On Mon, Jun 12, 2017 at 04:12:04PM -0700, Adam Williamson wrote:
Oh, I see the problem now, though it doesn't explain the '73
weeks
ago':
The 73 weeks relates to the state in pkgdb and I should have removed it
from the report. It is not useful here. :-/
[os-autoinst]
os-autoinst-openvswitch-4.4-18.20170410git97928a2.fc26.armv7hl requires
openvswitch
[os-autoinst]
os-autoinst-openvswitch-4.4-18.20170410git97928a2.fc26.ppc64 requires
openvswitch
so the problem is that the current stable openvswitch was not built for
two arches; the openvswitch package currently in updates-testing *is*
built for those arches, however. It seems pointless to send out an
update that disables the subpackage for two arches for a few days until
openvswitch is pushed stable.
As long as all updates are stable for the Final Freeze to fix the broken
deps there should be no problem.
Perhaps the proliferation of arches in primary Koji might cause a
rethink of how this automatic retirement works? It wouldn't seem to be
a good idea to automatically retire this package (which is in fact
quite important, and under active maintenance) on this basis.
It sounds like several broken deps will be resolved, soon, so there will
be a different picture. Since we only do up to two cleanups a year for
broken deps, it would a good chance to at least add proper
ExcludeArch/ExclusiveArch tags to packages that we cannot fix for the
Final Freeze. If this is not possible I am sure we will find a solution
that does not require to retire well-maintained packages.
Kind regards
Till