Investigated bug report regarding /usr/bin/dotnet being a relative symlink. Turned out to be a won’t fix.
distro
Investigating nvme-stas autopkgtests on different architectures, including iputils unit test failing on s390x.
Seeded the nvme-stas package on mantic as part of the MIR process. Currently waiting on the dasbus MIR to go through security review before promoting nvme-stas to main.
Code review of @andersson123’s MP which revives unit tests in autopkgtest-cloud.
Merged an MP from Zixing adding a couple of node packages to big_packages.
Submitted an MP modifying the stages at which pre-commit is run.
Submitted an MP modifying autopkgtest-web such that the duplicate request check is not used with github requests. This is change that was cowboy’ed because systemd-upstream was encountering issues with it. A proper fix is being tracked in a separate card.
Ubuntu
Further modifications to crichton, a bug bot, so that it can run on a system in PS5. Additionally, tested and moved the code for the bug bot to that system.
Confirmed bug 2032647 when doing FDE installer testing using the Mantic canary image.
Reported bug 2032677 which I observed after a fresh install of Mantic.
Reported bug 2032701 regarding there not being release notes about FDE install option.
Sync’ed memtest86+ from Debian as it’ll resolve bug 2031420 which is causing our automated dist-upgrades to fail.
Misc
Finished @vorlon’s conversion (which was started in 2019) of lpltk to python3, this was necessary to get arsenal running with python3 in PS5.
submitted a patch to debhelper to automate Provides: for 64-bit time_t libs
worked with mwhudson to fix duplication of all the snaps between squashfs layers in the canary image; successfully brought image size down from 5.7G to 4.0G
some additional triage of dhclient migration work for 23.10, identified that avahi-autoipd is almost certainly useless in our current network stack and should be unseeded
consulted on cloud-minimal seeds wrt currently having a hard dependency on systemd-timesyncd that disallows users to replace it with chrony
discussion with IS about the state of porter boxes (canonistack)
Archive
Added gtk3-nocsd to the i386 whitelist in response to a misfiled bug report I happened across
assisted with bootstrap of versioned rustc-1.68 package for mantic
assisted with a bootstrap of several node packages for +1 maintenance
further discussion around fortune-mod in stable releases
Infra
spent a whole day debugging why rmadison was not working, learning in the end that rsync --delete is unreliable without the addition of --ignore-errors and therefore the ubuntu-archive-team frontend’s disk had filled up
SRU
SRU day on Friday, not very successful overall and didn’t make it all the way through the queues
Review of a new proposed exception policy for ubuntu-advantage-tools (still in progress)
golang-defaults has been updated to golang-1.21 and migrated.
Implement scripts to filter out packages built by old Go toolchains. mwhudson helps me to run these scripts to rebuild packages with golang-1.21.
I’m watching the build results and proposed migrations for these packages.
Fix packages FTBFS with golang-1.21 in Debian and ask for syncing to Ubuntu.
ubuntu-zfs-enhance-support.patch: Adjustments for 2.12 library (LP: #2029260)
zfs: on_exit: Unmount ${MNTDIR}/boot before ${MNTDIR} (LP: #2031042)
Temporarily rmmod peimage for os-prober chainloader entries (LP: #2030810)
And adding deb822 ubuntu sources support to software-properties (and python-apt’s aptsource’s distro module). API unaligned with deb822 format makes it annoyingly frustrating.
Spent time running ML-like performance benchmarks on ppc64el and s390x, for the evaluation of an upstream JIT compiler performance patch, followed by some number crunching. Surprisingly (or may be not) the gains on s390x are decent [3-10%], but those on ppc64el aren’t very exciting [1-3%], with some benchmarks showing negative gains on both archs. We used 17.0.8 as the baseline.
Created MR 449637 to propose man pages support for the openjdk-fips package. I am still working on the review comments and expecting a significant change to the MR.
I began prototyping an openssl-based JNI wrapper library with FIPS support. The goals of this prototyping activity are (a) learning the openssl crypto API and (b) creating reusable code for some prospective FIPS work. This week, I’ve been prototyping basic functionalities of a simple DRBG (Deterministic Random Bit Generator) wrapper.
Proposed Migration
Spent an hour or so looking at the sudo proposed migration block and realized that the one autopkgtest regression (on armhf/arm64) had been already sorted out and the package is ready for migration (waiting on glibc only). Thanks to @ginggs
djbdns: investigated and fixed a problem with the new glibc in the tinydns binary on all archs and tests timeouts on arm*. (sponsored and reported to Debian).
vim: trying to fix the build on ppc64el.
Netplan
Rebased the netplan diff work and did some refactoring after the support for the new python bindings was merged. All the changes required by netplan diff were also merged.
Writing a testsuite to test upgrades to Mantic and check if network-manager+netplan will not break.
Network Manager
Found a bug on Jammy preventing veth pairs to be activated. Also prepared an MP and PPA with a fix from upstream. This bug is affecting Netplan’s CI on Github LP: #2032824PR#403
Investigated an issue affecting autopkgtests on the new version of NM cause by a change in behavior. NM will not install routes anymore if the interface doesn’t have an IP address. LP: #2031421
Upload 0.142ubuntu11 to use dracut-install in manual_add_modules to speed up mkinitramfs significantly (LP: #2031185) and harmonize kernel module installation with dracut’s behavior (LP: #2031841)
Upload 0.142ubuntu12 to change dracut-core dependency to newly created dracut-install package (LP: #2031304)
figure out how to test proposed fix for LP: #2031576 - simulate what livecd-rootfs does, use a ‘dangerous’ model to allow updating the pc-kernel to the special edge channel version
subiquity:
PR: #1768 - relax required models to not need network in some cases
PR: #1769 - upload my bind-patch test script for modifying a snap in the live environment
help debug issues with desktop iso - find model config issue, testing help, identify source of imaging failure
LP: #2032961 - Investigate IRC reported issue with late-commands. They need to be run with a clean environment.
proposed-migration
p11-kit: on tip from Steve, file LP: #2032576 to fix build issues with p11-kit, which gnutls28 needed. fixed and uploaded.
util-linux: run some migration tests on mir to clear the last regressions
gnuttls28: investigate test problems on i386. My initial tests were done in schroot, but how do we debug i386 issues on the wiki is more accurate. Device a fix for using the wrong architecture openssl and for not running 64 bit time_t tests (dpkg-architecture gives amd64 answers in this environment) and upload.
You test the canary images but the canary is missing in http://iso.qa.ubuntu.com/qatracker/milestones/446/builds.
I would test canary ISO if I knew what they are for and where to put the results.
Why do you test the canary ISO and not the daily ones?
Thanks for the interest in helping us to test this image. The canary images are intended to be short-lived development branches for the engineering team to test out new features before landing them in the regular Desktop image. In this case, the current canary contents are expected to land in the regular Desktop image in the coming week. We therefore will not be spending time to expose these images as an option on the ISO tracker, it will instead be possible to test the new functionality when it lands.