Previous status: Foundations Team Updates - Thursday 2024/07/25
Next status: Foundations Team Updates - Thursday 2024/08/08
imagecraft
- PR # 47 - Migrate to core24 and python3.12
- Merged. Even though the bug in craft-application is not fixed, considering imagecraft is in early “alpha” we are OK to be limited running on 24.04 only.
- PR # 51 - Make the snap able to run on systems without python3.12
- Done, reviewed and merged. Thanks @tigarmo.
- PR # 50 - Document imagecraft commands
- Done, reviewed and merged.
- PR # 53 - Make imagecraft build an image
- WIP. Fixing several small bugs one after another to each time go a step further in the build process.
- Also documenting how to build an image at the same time.
ubuntu-image
- GPT header handling
- PR # 237 - Fix YAML validation to enable having no gadget and no artifact at the same time
- Small bug I found while making imagecraft work
- Done, reviewed and merged.
Misc.
- Attended mid-cycle meetings
Distro
- initramfs-tools: All my pending patches were merged. So I prepared some more merges:
Apport
I spent a lot of time on apport-retrace needs more than 1 GB memory (when using sandbox). apport-retrace builds a dictionary that maps paths to Debian packages. For jammy there will be eight million entries in the dict consuming 1.7 GB of memory. I distilled down that code path and wrote different implementation using different databases (SQLite, LevelDB, LMDB, NDBM, GNU DBM). Benchmarks for them can be found on apport/Benchmark.md at benchmark-databases · bdrung/apport · GitHub. SQLite was the fastest on a Raspberry Pi Zero 2W. For SQLite I implemented different database layouts (from simple key-value to normalized directory,name,package). Result of that: You can trade creation time with saved disk space. Since I was too curious I spend some nights on rewriting those SQLite benchmarks in Rust. The Rust implementation is obviously faster. Surprisingly the creation time for the different database layouts is nearly the same. So I conclude that increased creation times in Python are caused by the Python code (and not by the more complex SQL statements).
- Reviewed and merged launchpad: don’t systematically order retrace on private reports (FR-7945)
- Reviewed and merged “Fix” the apport-retrace tests on s390x
- Reviewed, fixed, and merged CI: only run the linters on a single configuration
Sponsoring
- Sync golang-entgo-ent 0.11.3-5 (universe) from Debian unstable (main)
- Sync golang-oras-oras-go 2.5.0-1 (universe) from Debian unstable (main)
- Sync clblas 2.12-5 (universe) from Debian unstable (main)
- Sync das-watchdog 0.9.0-5 (universe) from Debian unstable (main)
- Reviewed and uploaded Merge ~liushuyu-011/ubuntu/+source/hfsutils:ubuntu/devel into ubuntu/+source/hfsutils:ubuntu/devel
- Reviewed and uploaded Merge ~liushuyu-011/ubuntu/+source/xfsdump:ubuntu/devel into ubuntu/+source/xfsdump:ubuntu/devel
+1 maintenance
- vowpal-wabbit rounding changes in Python 3.12 break unit tests:
Updating to a newer version (to see if that was fixed upstream) revealed major work. - devicexlib: Created bug report for the failures:
- django-assets: Fails to build with Python 3.12. I imported the bug from Debian and requested its removal from oracular-proposed.
- ubuntu-archive, is it possible to move a single binary package to multiverse (but keep the others in universe)? matlab-jnifti should be moved to multiverse (is in contrib on Debian). same for matlab-brain2mesh
- kissplice 2.6.2-3ubuntu1: Fix syntax warning on regexp (Closes: #1074658, LP: #2075171)
- ncbi-acc-download: Imported the existing bug report from Debian and started the analysis. It took quite some time to figure out that the newer requests relies on Content-Type to decode the request. The test should set the Content-Type header. Uploaded 0.2.8-2ubuntu1 with the fix and forwarded the fix to upstream and Debian.
- bdebstrap: I know that dude. Let me beat him to fix that failure. There was even a bug filed upstream today.
NVMe/TCP PoC
- Making more progress on automating the deployment of VMs and documenting the steps to take to run the PoC.
- In an effort to SRU https://bugs.launchpad.net/ubuntu/noble/+source/virt-manager/+bug/2073461 to noble, I ran into multiple issues that are causing the package to FTBFS in noble. The main problem is that python-argcomplete is broken on noble (see bug 2063129 and bug 2075315. I filed an SRU for python-argcomplete and submitted a patch (need a sponsor :))
- I opened a MP for virt-manager, but I marked it as an hold since it requires the python-argcomplete fix to land first.
distro
- Requested rebuild of netgen (and other packages) against mpi-defaults 1.17. The default implementation changed from openmpi to mpich on 32-bit arches.
