The previous status is here: Foundations Team Updates - Thursday 02 February 2023
- Created PR to resolve issue with /etc/resolv.conf
- Reviewed snapd PR for supporting max-revision-assertions and determined that there is nothing to be done in ubuntu-image to support this. All of the information is sourced from existing snap files, so no extra input from u-i is required.
- Wrote a spec for supporting specifying snap revisions for core and classic builds
- Began prototyping this revision support - working for core builds
- Began prototyping fully preseeding snaps in ubuntu-image builds. Spec to come
- Running into weird cgroups issue that I haven’t figured out yet
- Investigating snap FTBFS on riscv64
- Traced this to a bug in core22 and reported it
- Investigated snap performance on RISC-V
- Almost exactly the same performance as a static binary
- Merged icu from Debian
- proposed-migration: reprepo vs libarchive
- Have some ideas to fix but need queues to settle down
- Candidate interviews
- fuse migration (I got stuck there), https://pad.lv/2006483
- MIR: a bunch of PRs and reviews for process improvement: https://github.com/canonical/ubuntu-mir/pulls
- release and deploy v0.106 \o/
- bug grooming
- environment analysis
- vxlan tristate (merged), https://github.com/canonical/netplan/pull/311
- permanent MAC (merged), https://github.com/canonical/netplan/pull/278
- netplan_delete_connection (memfd fixes, merged), https://github.com/canonical/netplan/pull/322
- autopkgtest stability testing (merged), https://github.com/canonical/netplan/pull/223
- route duplicates (merged), https://github.com/canonical/netplan/pull/320
- NM+netplan autopkgtest review for @danilogondolfo (FR-2786)
- 0.106 pre-testing, https://autopkgtest.ubuntu.com/results/autopkgtest-lunar-slyon-testing/?format=plain
- SolutionsQA/MAAS policy-based routing (PBR) debugging (@marosg)
- python3-defaults with Python 3.11 as default migrated
- Merged a migration hint for datefudge for @adrien
- Added finalcif, igraph, numcodecs, rust-png and cloudkitty to big_packages
- Also mailman3 on ppc64el for @ogayot, but that is still failing
- Investigated whether sync of python3-defaults with Python 3.11 only could be safely migrated without re-running all of the autopkgtests it triggered. @vorlon added a skiptest hint and removed the pending tests from the queue
- proposed migration: setuptools vs numpy / ros-bloom / jenkins-job-builder
- numpy just needed a retry with the newly migrated version
- a regression in pandas appeared, and it needed the same
- ros-bloom was fixed in Debian and was auto-sync’d
- I proposed a patch for jenkins-job-builder in Debian bug #1030261 which was included in the next upload
- I believe everything is in place now for migration, just waiting on autopkgtests
Go 1.20 transition
- Update packages in Debian to fix building with Go1.20
- golang-golang-x-tools 0.5.0+ds-1
- golang-github-r3labs-diff 3.0.0-3
- golang-golang-x-vuln 0.0~git20230201.4c848ed-1
- golang-honnef-go-tools 2023.1-1
- golang-refraction-networking-utls 1.2.1-1
- golang-step-crypto 0.24.0-1
- golang-google-protobuf 1.28.1-3
- golang-github-google-go-github 48.1.0-2
- gokey 0.1.2-1
Patch also sent to upstream https://github.com/cloudflare/gokey/pull/63
- golang-github-russellhaering-goxmldsig 1.2.0-1
- golang-github-crewjam-saml 0.4.12-1
- golang-github-mmcloughlin-avo 0.5.0-1
- golang-github-henvic-httpretty 0.1.0-1
Patch also sent to upstream https://github.com/henvic/httpretty/pull/18
- golang-ginkgo 1.16.5-2
- golang-gopkg-square-go-jose.v1 1.1.2-4
- golang-gopkg-square-go-jose.v2 2.6.0-2
- google-osconfig-agent needs a patch to build with Go1.20 https://pad.lv/2006676
Patch also sent to upstream https://github.com/GoogleCloudPlatform/osconfig/pull/457
- Should we remove consul from Ubuntu? https://pad.lv/2006724
- It’s time to update golang-defaults to 1.20! https://pad.lv/2006713
Go packages migration
- golang-github-mwitkow-go-conntrack 0.0~git20190716.2f06839-3 to fix flaky test
- Review server team’s dh-golang MR .
- Help server team fix building xe-guest-utils.
- LP #1998513 “SRU provide RISCV_EFI_BOOT_PROTOCOL in Focal” in now fix-released.
- LP #2006693 “Merge u-boot-menu 4.2.1 from Debian” needs sponsor
AMD GPU support on RISC-V
Due to missing llvm-pipe for riscv64 some Mesa packages are not available for Jammy.
- Started testing in Jammy and Lunar.
The 5.19 kernel complains about a “soft lockup” when a Radeon is installed. Increasing the kernel watchdog threshold works around the issue.
- Created a merge request to add a kernel command line parameter in the installer image.
