What to do with a noarch package that fails to build on ARM

[Dan Horák]

On Tue, 20 Aug 2013 15:43:19 -0400
Josh Boyer <jwboyer at fedoraproject.org> wrote:

> On Tue, Aug 20, 2013 at 3:36 PM, Dan Horák <dan at danny.cz> wrote:
> > On Tue, 20 Aug 2013 14:20:12 -0500
> > Andrew McNabb <amcnabb at mcnabbs.org> wrote:
> >
> >> I maintain the python-pexpect package, which is noarch.  Koji is
> >> trying to build the Fedora 20 packages on ARM machines, and for
> >> some reason, one of the tests is failing.  I've opened a report
> >> with upstream, but neither the upstream maintainer nor I have
> >> access to any ARM machines to figure out what's going on.
> >>
> >> What's the right way to deal with this situation?  Is it possible
> >> to block the noarch package from the ARM architecture?
> >
> > the really dirty workaround is to try building so many times until
> > a x86 builder is used
> 
> The end solution to this is to actually build all packages on all
> arches and not have the secondary arches suck down completed builds
> from primary koji.  I realize this would increase the build load on
> those arches significantly, but it would catch these issues.

I think the ppc and s390x builders are capable of processing the regular
load (rawhide + updates), but they would be the bottleneck for the mass
rebuilds, especially on s390x our resources are limited


		Dan