So, these updates didn't go to updates-testing like they should have. ;(

I tagged them in now, so they should go stable in tonights push.

@mattia any idea what happened here? freeze or gating?

Probably a failed compose that caused the update status and koji tags to be out of synch. Or some glitch with Koji that prevented Bodhi to move the builds from one tag to another.
I still haven't got any idea how to make the composer more reliable to prevent such behavior...

Also affecting my build perl-Compress-Raw-Bzip2-2.210-1.fc40, which has been in testing for 3 weeks now.

Ah, so I have a theory as to what happened here.

These updates all failed gating and then about a week later passed.

So, at that point bodhi saw that they were 'days to stable' ok, so requested stable.

I think there's somewhere bodhi should be checking 'is this update in testing' before 'should this autorequest stable now'

Does that theory seem to fit? Should I file a upstream bodhi issue on this?

I fixed perl-Compress-Raw-Bzip2-2.210-1.fc40 just now.

That does sound reasonable, as there's no "... was pushed to testing" comment in the update.

Might it also be possible for bodhi to check that a package that has been submitted for testing and not unpushed actually gets put in testing?

So, what happened here (I think) is:

• those were all automatic updates submitted before bodhi activation point. That means that the update is marked as pushed to testing as soon as all builds are signed, but the builds are not tagged for that because no testing repository exists.
• usually the update is pushed to stable and the builds are properly tagged as soon as tests pass. In these updates gating blocked the push to stable.
• F40 reached the bodhi activation point, so all updates must spend time in testing (and their builds need to be properly tagged) before being pushed to stable.
• gating tests of those updates were marked pass, but now we have builds not properly tagged.

Maybe we must properly tag all builds in updates still in testing state when we change the Release property create_automatic_updates to False? @adamwill any thoughts?

yeah...there's almost a kinda generic category here of Awkward Things That Happen When We Change Bodhi Configuration. See also https://bodhi.fedoraproject.org/updates/FEDORA-2024-3d80a949c6 , which - as of right now - has the "This update can be pushed to stable now if the maintainer wishes" comment but actually cannot be pushed stable unless it gets one more +1 or waits in testing for another two days; this is because after it got that message, we went from the "between Updates-testing Activation and Beta Freeze" policy to the "after Beta Freeze" policy.

I kinda feel like we need to wire in some phase on startup where we run actions to handle known config state transitions, or something...