- Changes in livecd-rootfs need to be addressed next.
- Vvxlan settings fix merged PR#311
- Missing routes fix merged PR#320
- Code review for @slyon PR#322
- Working on the SRUs for the netplan set fix (and found few remaining issues to be fixed in process)
- Fixed a few things in the NetworkManager+Netplan autopkgtests
- Pushed a change to temporarily disable a libcap2 test that was failing on armhf due to the kernel version we use there (merged)
- Pushed a merge request to Debian to update Rich to 13.3.1 (wasn’t accepted but the maintainer told me he will update the package)
- Prepared a build of GHC with LLVM 13 backend for s390x:
- Again more time on the crypto configuration specification (last touches to be done this evening)
- Prepared a workaround for anacron 2.3-33 disabling itself when being upgraded (if the user has upgraded to an intermediate version, it’s not possible to fix this automatically) https://bugs.launchpad.net/ubuntu/+source/anacron/+bug/2006589
- Attended fosdem; had discussions on 32-bit time_t migration and attended talks about distro and cgroup v2 among others
- distro-info: Uploaded 1.5 to unstable to skip mypy on i386 (not available on Ubuntu)
- Make Python version PEP440 compliant (LP: #1991606)
- chdist: Default to Ubuntu 22.04 “jammy” in sources.list example (Closes: #826313)
- Do not install Python test cases
- suspicious-source: Fix detecting .vfb and .vtp
- Format Python code with black and run black during package build
- Sort Python imports with isort and run isort during package build
- Uploaded 2.23.0 to unstable
- Merged to Ubuntu and uploaded 2.23.0ubuntu1 to lunar
- Drop Asia/Rangoon from debconf options (was renamed to Asia/Yangon)
- Update conversion targets to America/Indiana/Indianapolis
- Remove incorrect fuzzy debconf translations
- Test convert_timezone for consistency
- Build timezones that differ pre-1970 (LP: #2003797)
- Update English and German debconf translations
- Test timezones using Python’s zoneinfo module
- Stop creating /etc/timezone and remove it on upgrades as a one-time action, but keep updating it in case users restore it (Closes: #822733)
- Remove /etc/timezone on purge (LP: #1683595)
- Test debconf configuration with autopkgtest
- d/tzdata.config: Group matches by target timzones
- Upload 2022g-3 to unstable
- Fix configuration failure with relative /etc/localtime symlink (Closes: #1030742)
- Fix exit status 10 for invalid /etc/localtime symlinks
- Upload 2022g-4 to unstable
- ci: Add Ubuntu 23.04 “Lunar Lobster”
- test: Add test_init_error_log_is_tty()
- test: Add unit test cases for check_lock()
- refactor(apport): Avoid consecutive sys.stderr assignment
- test: Mark error case in test_get_file_package_diversion with no cover
- refactor(test): Introduce _apport_args() helper
- refactor(test): Make apport_path a public variable
- refactor(apport): Let argparse check type of arguments
- refactor: Use forward_crash_to_container in test_coredump_from_socket
- Working on ddeb-retriever to build an InRelease file (LP: #2002833) to make http://ddebs.ubuntu.com/ more reliable. I spend most time how to run it locally without building a complete ddeb archive (taking days and terabytes of storage).
- MP: 436862 (LP: #2004609) - fix swap partition creation with v1 storage config
- MP: 436927 - Fix integration tests, which were using a mix of sfdisk and sysfs to get partition state. Moving to pure sfdisk gives better results.
- LP: #2005131 (no MP yet) - start adding support to open an existing dm_crypt
- LP: #2004659 - root cause issue with installation of packages when we are online and those packages aren’t on the ISO. Due to an unfortunate sequnce of events, subiquity can be incorrectly told that there is no default route. Plan to fix this is to treat route_change events as a notification that there is a change, and then to have subiquity go retrieve the default route state rather than relying on the info from the route_change event directly. LP has more details.
- PR: 1551 - fix CI on Lunar. Removed unneeded wheel builds, subiquity should be fully using archive packages.
- PR: 1555, PR: 1557 - api test cleanup
- PR: 1556 - add some automated testing for unformatted partition creation on the v2 API, and unblock creation of those partitions
- mini-iso-tools - start improving and de-hardcoding the JSON handling of simplestreams
- File LP: #2006404 about python3.11 and a
fatal error: longintrepr.herror that can be seen in some cases
- reworked and merged PR 1547 to get output of mirror testing translated in the user’s selected language.
- opened PR 1552 to add support for multiple candidate mirrors ; with automatic selection for autoinstalls and desktop installer.
- helped investigation on LP: #2004659 (failure to install packages that supposedly are in the pool) and discussed with @dbungert possible solutions.
- Uploaded mailman3 (thanks @tsimonq2 for sponsoring!) to fix autopkgtest regression. Sadly, another test is failing, presumable on ppc64el only.
- investigated mpdecimal FTBFS.
- The upstream tarball only contains pre-built documentation that is not compatible with recent sphinx versions.
- @doko contacted upstream.
