Previous week’s status can be found at Foundations Team Updates - Thursday 14 Sep 2023
cbmc (LP: #2036745)
- Tests on some architectures fails due to variable sorting of output. Upstream ticket created.
- Rebuilding without LTO resolves the original problems. But building against glibc-2.38 fails. This is a common problem of many packages see LP: 2032624.
spades (LP: #2036726)
- Configuration fails due to missing build dependency libzstd-dev. Debian merged my patch. New upload to Ubuntu needed.
py3dns (LP: #2033649)
- Extending the timeout does not resolve the issue
debug-me (LP: #2033639)
- Disabling LTO on ppc64el resolves the issue. Debdiff created. Looking for sponsor.
23.10 image testing
- Testing RISC-V images (no issues up to now)
- Review written interviews
- Reviewed an MP from @andersson123 adding a unit test for duplicate distro requests.
- Submitted an MP which enables the unit tests in lpci. Ironically, this failed on Launchpad but passed locally.
- Updated the above MP so that it’ll pass in Launchpad.
- Merged an MP adding pytorch to long_tests for amd64, ppc64el.
- With @andersson123 investigation into why pytorch amd64 tests were running forever. There were some hung tee processes (reading log files) due to an inability to connect to the instance running the test anymore.
- Discovered pytorch/amd64 finished running over the weekend but that it OOM’ed - added it to big_packages and it passed, then dropped it from long_tests.
- Worked and worked again with IS regarding the migration of the Error Tracker environment to PS6.
- Modified the daisy and errors charms to behave better (not barf) when the cassandra database servers are set manually instead of being part of the relationship.
- Continued investigation into issues with the ISO tracker sending email or sending them with the wrong From address. I think I finally found the right knob!
- Merged an MP from @andersson123 adding more tests for the new installer so that we are running the same tests (where applicable) for the legacy image and the new image.
- Merged an MP from @andersson123 adding a post install test for gjs.
- Updated test cases on the ISO tracker for the above.
- Merged an MP which modified how we run the screen reader test in the Ubuntu Desktop legacy image. Then futzed about with commenting out the steps in the ISO tracker.
- Reported a bug regarding a UI issue with a checkbox when selecting Advanced installation options with ubuntu-desktop-installer.
- Confirmed a bug about the screen reader not working with ubiquity after choosing “Try Ubuntu”.
- Reported a crash regarding gnome-shell from an Ubuntu Legacy desktop “Try Ubuntu” session. Did some additional research into why the retracing failed and opened a task about apport as this bears further investigation.
- Manually promoted the Ubuntu daily-live image while @paride is sorting out the Utah issues with the new image format / kernel.
- Reported a bug regarding whoopise failing to open crash reports on the Ubuntu Legacy live system.
- Assisting with the Ubuntu 22.10 Beta.
- Conducted an interview.
- Reviewing candidate applications.
- Finished forwarding porting UEFI network stack to 2.12, submitted merge request on Debian Salsa.
- Addressed issues with previous peimage changes in 2.12.
- Addressed the lack of image unloading with peimage in 2.12.
- Investigated the implications bringing GRUB 2.12 to previous stable releases
- Some minimal NX pondering
- Secure Boot meeting on Tuesday.
- fixed s390x not rebooting to target system after install (https://github.com/canonical/subiquity/pull/1799)
- fixed client using basic mode on SSH connections on s390x (https://github.com/canonical/subiquity/pull/1802)
- Coordinated with desktop to fix install of third-party codecs when connecting to a network past the network page. https://bugs.launchpad.net/subiquity/+bug/2017278
- Added support for recovery key support to curtin when zkey is used on s390x.
- Investigated bug reports
- https://bugs.launchpad.net/subiquity/+bug/2036424 - where build on s390x would cause some binaries included in the snap to crash
- ubuntu-budgie (and potentially other flavors) not getting fixes merged recently (because of branch naming changes in ubuntu-desktop-installer).
- ubuntu-budgie not getting ubuntu-restricted-addons installed despite being online. Turned out that the apt sources have multiverse disabled. If this a budgie specific requirement, we should probably get the checkbox removed?
Following up for my last +1. It’s caused by gcc-13 defaults to -fexcess-precision=standard for C++. Uploaded 1.2.2-2ubuntu4 to restore -fexcess-precision=fast. https://launchpad.net/bugs/2036181
libgcrypt20 & gcc-mingw-w64
Both FTBFS because binutils-mingw-w64 no longer recognizes option like ‘–insert-timestamp=1686475264’.
Reported to Debian and both maintainers have fixed them by removing such option in d/rules.
groff 1.23 changes warning message format and it causes lintian FTBFS.
- 24.04 roadmap planning wrt Go toolchain
- Reverted deb822 migration by default in ubuntu-release-upgrader (LP: #2036288) and curtin
- Worked on patch for systemd SRU regression in systemd-networkd-wait-online (LP: #2036358)
- Uploaded to jammy and lunar, currently in -proposed
- Uploaded systemd to mantic to fix autopkgtest (LP: #2035778)
- All Systems Go! conference in Berlin last week
- reworked state encoding (move from gob to JSON) to improve maintainability
- replaced the linter and add rules to help improve code quality (+ fixed every issues found by the linter)
- aptdaemon (“false” positive). Confirmed it was building fine.
