Click-click-click on test retries, trying to be mindful of the test queues (not overloading, nor letting them empty), which I guess was useful useful this week
Reporting issues with test runners and following their status (thanks a lot to @hyask, @andersson123 and @juliank)
Took part in the archive unstucking; in particular but not limited to:
forensics-extra-* which testsuite is only to install all its depends (and recommends?), which is useful for the packager but probably a terrible test at the distribution level: a single issue with Qt and linssid became an opaque error for zlib
report-tracker4/5 for which one test fails and which has blocked dozens of perl packages but the use of all-proposed on all arches makes it difficult to track: Test failures in noble-proposed (all-proposed)
follow-up for the removal of some R packages which are marked as not supporting 32-bit arches anymore: triggered migration-reference/0 test in order to make the revdep test non-regressions
overall, many test retries which have been made much easier by my rewritten excuse page
Discussion with Helmut about the details of abi-compliance-checker analysis and its output
Tabular and predictable layout Status filters! Architecture filters! The date of tests directly visible on the page without hovering over links! Links to the launchpad pages of the tests! Statistics! Re-written status messages that aren’t just API leaks from Britney but actually make sense!
I opened PR #1954 - Which is about catching a common error where users pass autoinstall config in cloud-config directly and continuing to make error cases more obvious to users
I opened PR #1956 - Which is about standardizing the accepted autoinstall formatting. Directly relates to the above, as we want to encourage users to use the cloud-config delivery format everywhere to avoid that issue.
Follow up work to get PRs #1945, #1946, and #1947 from last week merged
Subiquity bug triage
Distro
I participated in the package analysis for time_t hinting. I didn’t think to count them like everyone else did! I should go back and do that.
Lots of queue cleanup and fire fixing, particularly looking into amd64 throughput and killing forever looping tests
MP for hotfixing all noble tests to be all-proposed, then later reverting (it’s not merged or deployed yet)
Fixing logic in the worker code to disable retries
MP adding more fail strings to better recognise obscure test failures
testing a fix to the lxd metrics and the way they’re collected
made an MP with two scripts, pull-amqp and push-amqp, which should make it easier for us to move tests between queues or re-queue with new args
MP for adding a login button to the navbar of auc
MP for adding a user-specific page, per request of @schopin (nearly done, just needs review)
MP for fixing display of exit code 14 results, as alerted to us by @schopin
ubuntu-release-metrics
MP to change the cloning process of the repo in production - is now cloned by the ubuntu user and admins can do git stuff.
In the same MP, I fixed the charm itself. The charm branch hasn’t been updated in years and I made the charm building and deploying process a bit more up to date.