synced python-testtools to fix failure in (build time) test-suite (causing FTBFS when building without nocheck) (thanks @ginggs for sponsoring this sync too!)
submitted debdiff for lvm2 to fix FTBFS against autoconf 2.72 (sponsor needed if we want to land this for oracular :))
fixed a number of stupid mistakes in my gsasl and libgcrypt20 MRs for FTBFS and @enr0n sponsored them
investigated test issues with gsasl on armhf but didn’t succeed; my main suspect is glibc but several months ago unfortunately
Misc
Studied the python cryptographic libraries ecosystem, especially for TLS; they tend to set their own default cipher list which obviously ruins any consistency set at the distribution levelr. It looks like changes are only needed in python’s ssl and the rust openssl bindings (mostly used by pyopenssl; yes, that’s quite an unexpected detour!)
Studied the gnutls28 autopkgtest which doesn’t use the upstream testsuite mechanism but a home-grown, which sometimes leads to surprises and makes reporting bugs upstream more difficult. Patches in progress.
Worked on improving the CI on the meta-release files, to prevent another accident like this to happen again.
auto-upgrade-testing
Got that back on track after the loss of venonat, our good old Jenkins runner. Got replaced by @paride, but needed a bit of tuning to get everything to work.
autopkgtest
Reviewed and tested an MR by @paride, adding support for adding PPA easily.
Obviously with version number shenanigans when Noble and Oracular have the same version.
You apparently cannot promote a package that is in $release-updates, but also still in $release-proposed. The cleanup is not automatic.
For a simple MIR at first glance, I definitely learned a lot in the process! And now we’ll finally be able to SRU autopkgtest to Jammy and Noble with long awaited bug fixes!
Error Tracker
Fixed and deployed two missing lines for noble and oracular. This was a good way of investigating how to actually deploy code in production in the error tracker.
Explored how this whole thing is wired. Started a personal note to track the no less than 7 repositories that hold the code, excluding apport and whoopsie.
Got a ping from @ballot to see about cleaning the Cassandra DB storing the data. I still need to level up with all these scripts, tables, and data everywhere, but I’ve figured something.
nbd-client autopkgtests: fix MP up, but turns out that resizing /tmp on canonical autopkgtests infrastructure doesn’t work. Looking for workarounds (like changing cwd) or will just ask QA to increase the /tmp size
Investigating the inhibition of growroot-almost during first boot
Making rpi-imager snap to bump it to v1.9
Got the FTBFS MPs merges
Pi planning with @waveform
Had a public holiday on 2nd Oct.
Continue to try and replicate the tmpfs bug, which has proven to be some trouble. There are some other lingering issues w/r/t /tmp being a tmpfs, with prior assumptions about the fs type no longer being valid.
Before /tmp was a tmpfs, the directory was cleared “manually” at boot by the system. Now that the default is different /tmp does not get cleared, which is at odds with some assumptions autopkgtest makes about the system.
Some misc improvements with dependencies and chasing down a possible dependency bug in autopkgtest-build-lxd.
autopkgtest-cloud
Still moving the statistics calculations out of the webserver itself into a systemd unit, but requirements are better defined now and an MP is forthcoming.
Misc
Working on getting access to the engineering bastion to add some firewall rules.
Made a first-pass change to the git-ubuntu CI using @rbasak’s snap-other-archs.py script to build git-ubuntu for non-amd64 systems.