Upcoming hardware
- Configure and package vendor kernel
- Build live installer image
Mid cycle sprint
Greenhouse
- Review applications
Distro
- Fixed LP: #2071815 in openssh: 1:9.6p1-3ubuntu18 : openssh package : Ubuntu
- Merge systemd (with some follow up fixes): 256.4-1ubuntu3 : systemd package : Ubuntu
- Another quick systemd merge to fix autopkgtest regressions: 256.4-2ubuntu1 : systemd package : Ubuntu
- Merged openssh (and uploaded a follow up fix): 1:9.7p1-7ubuntu2 : openssh package : Ubuntu
- Investigated https://bugs.launchpad.net/ubuntu/+source/snapd/+bug/2043820
- Trying to decide best way to fix it. Need to discuss with team.
- Investigated Bug #2055239 “Warning: The unit file, source configuration file ...” : Bugs : ubuntu-advantage-tools package : Ubuntu and tested a patch from upstream to fix it
- Planning to SRU this.
Sponsoring
- 2.3.6-1ubuntu1 : beancount package : Ubuntu for @adrien
- 4.12.1-2~exp1ubuntu5 : fence-agents package : Ubuntu for @adrien
- 2:24.0.0-0ubuntu3 : cinder package : Ubuntu for @adrien
- 2:2.7.2-2ubuntu1 : cryptsetup package : Ubuntu for @mkukri
- 24.10.3 : ubuntu-x13s-settings package : Ubuntu for @tobhe
Misc
- Conducting interviews
iso testing
- Lots of retriggering flaky tests to help with image promotion
- pinged flavors re cinnamon build failure
autopkgtest-cloud
- Deploying new charms after @uralt 's new changes
- Investigating issues with bos01-s390x
- Investigation into any issues and/or improvements to our statistics page
- Continuing work on moving from mojo to terraform
- Cherry picking recent changes to
autodep8
and placing them on theubuntu-release
master branch for usage inautopkgtest-cloud
- statistics analysis on the performance of
bos03-s390x
- seems pretty healthyat least much healthier than expected!
- Fixing an issue with
bos03-s390x
not being able to reachsalsa
hardware-installer-testing
- Wrote and tested the OEM installation testcase - all is well
- Wrote testcase for TPM FDE installation - untested as of yet
- New changes to testflinger which makes the hardware-installer-testing jobs more production ready - previously was testing from first slide of installer, now I have all the resources required to run a job from any state, download and flash iso, and boot into live session!
- This also required a bit of refactoring in my repository, there will be quite a few changes in the coming days but we’re looking close to production ready
+1 maintenance
- submitted a patch upstream for
freerdp2
which fixes the recent failing tests
TCK on armhf/lxd
- Completed investigation of failures (2/2) in openjdk-21 FFI tests (thanks to help from @liushuyu-011). Both seem to be expected failures and have already been excluded.
- Investigation pending for an apparent template interpreter failure, related to volatile longs in Java.
Java OpenSSL FIPS Provider
- Minor improvements
- Setup a GitHub Action to run CI build and test
- Added copyright headers
OpenJDK CRaC
- Post-review work in progress. Thanks @vpa19771 for the initial review.
- Merged July openjdk security updates into openjdk-17-crac and openjdk-21-crac upstreams.
GraalVM
- Put up a braindump of a specification for GraalVM snaps
Distro
Merge task for user-setup.
