On Thu, Dec 01, 2022 at 04:51:18PM +0000, Sérgio Basto wrote:
On Thu, 2022-12-01 at 13:21 +0800, yanqiyu(a)fedoraproject.org wrote:
> Hi all,
>
> It seems update fcitx5-qt-5.0.16-2.fc37 from [1] somehow did not make
> it to stable repository while other package in same update did. This
> caused broken dependences during update [2].
>
> Seems that bodhi is doing something weird. I might have an idea how
> this happened.]: I decided to update fcitx5-qt to 5.0.16 and created
> a
> update [3] containing fcitx5-qt-5.0.16-1.fc37, during [3] waiting to
> make it to stable, qt6 rebuild happens and created another update
> with
> fcitx5-qt-5.0.16-2.fc37 [1]. [1] gets a lot karma and pushed to
> stable
> soon. And when [3] gets pushed afterwards, somehow fcitx5-qt-5.0.16-
> 1.fc37 overrides fcitx5-qt-5.0.16-2.fc37.
>
IMO , when fcitx5-qt-5.0.16-2.fc37 went to bodhi should have obsoleted
fcitx5-qt-5.0.16-1.fc37 .
It couldn't, because that update was in an update with other packages.
In that case, bodhi isn't sure what to do with those other packages.
I'm proposed in this case to just reject the new update and force the
maintainer(s) to sort this out. By removing the 'older' update or
editing the other packages into a newer update or something.
But if fcitx5-qt-5.0.16-1.fc37 was already marked got to stable ,
was
not obsoleted , and a race condition might happened .
It wasn't stable, but bodhi won't obsolete a update with multiple builds
in it. ;(
In resume I think is a bug
> I don't know if this behavior should be considered as a bug, but
> anyway, it seems that manual override from releng is needed to fix
> the
> broken dependencies.
Yep. I've retagged the -2 one in and it should be fixed in the next
compose.
kevin