[Date Prev][Date Next]   [Thread Prev][Thread Next]   [Thread Index] [Date Index] [Author Index]

Re: Mock Configs [Was: Re: Plan for tomorrows EPEL SIG meeting]

On Wed, 2007-06-13 at 21:29 -0500, Dennis Gilmore wrote:
> Once upon a time Wednesday 13 June 2007, Michael E Brown wrote:
> > On Wed, Jun 13, 2007 at 11:56:46AM -0400, Jeff Sheltren wrote:
> > > On Jun 13, 2007, at 11:45 AM, Michael E Brown wrote:
> > > >>Hi Michael, maybe we can get this decided in the meeting today.  I'll
> > > >>send an updated EL5 mock config to you and the list once we determine
> > > >>exactly what we need to exclude :)
> > > >
> > > >Quickly. I have 0.7.1 ready to go as we hit a bug in 0.7.0. We will be
> > > >releaseing 0.7.1 sometime today in the next couple hours.
> > > >
> > > >Current configs that were posted a couple weeks ago (sans exclude=)
> > > >are
> > > >current staged for 0.7.1.
> > >
> > > Hi Michael, if you have to push a 0.7.1 today, I'd lean towards
> > > simply copying the exclude line from a fedora x86_64 mock config and
> > > re-using that in the EL5 x86_64 config.  That "works for me", and
> > > it's an improvement over the behavior without having an exclude line.
> > >
> > > If anyone sees any problem with this, please respond soon so Michael
> > > can get the update pushed :)
> >
> > Will do.
> >
> > Let me know if you guys want something different and I'll get it ready
> > for 0.7.2.
> the exclude is because some packages require glibc.i386  in the 64 bit chroot 
> to build.  gdb gcc and a few others 

Actually ... (3) i[3,4,5,6]86 things are potentially required in the
x86_64 tree for building x86_64 RPMS: glibc.i386, glibc.i686,
glibc-devel.i386.  That exclude line does allow for all those to get
called in when required, but blocks all other i386 arch content.  It
might not be "required" for EPEL (if you don't build any packages that
need both 32 and 64 bit parts), but it is certainly required for Fedora
and CentOS/RHEL base OSes if building in mock. 

Also, if koji uses the same "Stage" repo functionality as plague (I am
not sure how koji does its "generated" repos {or even if it does self
generated repos}), you might also need to exclude x86_64 rpms in the
i386 mock configs.  (In plague the Staged feature {at least in our
setup) does one set of repo metadata that contains all i386, x86_64, and
all other arches.  This means that when solving deps, all arches can be
looked at and potentially called out of arch)

Again, this may or may not also be how koji does staged repos ... just
food for thought.

Johnny Hughes

Attachment: signature.asc
Description: This is a digitally signed message part

[Date Prev][Date Next]   [Thread Prev][Thread Next]   [Thread Index] [Date Index] [Author Index]