When a package is retired from an epel branch, it is not blocked from the corresponding epel-build branch. Please fix whatever script is doing this.
This has been happening for at least two years as shown in this epel issue. https://pagure.io/epel/issue/208
Here is an example of what things look like.
$ koji list-pkgs --show-blocked --package xxhash | grep epel xxhash epel8 ellert [BLOCKED] xxhash epel8-build ellert xxhash epel9 ellert [BLOCKED] xxhash epel9-build ellert
If we look at the koji history we see this:
Thu Nov 17 21:17:01 2022 package owner changed for xxhash in epel8 by releng [still active] Thu Nov 17 21:17:01 2022 package list entry for xxhash in epel8 updated by releng [still active] Thu Nov 17 21:17:02 2022 package owner ellert set for xxhash in epel8-build by releng [still active] Thu Nov 17 21:17:02 2022 package list entry created: xxhash in epel8-build by releng [still active] ... Mon Nov 28 21:16:53 2022 package owner changed for xxhash in epel9 by releng [still active] Mon Nov 28 21:16:53 2022 package list entry for xxhash in epel9 updated by releng [still active] Mon Nov 28 21:16:53 2022 package owner ellert set for xxhash in epel9-build by releng [still active] Mon Nov 28 21:16:53 2022 package list entry created: xxhash in epel9-build by releng [still active]
When do you need this? (YYYY/MM/DD) 2024/06/30
When is this no longer needed or useful? (YYYY/MM/DD) Never
If we cannot complete your request, what is the impact? This is affecting koschei, as well as other script/automations that utilize the -build tag. We will need to keep opening releng tickets asking for the packages to be blocked manually.
The problem here is that the package should not be added to the -build tag. It should only be inherited from the base tag.
So, the bug is that something is adding packages to the build tag. We need to find what is doing that to fix this.
Metadata Update from @phsmoura: - Issue tagged with: low-gain, low-trouble, ops
Is this still happening?
As far as I can tell, since https://pagure.io/releng/issue/12192 was resolved, this is no longer happening. Packages I have retired in epel9 recently are processed correctly.
However, packages that were retired in epel9 earlier are still tagged into epel9-build while being blocked in the epel9 tag.
Packages being explicitly tagged with epel9-build seems to be an abnormal state - other epel9 packages are only "inherited" into epel9-build from epel9, as far as I can tell.
epel9-build
epel9
I did some digging, and finally figured out how to query koji for affected packages:
$ koji list-pkgs --tag=epel9-build --noinherit Package Tag Extra Arches Owner ----------------------- ----------------------- ---------------- --------------- lm_sensors-epel epel9-build tdawson q4wine epel9-build krege resalloc-openstack epel9-build praiskup rpkg epel9-build onosek cxxopts epel9-build xvitaly rust-exa epel9-build decathorpe rust-event-listener4 epel9-build decathorpe rust-event-listener-strategy0.4 epel9-build decathorpe
List of packages tagged into epel9-build that are retired in epel9 / blocked in the epel9 tag:
These all appear to have been retired in the epel9 branch in the last year.
List of packages tagged into epel9-build that are not retired in epel9 / blocked in the epel9 tag:
The first three appear to have been retired in epel9 at some point in the past but were unretired later. I can't figure out why rpkg would be in this state too, it's weird.
Would it be possible to untag at least the three packages where their state is obviously wrong (i.e. the three rust-* packages)?
(It looks like cxxopts was partially - but not fully - unretired: https://src.fedoraproject.org/rpms/cxxopts/commits/epel9 - so it should have been in the second list all along)
You mean block them?
I think rpkg was the same thing... retired and then unretired.
Yes - for the package where this is an obvious mistake that should be fine, right?
They're messing up my packager dashboard because they're showing up in koschei as failing to build.
Done.
With that... can we close this? All the rest of them should be ok?
At least the things I care about look fine now, thank you! The remaining packages seem to be in other weird states, so from my point of view the ticket can be closed :100:
Metadata Update from @kevin: - Issue close_status updated to: Fixed with Explanation - Issue status updated to: Closed (was: Open)