Misc
Spec reviews
Visa appointment prep
Couple of weeks here as I was off in the middle:
- Worked on adding FIT support to u-boot-rpi
- More work on Boards Documentation
- Merged reference section into how-to section after discussion with @xypron and @rkratky
- Updated flash-images PR with some suggestions
- Added Nezha D1 PR
- Patch Pilot shift on Monday
- Merged simplified branch of the classic pi-gadget
- Also merged changes to enable gnome-initial-setup in oracular
- Did a little sphinx debugging for imagecraft
- Worked on u-boot merge from debian (LP: #2075352); merge now finished but dealing with FTBFS on armhf
- More investigation of build failures on arm64+raspi images
- It’s not a snapd issue (as first thought)
- … or an adduser issue (second try)
- … or a systemd issue (third time was notably not the charm)
- It is LXD failing to launch (useable) oracular containers under focal’s 5.4 kernel (works fine from jammy onwards under 5.15)
- And no, the latest LXD doesn’t fix the issue so either LXD needs fixing to work under 5.4, or the image builders need to be running some later
- Filed GitHub issue to track, and provided some further debug logs on request
- Interviews & take home tests
- Meetings
- Customer Pi compatibility work
.NET
- packaging .NET 9 for oracular
- drafting release process spec
- attended .NET security partners meeting
Documentation
- working on git-ubuntu command reference for the Ubuntu packaging guide
- working on how-to patch management for the Ubuntu packaging guide
Distro
- Bug #2073950 “Merge curl 8.9.0-1 from Debian unstable” : Bugs : curl package : Ubuntu
- Bug #2075338 “merge coreutils 9.4-3.1 from Debian unstable” : Bugs : coreutils package : Ubuntu
- closed Bug #416981 “coreutils: du(1) doesn't summarize correctly” : Bugs : coreutils package : Ubuntu
- closed https://bugs.launchpad.net/ubuntu/+source/coreutils/+bug/1450179
- closed Bug #1474519 “realpath fails with inexplicable error message “No...” : Bugs : coreutils package : Ubuntu
- closed Bug #2002428 “Autopkgtest failure due to moved testcase” : Bugs : coreutils package : Ubuntu
dotnet
- Attended the weekly .NET partners sync meeting.
- Investigating the behavior of updates to .NET under the new snap architecture and how the .NET Installer will handle and update systemd-mount unit changes in the content snaps.
- Ran autopkgtests on the stage 2 .NET ppc64el package to start SRU process, but decided to hold off on it until the next security servicing release lands in Aug 13th to avoid that version superseding the current one in case the SRU does not finish by then.
Documentation
Netplan
- Review: PR#494 - Add
netplan try
to netplan tutorial
Launchpad
- New docs set
SRU docs
- Figuring out Launchpad + ReadTheDocs integration
- Setting up docs checks through
lpci
Subiquity
- PR: #2037 - Update autoinstall reference with usage examples for a few different autoinstall keys
- Refresh PR: #1901 - use legacy behavior for the validation script in CI until we can fix existing bugs that are uncovered by this change
- Bug triage
Distro
autopkgtest-cloud
- Worked on a less brittle solution for
vrf
module loading on armhf workers - Replaced a broken LXD armhf remote: got bitten by the
core.trust_password
removal and worked around it (details on the bug) - Closed 32 various open bugs from our backlog, bringing the amount of opened bugs to a single page. Many of them were old and not relevant anymore, but some of them were actually fixed in the meantime.
- Made some stats on the boot time in bos03-arm64 for @bdmurray
- Fixed an issue with
autopkgtest-buildvm-ubuntu-cloud
ubuntu-dev-tools
- Updated my MP, further improving
import-bug-from-debian
(no new feature, just code quality)
curtin
- Checked the availability of nested virtualization on every datacenter/arch (not available for arm64 and s390x :/)
openjdk:
- reviewing openjdk-{17,21}-crac packaging
- working on lintian warnings cleanup, one more MR in progress
- prototyped openjdk ROCK
- did a small investigation on enabling -dbgsym packages for openjdk
-fno-omit-framepointer:
- enabled option for libio-dirent-perl
autopkgtest-cloud
- Added a new page for viewing the results of a single test by UUID
- Turned test UUIDs in the package results page into links to the new endpoint
autopkgtest
- Worked on improvements to the autopkgtest-buildvm-ubuntu-cloud UX, mainly focusing on not having to redownload an ISO only to waste it due to missing dependencies or lack of space.
- In a similar vein, implemented caching for downloaded images to stop redundant downloads.
Hi all, this is my first post here. I’m Ural and I recently joined the Ubuntu Release Management team as a Graduate Software Engineer. I’ve been mostly learning the ropes over the past month, but I’m excited to get to contribute more as time goes on.
openjdk:
openjdk-crac:
- reviewed crac-criu
OpenJDK ROCKS:
- raised MP to forward port openjdk slices to 24.40
- raised MP with a smoke test for openjdk-8-jre-headless slice
- working on openjdk-21 slices.
distro:
- raised removal bug for tools-nrepl-clojure after removing it from reverse build dependencies
- investigating boost vs cmake build failures on armhf.
misc:
- finished phoronix orchestrator