Welcome back web readers! Here we’ll post our updates from the past week here. Excelsior!
Last week’s update is available here: Foundations Team Updates - Thursday 31 March 2022
We are two weeks out from release, which means it’s important to clean up all the remaining binaries in the release pocket that are no longer built from source. This generally involves looking at the reverse-dependencies listed on https://people.canonical.com/~ubuntu-archive/nbs.html and resolving why they have not been rebuilt to drop dependencies.
- mozjs78: noticed lto-wrapper listed in the output of the builds on failing architectures, then noticed the architectures where it was built successfully are the archs where we don’t use LTO. Disabled use of LTO in the package, then also fix the test target to invoke the js interpreter from the binary install directory instead of from the source directory, which for some reason is not marked executable - we can ignore this instead of debugging the upstream build system. This unblocks removal of libicu67.
- tracked down the sra-sdk autopkgtest failure with the new ncbi-vdb to be a misbuild of sra-sdk itself with LTO, not a bug in ncbi-vdb; rebuilt sra-sdk without LTO and both packages migrated, unblocking removal of mbedtls NBS binaries
- Finally took a look at the cluster of blocked D library packages, to notice that 3 of the 4 were candidates but britney wasn’t working out on its own how to migrate them together. Added a hint for this and the 3 migrated. Then removed diet-ng, which is also FTBFS in Debian and not in testing. This let us remove the old libphobos binaries.
- As of Tuesday, we have zero NBS binaries in the release pocket for jammy!
- Looked at gyoto/ppc64el, one of 3 packages which block removal of python3.9. Failing to build because of mutually-incompatible options being passed to g++ when LTO mode is on; but these flags don’t come from the build of the gyoto package, instead they are hard-coded into the lorene library, causing the
check-lorene test to fail. Worked on unpicking this, so that impossible options were not being inherited from liblorene.
- skimming the bottom of update_excuses, noticed that two related packages,
libperinci-sub-normalize-perl, had been blocked in -proposed for 170 days due to one of the packages failing its autopkgtests; but somehow no one had ever tried triggering the autopkgtest with both packages together? (+1 maintenance?) Did so now, and the packages migrated successfully.
- it was pointed out that mozjs78 previously built on i386 but was now not being scheduled there. This led to a round of cleanup of old i386 binaries in the release pocket
- got a request to clean up particular NBS binaries in the -proposed pocket, so went through update_excuses again and removed all NBS binaries I found; this should let around ~14 packages (mostly kernels) migrate
- misc push-button NEW processing of packages synced from Debian
- now that NBS is empty, also worked on zeroing https://people.canonical.com/~ubuntu-archive/proposed-migration/jammy_uninst.txt and https://people.canonical.com/~ubuntu-archive/proposed-migration/jammy_outdate.txt which required some further i386 work
- discovered in the process that the libpth20 i386 binary went missing in jammy by my hand despite having reverse-dependencies in the archive that should have prevented it being picked up as a candidate for removal. Best guess is that there was a window when there were not any reverse-dependencies in the jammy release pocket, and therefore it got removed. Binary is back now, though, and the count of uninstallables on i386 is now zero, which it hadn’t been for a few cycles.
- pestered fellow AAs to clean up their manual whitelist entries for packages on i386 (which would ideally only be used during bootstrapping, with things eventually moving to the seed).
- started trimming down the priority mismatches for jammy release. These are less important in the past now that we don’t use debian-installer which relies on their correctness, but it is a good review to make sure our base system is not being unexpectedly bloated.
process-removals for all packages removed from Debian unstable since the start of 2021, picking up a few more removals of obsolete packages that are no longer blocked from removal in Ubuntu due to reverse-dependencies.
- fixed remaining references to exfat-utils so it could be removed cleanly.
- Processed package removal requests from https://bugs.launchpad.net/ubuntu/+bugs?field.subscriber=ubuntu-archive&orderby=-id&start=0
- Noticed via an open ubuntu-archive bug that numba made it back into the archive, so resurrected all of the source packages to -proposed that were removed because of lack of numba. One or more of these probably needs follow-up to be migratable to the release pocket, but now it’s on the +1 maintenance todo list instead of being lost in the ether.
- along the way, noticed that some packages were stuck in -proposed due to spurious autopkgtest "regressions’ reported on i386. Did a sweep of these, got 8 packages to migrate out of -proposed.
- Spent a little time cleaning up stale hints files in proposed-migration. We are now down to only 72 hints total in the devel release, and no force-reset-test hints left, so I will look at deprecating this delta from Debian.
- Discussion with enr0n, jawn-smith about use of britney hints vs baseline autopkgtest retesting. Rejected an MP for a hint because the retest proved that snapd autopkgtests had regressed in the impish-updates pocket, so no manual override was required; thus unblocking a systemd SRU.
- Discussion about systemd-oomd behavior regressing experience in certain situations where memory pressure was not causing issues, and possible reversion for jammy release
- Technical board meeting with discussion about third-party-archive policies for flavors. Ran into an issue (RT#149245 filed by @sil2100) that prevents logins to wiki.ubuntu.com.
- Went through the process of booking travel to attend the roadmap sprint in Copenhagen in the beginning of May, first trip in 2 years so having to jog my memory about how to do things!
Awesome job. I would like to thank everyone for all of your
hard work and dedication for Ubuntu, and the determination
too make Ubuntu more…!