releng/tickets

Fork 2

Multiple builds for same package pushed in initial F44-updates push and the older build won (2) #13332

New issue

Open

opened 2026-04-29 00:16:43 +00:00 by gotmax23 · 6 comments

gotmax23 commented

2026-04-29 00:16:43 +00:00

Contributor

With some not-very-good code, I went through the f44-updates tag and checked if there were any other packages (besides python3.14 from #13331) where the newest NEVRA was not the latest tagged build and found one more.

ruby-build-20260327-2.fc44 is tagged instead of ruby-build-20260412-1.fc44. Can you please fix that one as well?

Also, we should probably track the larger problem here or in Bodhi's tracker to figure out how to prevent this from happening in the future.

With some not-very-good code, I went through the `f44-updates` tag and checked if there were any other packages (besides python3.14 from https://forge.fedoraproject.org/releng/tickets/issues/13331) where the newest NEVRA was not the latest tagged build and found one more. `ruby-build-20260327-2.fc44` is tagged instead of `ruby-build-20260412-1.fc44`. Can you please fix that one as well? Also, we should probably track the larger problem here or in Bodhi's tracker to figure out how to prevent this from happening in the future.

gotmax23 referenced this issue

2026-04-29 00:17:23 +00:00

Multiple builds for same package pushed in initial F44-updates push and the older build won #13331

jnsamyak self-assigned this

2026-04-29 02:53:31 +00:00

jnsamyak commented

2026-04-29 02:56:19 +00:00

Owner

koji latest-pkg f44-updates ruby-build 
Build                                     Tag                   Built by
----------------------------------------  --------------------  ----------------
ruby-build-20260327-2.fc44                f44-updates           mikelo2

I see the latest one in f44-updates is the newer build only

``` koji latest-pkg f44-updates ruby-build Build Tag Built by ---------------------------------------- -------------------- ---------------- ruby-build-20260327-2.fc44 f44-updates mikelo2 ``` I see the latest one in f44-updates is the newer build only

jnsamyak added the

labels

2026-04-29 02:58:00 +00:00

jnsamyak added this to the Sprint ( 2026-04-20 to 2026-05-03 ) project

2026-04-29 02:58:15 +00:00

jnsamyak commented

2026-04-29 03:00:27 +00:00

Owner

koji latest-pkg f44-updates ruby-build                       
Build                                     Tag                   Built by
----------------------------------------  --------------------  ----------------
ruby-build-20260412-1.fc44                f44-updates           mikelo2

Should be fixed now; IF you see more, please reopen this ticket thank you.

``` koji latest-pkg f44-updates ruby-build Build Tag Built by ---------------------------------------- -------------------- ---------------- ruby-build-20260412-1.fc44 f44-updates mikelo2 ``` Should be fixed now; IF you see more, please reopen this ticket thank you.

jnsamyak closed this issue

2026-04-29 03:00:27 +00:00

gotmax23 commented

2026-04-29 03:28:52 +00:00

Author

Contributor

Thank you for fixing it! Can we keep one of the tickets open (or make a new one) to track a permanent fix? The root cause is a race condition in Bodhi where, in certain cases, Bodhi allows creating two conflicting updates for the same package, and then if both updates are in testing -> stable at the same time, the older update might supersede the newer one. This is exacerbated by long updates freezes where testing -> stable pushes don't happen and these conflicting updates can accumulate. We should either fix Bodhi or create a releng process/script to identify these cases and retag the correct build.

Thank you for fixing it! Can we keep one of the tickets open (or make a new one) to track a permanent fix? The root cause is a race condition in Bodhi where, in certain cases, Bodhi allows creating two conflicting updates for the same package, and then if both updates are in `testing -> stable` at the same time, the older update might supersede the newer one. This is exacerbated by long updates freezes where `testing -> stable` pushes don't happen and these conflicting updates can accumulate. We should either fix Bodhi or create a releng process/script to identify these cases and retag the correct build.

👍 1

gotmax23 commented

2026-04-29 16:06:56 +00:00

Author

Contributor

Can we keep one of the tickets open (or make a new one) to track a permanent fix?

I'll just reopen this ticket for now. Feel free to open a new one and close this one again if that's preferred.

> Can we keep one of the tickets open (or make a new one) to track a permanent fix? I'll just reopen this ticket for now. Feel free to open a new one and close this one again if that's preferred.

gotmax23 reopened this issue

2026-04-29 16:06:56 +00:00

kevin commented

2026-04-29 16:24:17 +00:00

Owner

Well, my solution/idea on how to handle this is https://github.com/fedora-infra/bodhi/issues/1266 but it never seemed to get anywhere...

jnsamyak removed their assignment

2026-05-12 17:35:03 +00:00

jnsamyak added

and removed

labels

2026-05-12 17:35:17 +00:00

jnsamyak commented

2026-05-12 17:36:24 +00:00

Owner

In order to open a new ticket with a much neater description, I have put this to backlog, once there is a new ticket, we will close this one, or we can just reuse this one with proper action item.