Thanks for the excellent explanation, Adam. It makes me to understand the problem now.
Jiri On Fri, Dec 2, 2022 at 10:03 PM Adam Williamson <adamw...@fedoraproject.org> wrote: > On Fri, 2022-12-02 at 21:25 +0100, Jiri Kucera wrote: > > Yes, builds in [1] were built with the `f38-build-side-60497` side tag. > In > > [1] there are two errors that were not here in time I hit the submit > button > > (maybe I should wait a bit longer): > > * `nothing provides libqgpgme.so.7 needed by > > kdepim-addons-22.08.3-1.fc38.i686` - this one was > > caused by not building `kdepim-addons` on `i686` since missing > > `libphonenumber` on `i686`. > > `libphonenumber` is not built for `i686` anymore due to `ExclusiveArch: > > %{java_arches}`. This > > can be fixed by skipping building the Java binding for `i686` only. > > * ``` > > Undeclared file conflicts: > > kleopatra-*.i686 provides ... which is also provided by > kleopatra-*.x86_64 > > ... > > kmail-*.i686 provides ... which is also provided by kmail-*.x86_64 > > ... > > ``` > > These must have appeared also in the update before, but I cannot find > > `rpmdeplint` tests > > here: https://bodhi.fedoraproject.org/updates/FEDORA-2022-e8d23a026e > > > > I submitted [2] approx. 22h after [1] became stable. Have no idea why the > > builds from pre-[1] rawhide were picked up. However, `rpmdeplint` > > repoclosure failures are happening only on `i686` so maybe this is > somehow > > connected with `kdepim-addons` not built for `i686`. > > > > Regards and sorry for the chaos, > > Jiri > > > > [1] https://bodhi.fedoraproject.org/updates/FEDORA-2022-4c1b011b1b > > [2] https://bodhi.fedoraproject.org/updates/FEDORA-2022-603eea89a3 > > Ahhh, I see what happened! > > So, the problem is this. When a Rawhide update "goes stable" in Bodhi, > all that really *means* is, it gets the release tag applied (so 'f38' > at the moment). The next time the buildroot repo is composed after that > (which happens frequently), the packages from the update will be added > to it. > > However, the packages from the update will only show up in the main > 'rawhide' repository after the next successful Rawhide compose, which > is an uncertain interval. One compose is run per day automatically, so > if those all succeed, you'd be sure of an update making it to 'rawhide' > no more than 24 hours after it reached 'stable'. But sometimes they > fail. When they fail, we might fix the problem and try again, or it > might magically go away with the next day's compose, or we might not > get a successful compose for a while. > > Right now, we haven't had a successful compose for several days due to > various problems, so the current packages in the main 'rawhide' repo > are the ones from the last successful compose, which I think was > 20221130.n.0. Nothing that's gone "stable" since then is actually in > the repo. > > openQA update tests for Rawhide updates use the latest packages from > the main 'rawhide' repo, plus the packages from the update being > tested. They do *not* include packages from the buildroot repo. > > So in the openQA tests, the first update - with all the builds in it - > passed the tests just fine. But the second update, which only had gpgme > in it, failed, because it didn't include the rebuilt dependent packages > from the first update, even though they were now "stable". > > Overall I'd say this isn't your problem as the packager; everything you > did was totally reasonable. In theory what you "should have done" to > make the tests pass is wait till a Rawhide compose had completed before > submitting the second update, but obviously that's not a reasonable > thing to ask. It's more of a problem for me and releng to think about. > > I may think about having openQA Rawhide update tests enable the > buildroot repo that includes packages from the release tag; this would > make it include packages that have gone 'stable' since the previous > Rawhide compose. I'd have to think if there are any potential drawbacks > to doing that. Ironically, Fedora CI currently does that but is > considering *not* doing it any more due to "unpleasant surprises": > https://pagure.io/fedora-ci/general/issue/376 . I'm not sure exactly > what the surprises were, I'll have to look into it. > -- > Adam Williamson > Fedora QA > IRC: adamw | Twitter: adamw_ha > https://www.happyassassin.net > > >
_______________________________________________ devel mailing list -- devel@lists.fedoraproject.org To unsubscribe send an email to devel-le...@lists.fedoraproject.org Fedora Code of Conduct: https://docs.fedoraproject.org/en-US/project/code-of-conduct/ List Guidelines: https://fedoraproject.org/wiki/Mailing_list_guidelines List Archives: https://lists.fedoraproject.org/archives/list/devel@lists.fedoraproject.org Do not reply to spam, report it: https://pagure.io/fedora-infrastructure/new_issue