintian tags to clean up, then will apply for sponsorship
  • Tested libcamera 0.0.5 on mantic (works fine with cameras that worked before, i.e. all but v3 module)
  • Tested docker snap report for xnox
  • Investigated debhelper issue for public cloud team (LP: #1959054)
  • Working on SRU of flash-kernel issue to focal, jammy, kinetic (LP: #2007827)
  • Interviews
  • Pi meetings
1 Like

autopkgtest

  • Updated the autopkgtest-cloud mojo/service-bundle file in git with changes that had been made on production.
  • Modified the mojo/service-bundle to set a worker-big-flavor of autopkgtest-big as that flavor now exists in every DC in which we run our tests. Submitted an MP with the same change.
  • Updated novarcs.tar (openstack / nova environment variable files) on wendigo to include lcy02 and drop lcy01.
  • Recovering from the autopkgtest-cloud-worker-lrg units running out of disk space.
  • Added an autopkgtest-cloud-worker unit (which uses a 200GB ceph volume for /tmp) to help with the autopkgtest-cloud-worker-lrg units running out of disk space.
  • Submitted an RT regarding the lower ports quota in prodstack5 which is slowing down the processing of the amd64 queue.
  • Pinged IS regarding restarting neutron on s0lp4 in scalingstack-bo02 as the s390x queues were slowly processing again. See RT 155441 for details.

Ubuntu

  • Add libreoffice/i386/mantic to never_run to match previous releases.
  • Hinted some software-properties phased-update false positives.

Misc

  • Some time off.
  • Conducted an interview.
  • Reviewing candidate applications.
  • Submitted an MP adding mantic images to openstack glance for IS.
1 Like
1 Like

Distro

  • Landed a change to livecd-rootfs to expand task definitions more directly, and fixed some fallout
  • Reviewed an openjdk-8 merge for @vpa1977
  • Reviewed some time_t-transition related scripts
1 Like

This report contains the changes since my last report (not just from the last seven days).

Distro

  • distro-info-data: Add Ubuntu 23.10 Mantic Minotaur SRUs (LP: #2018028)
  • apport: Verified apport SRUs (2.20.11-0ubuntu27.27 for focal and 2.20.11-0ubuntu82.5 for jammy)
  • tzdata: Prepared SRU to fix inconsistencies introduced by the previous SRU (LP: #2017999)
  • duecredit: Synced 0.9.1-1.1 because it included the Ubuntu change.

+1 week

  • Sponsored aespipe fix for @xypron (LP: #2019319) from last week’s shift
  • Added tag ‘update-excuse’ to @xypron’s bugs #2019319, #2019195, #2019079, #2019219, #2018702, #2018713
  • Follow-up on Lukas Märdian +1 shift: matplotlib 3.6.3-1 was not synced. My sync was rejected: “same version already has published binaries in the destination archive”. I did a fake sync by uploading 3.6.3-1fakesync1 with only has the version number changed.
  • https://ubuntu-archive-team.ubuntu.com/nbs.html is emtpy. Nothing to do here.
  • kanboard: It fails to build due to the missing php-pimple build dependency. php-pimple was removed from the archive in bug #1931913. We are blocking the sync until we are able to sync the PSR stack back with Debian. Since we are still waiting, I requested (in bug #1931913) to remove kanboard until then.
  • mps-youtube: Version 2.9.4-1 fails to build from source due to failing unit tests, because some dependencies are missing (at least pip; probably also python-pkg-resources, see Debian bug #832087). There are four open release critical bugs in Debian for mps-youtube. A look upstream indicates that there is more work needed to get it running (see https://github.com/mps-youtube/yewtube/issues/1143). There is an active fork named yewtube. So I requested to remove mps-youtube (LP: #2019754)
  • cryfs: Fails to build from source due to failing test. C++ exception with description “Error calling mlock. Errno: 12” thrown in the test body. errno 12 says: ENOMEM 12 Cannot allocate memory. I opened bug #2019853 for it.
  • libclamunrar: The binary package libclamunrar11 1.0.0-2 depends on libclamav11 which requires clamav >= 1.0 to provide it. Poked the server team to sync/merge the new clamav version. They have the merge on their radar (see bug #2018063).
  • flask-appbuilder: flask-appbuilder 4.1.4+ds-3build1 fails to build from source against python3-flask-sqlalchemy 3.0.3-1. I tried to update the package to 4.3.1 (plus the ustream commits until 2023-05-16) to make it work with flask 2.2.2, but this is not enough. I filed a bug against Ubuntu and Debian (see bug #2019860) and requested to remove the package until it is made to work with flask-sqlalchemy 3.0 (LP: #2019862).
  • Synced several packages for @enr0n (Ubuntu delta was merged in Debian): bug 2019241, bug 2019242, bug 2019243, and bug 2019245
  • python-ansible-pygments: Sponsored fix for @ogayot: https://salsa.debian.org/python-team/packages/python-ansible-pygments/-/merge_requests/1
  • Sponsored sync requests from https://reports.qa.ubuntu.com/reports/sponsoring/:
  • golang-github-crowdsecurity-go-cs-bouncer: It build depends on golang-github-crowdsecurity-crowdsec-dev (>= 1.4.2). crowdsec fails to build. This is caused by vendored sources in docker.io. I found https://bugs.launchpad.net/ubuntu/+source/docker.io/+bug/1946376 for it and added golang-github-crowdsecurity-go-cs-bouncer as affected as well. Lucas Kanashiro from the server team is working on splitting the docker.io library (use no vendoring) from the application (use vendoring for easier backporting).
  • git-annex: It fails to build with “requires dynamic R_X86_64_PC32 reloc against UtilityziYesod_widgetFilezuw_closure’ which may overflow at runtime”. I opened bug #2019992 for it without further investigation.
  • stress-ng: bug report #2019079 already described the test case failure on armhf and pointed to an upstream commit that fixes it. I successfully tested the fix in a PPA and uploaded 0.15.07-1ubuntu1.

Apport

1 Like

Been off on vacation in sunny Fuerteventura this week (Mon through Thurs), but today I’ll be looking into doing some preliminary work on an upstream MR to Debian that would improve the logging in autopkgtest. I’ll also probably spend some time linting the autopkgtest-cloud repository - the goal is to soon have the repo passing pylint, yamllint and shellcheck, then we can enable this in CI.

1 Like