On Wed, Sep 9, 2020 at 1:34 PM Kevin Fenzi <kevin@xxxxxxxxx> wrote: > > On Wed, Sep 09, 2020 at 08:07:24PM +0200, Miro Hrončok wrote: > > Hello. > > > > I've noticed at https://src.fedoraproject.org/rpms/python3.9 that the latest > > ELN release of python3.9 is behind Fedora. > > > > I've assumed the build has failed, but it succeeded: > > > > https://koji.fedoraproject.org/koji/buildinfo?buildID=1592756 > > > > Except it it not tagged to eln. > > > > I've located the bodhi update: > > https://bodhi.fedoraproject.org/updates/FEDORA-2020-8e33e741d8 > > > > It is "obsoleted" with "This update cannot be pushed to stable. These builds > > python3.9-3.9.0~rc1-2.eln103 have a more recent build in koji's eln tag." > > > > I've checked the koji's eln tag and it has python3.9-3.9.0~rc1-1.eln103 now. > > > > Maybe some different build was tagged when the update was created (no idea > > how to tell). If it was the Fedora build, it indeed has higher > > version-release if that's what meant by "more recent" -- however there is no > > "more recent" (more recently started or more recently completed) build of > > python3.9 in Koji. > > > > This does not show anything suspicious: > > > > $ koji list-history --package=python3.9 | grep eln > > > > Is this expected? Should the update be edited and pushed again? Or should > > the build be tagged manually? Am I expected to deal with that, as the > > package maintainer? > > > > This particular build only fixes a minor user-facing issue and does not > > affect builds of dependent packages so we can "leave it be" and wait for the > > next build, however that might not always be the case. > > I guess this is a bug, but What I think happened is: > > f33 python3.9 finished: > Thu Aug 13 20:40:05 2020 python3.9-3.9.0~rc1-2.fc33 tagged into f33 by bodhi [still active] > ( eln was inheriting from f33-build at the time ) > > Then, the eln build finished: > Thu Aug 13 23:12:34 2020 python3.9-3.9.0~rc1-2.eln103 tagged into eln-updates-candidate > by bpeck/jenkins-continuous-infra.apps.ci.centos.org > > but at this point bodhi looks and sees... hey f33 which we inherit > from already has a build of this version, so lets not push this to > stable. > > I think bodhi needs to be a bit smarter about checking for "more recent > builds". Possibly not looking at inheritence? > > I could be wrong, but thats what it seems like from a quick glance... > It was a bit of a race condition. The automatic eln rebuilder got stuck, I'm not positive of the reason. But when it was restarted, it went through all of the package builds that it had missed, and built them. I'm not positive why, but if you look at the start times of the builds, python3.9-3.9.0~rc1-1.eln103 was started AFTER python3.9-3.9.0~rc1-2.eln103. By one minute. Since ...rc1-2 was started before ...rc1-1, bodhi assumed that the one that started it's build last, was supposed to be the one that gets merged. Anyway, bodhi was just doing it's job, and it did it correctly. I think we need to look at our automation and when it get's stuck, make sure the backed up builds it does are in build order. Troy _______________________________________________ devel mailing list -- devel@xxxxxxxxxxxxxxxxxxxxxxx To unsubscribe send an email to devel-leave@xxxxxxxxxxxxxxxxxxxxxxx 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@xxxxxxxxxxxxxxxxxxxxxxx