- syslinux - ‘-fcf-protection’ is not compatible with this target (i386) https://bugs.launchpad.net/ubuntu/+source/syslinux/+bug/2036757 . Patch should soon be approved.
Returning from vacations today so not much to report
- Code review for a community contribution PR#409
- Prepared some mock ups and added to the netplan diff spec
- Helped an internal team to investigate a systemd-resolved issue
- Going through the bug reports that were opened during my absence
- I tried to debug why a user cannot import photos in digikam any more after an upgrade (LP: #2035406). I could reproduce the failure (bug #2036244) but that failure seems to be unrelated to the systemd/udev version.
- apt_dpkg: Skip parsing non-deb files with dpkg-deb
- bash-completion: Do not use ls in completion (fixes bug #1850804)
- Working on python-distutils-extra
Another short week
- Was at DebConf 23 from Sep 10 - Sep 17 (conference report coming soon)
- Travel day on Sep 18, public holiday on Sep 19
Here are some updates of developments sparsely spread across the past two weeks.
We got feedback from Microsoft on the performance evaluation of a OpenJDK performance patch submitted by a developer from Microsoft. We’d seen degradation and inconsistent gains across architectures supported by Ubuntu, when we ran the suggested benchmarks. From the MS response, it is clear that we did not run them in the right mode. The re-run will be done in the next pulse.
- The openjdk-11-fips package is now with the Security team for review
- (LP: #2009141) - figure out a amd64 reproducer of the udev problem we’re seeing on s390x. Assemble a test build including pyudev changes from @ogayot. Helps but still have more issues, which can be seen on amd64 so investigation continues.
- assemble a working s390x working around LP: #2036424 - this needed core22 from edge channel at build time. Unclear if this is a glibc/patchelf/snapcraft/other issue at this point.
- investigate failing vmtests related to cloud-init SRU testing. MP: #451677 opend and merged, maybe more needed.
- MP: #451562 - offer TPMFDE + minimal combination for Desktop ISO
- LP: #2036292 - review&sponsor for @cpete
- 24.04 planning
- Released apt 2.7.6 to Debian unstable.
- Rework the error about unmerged-usr that I added Saturday into two warnings that run before the Y/n prompt and after install, respectively.
- Reviewed a bunch of merges from Mate
- Released grub 2.12~rc1-10 to Debian unstable:
- Cherry-pick the shebang fix to keep the unmerged-usr people and CI happy
- Install empty /etc/default/grub.d - it got lost by accident
- @mkukri’s first peimage refactoring merge
- Tried to find evidence that we actually need to use the EFI handover protocol on older kernels because we can’t seem to break boot or EFI RT access.
- Writing a lot of text and talking to a couple of people
- Helped the desktop team to debug the gnome-shell kills when screen is being locked
- Patch piloting
- No-change rebuilds of remaining packages depending on the transitional
libevent-core-2.1-7aso we can drop it in favor of the fixed
ubiquityupload to fix on issue where some flavors randomly got
grub-efi-amd64in their live seed, via transitive dep from
ubiquity, instead of
grub-pc, via direct recommends from
ubiquity. This broke installs of Ubuntu Unity and UbuntuStudio who were the unlucky winners, because some ancient
grub-installercode tries to dpkg --purge
grub-efi-amd64and falls over. That code should be killed but in the meantime the bug is fixed.
- patch pilot shift
- Full SRU shift on Friday
- Follow through on systemd SRU revert, the reported regression turned out to be a false-positive
- starting up a spreadsheet to schedule exception-reviewer vanguards in the Release Team, to address some concerns raised
- work on the 23.10 Beta
- candidate interview
- meet new CISO
- first armhf time_t shift
scheduled/attended MIR discussion
24.04 roadmap planning with @mateus-morais
attended weekly .NET Partners Sync
Ubuntu Packaging Guide
PR #32 Add explanations
PR #34 Style guide
Discussion #33 VS Code Snippets
attended weekly Ubuntu Packaging Guide meeting
- prepared/applied for promotion
- Packaging .NET 8 RC1 release
- Some work on an openssl SRU (but I didn’t have that much time and I need to resume that topic)
- Work on crypto-configuration demo and specification, including a patch to openssl to introduce a system default cipherlist/ciphersuite
- A lot of armhf time_t analysis (~ 300 more packages can be analyzed iirc and partial fixes for hundreds more, along with some automation)
PR #1797 Reduced the time taken on the Unattended Upgrades step by no longer restricting to using minimal steps
Some work to allow specifying the autoinstall file on the kernel commandline, PR coming today
bindgen = 0.65.1for the kernel team. Final packages uploaded to https://launchpad.net/~liushuyu-011/+archive/ubuntu/misc-1/+packages.
- Investigated i386
zshFTBFS and discovered it was a glibc issue. Opened LP#2036283
- Completely finished
cargo-related MIRs: LP#1991650 LP#1990655
- Reviewed chiselled container pull requests