- debian bug: https://bugs.debian.org/cgi-bin/bugreport.cgi?bug=1013395
- Merged systemd 252.5-2 from Debian, preparing PR for Ubuntu upload
- +1 Maintenance
zeal missing builds on several arches
The latest Debian upload restricted the arches to where qtwebengine5 is available: http://launchpadlibrarian.net/617552161/zeal_1:0.6.1+git20220714+6fee23-1_1:0.6.1+git20220714+6fee23-2.diff.gz. We need to remove zeal on those arches.
openlibm FTBFS on armhf
Debian already has a fix committed, but not release for a few months now. I verified the fix works for us, and have proposed uploading the fix with a
maysyncversion to allow the next Debian upload to autosync.
This FTBFS because some symbols listed in debian/libvbz-hdf-plugin0.symbols are missing from the compiled library. This .symbols file was added in the most recent upload, and from what I can tell, these specific symbols may have been added by mistake. I discussed with the Debian maintainer and they agreed it would be better to just not have a .symbols file in this case. This version has synced, and successfully built in -proposed.
FTBFS due to a test that has a different outcome on python 3.11. The upstream has a larger commit for python 3.11 support, so I took the relevant change.
aseba missing builds
This package went from Architecture: any, to specifying a subset which excludes s390x and ppc64el (https://salsa.debian.org/science-team/aseba/-/commit/d5ad0ce97c814ccac777cf5ee86eeaaf4b5a9bac). It still builds fine on these arches, so I sent a patch to Debian to re-enable. It was accepted, and the latest upload has since synced and migrated.
Investigated adding some packages to
big_packagesas @bdmurray requested. I will make a PR for autopkgtest configs.
mmc-utils FTBFS on ppc64el
Upstream already had a patch for this, but Debian has not released a new upstream version. I provided an Ubuntu patch cherry-picking the upstream patch, and forwarded to Debian to make them aware. Thanks to Sergio (sergiodj) for sponsoring.
minizinc FTBFS on ppc64el
This FTBFS on ppc64el due to a narrowing conversion in libboost1.74-dev. There is an upstream bug/PR to fix the issue, but it has been stale for a year now. The fix should be simple, but I didn’t think it made sense to upload a new boost1.74 at this stage in the cycle.
- Still investigating slack space at the end of our images (rpi-imager#480)
- Investigated state of
libcamerapatches; appears most (all?) patches from raspios are upstream for 0.0.4
- Worked on libcamera support in lunar:
- current v0.0.2 doesn’t work with the v2 or v3 Pi cameras (haven’t tested v1 or HQ yet)
- custom build of the raspios version (customized v0.0.3; actually closer to v0.0.4) also fails
- build of v0.0.4 from debian experimental also fails
- appears to be an issue with IPC to the IPA proxy in all cases, but more likely something going wrong in IPA init
- Proposed migration of
sbuild(LP: #2003201): appears to be a change in zstd headers/footers. Found potential error in libzstd build but doesn’t appear to change the actual build options in the package; still digging
- Tested backport of veusz (LP: #1982839); lunar version works on kinetic and jammy
- 22.04.2 release meeting
- uploaded 2.37 to lunar last Friday (thanks @bdmurray)
- spent a big chunk of my week looking into the various autopkgtests results
- patched a glibc regression in cross-toolchain-base-ports
- Usual batch of reviews & merge for @bdrung
- Reviewed the GH backend for WSL and snaps PR
- final review of netplan PR on netplan_delete_connection()
- Sponsored a s390-tools upload to Lunar for @Frank
- Rust and Cargo updated to 1.66.1 (LP#2005122) and 0.67.1 (LP#2005123) respectively. Currently, packages are in a personal PPA.
- Rust coverage/profiling (LP#2002694) enabled in LL. The packages which enabled profiling are in a personal PPA.
- Finished investigating the requirements of creating a minified Ubuntu image for Rust application deployment. Image creation pending.
- Currently exploring the possibility of creating a Ubuntu image for Rust application development.
Short week due to FOSDEM.
- Several infrastructure issues requiring attention and fixing
- Currently working around an issue due to Ceph breakage on PS4.5
- Resubmitted @vorlon MR adding
--architectureto autopkgtest upstream
- Still a WIP with many review comments to the old MR to be addressed
- Reviewed @seb128 MPs fixing ISO static validation testing for the Deskop ISOs
- Weekly triage rotation
- Reviewed SRU MP fixing qemu crashes on non-amd64 due to LTO
- Azure HA testing: credentials renewal
- ISC Kea packaging: work at fixing bugs on the Debian side in view of the MRE:
Short week due to public holiday
- helping to make chiseled JRE8
- opened merge requests to build aom with system libwebm:
- opened merge request (1/2) for Python migration:
- working on https://autopkgtest.ubuntu.com/results/autopkgtest-lunar/lunar/s390x/v/vim-youcompleteme/20230204_215317_554ef@/log.gz for vim migration
- prototyping aqa-test packaging for use in openjdk autopackage test