El jue, 19-10-2017 a las 19:29 +0200, Florian Weimer escribió: > On 10/19/2017 06:42 PM, Dennis Gilmore wrote: > > > There is long standing 8+ year old bugs we have never gotten around > > to > > to ensure multilib is correct in all cases. It is unfortuantly > > expected > > sometimes, it probably has happened a lot over the yearsm, but > > since > > defaulting to turning off installing multilib by default we have > > not > > had and issues filed in years, until you asked earlier this year. > > we > > may be able to resolve the issues with the move to use pungi for > > updates pushes rather than mashing the repos. > > I'm not sure if it's the same bug. Based on the reports, it feels > like > as if every other compose is broken. Either we have been > exceedingly > unlucky the past week or so, or we're hitting some new issue. > > Thanks, > Florian Florian,
updates does not have composes it has mashes, the behaviour is expected to be different, the problem boils down to the fact that if a package is made multilib indirectly, i.e. it is only pulled in because some external package that is multilib requires it, it will only be multilib in updates or updates-testing if the package requiring it is also in the same repo. the way we make the updates and updates-testing repos make the repo and does multilib in isolation, without any consideration for what was multilib in the base repo, as it has no knowledge it even exits, we have had a bug open for ~8 years to fix and until this year when you raised an issue about it had not had any reports of issues since we defaulted to not installing multilib by default. Dennis
signature.asc
Description: This is a digitally signed message part
_______________________________________________ devel mailing list -- devel@lists.fedoraproject.org To unsubscribe send an email to devel-le...@lists.fedoraproject.org
