David Woodhouse píše v Út 08. 06. 2010 v 13:52 +0100:
On Tue, 2010-06-08 at 14:09 +0200, Timo Schoeler wrote:
> -----BEGIN PGP SIGNED MESSAGE-----
> Hash: SHA1
>
> thus Josh Boyer spake:
> > On Tue, Jun 08, 2010 at 09:47:33AM +0100, David Woodhouse wrote:
> >> On Sat, 2010-06-05 at 13:39 -0400, Josh Boyer wrote:
> >>> On Sat, Jun 05, 2010 at 08:17:23PM +0300, Robert Hagan wrote:
> >>>> As somewhat of a novice at both Fedora and powerpc, does anyone
have
> >>>> news on when we may expect Fedora13 for powerpc to be available?
> >>> There is no ETA for Fedora 13 ppc/ppc64.
> >> I took a quick look at this. I was _hoping_ that it would be a simple
> >> case of doing a compose... but there are actually quite a lot of build
> >> failures.
> >
> > Yes, they have grown over the past couple of months. I tried reporting build
> > failures in bugzilla with limited success and I don't have time to fix
them
> > all at the moment.
> >
> >> Is there any tool to make sense of these and find the original failure
> >>from which the others cascade, and to resubmit the dependent failures?
> >
> > No. Other than the output from koji-shadow, which isn't really accessible
or
> > easy to read over.
> >
> >> Is there a generic "Fedora secondary arches" mailing list where
such
> >> things should be discussed?
> >
> > secondary(a)lists.fedoraproject.org
> >
> >> Let's pick the gcc-4.4.4-5.fc13 build failure as the first one to
> >> investigate.... it failed in deps due to not finding /usr/lib64/libc.so.
> >>
> >> Is this because the glibc64 package isn't tagged in
> >> SHADOWBUILD-dist-f13-build? Perhaps because Jakub never did a build of
> >> glibc64 for F-13? Can we remedy that?
> >
> > This is because koji-shadow is dumb and didn't bring in the glibc64
package.
> > It will likely work if submitted by hand.
> >
> >> Can we get the build repositories rsynced onto bombadil so that we can
> >> use mock manually there?
> >
> > The repos for the tags? Not sure. My bandwidth is the limiting factor for
> > most of this and I'm hoping to get new hardware placed into the PXH2
datacenter
> > in not horribly long.
How much bandwidth does it take to upload the built packages? Surely you
can't build them faster than you can upload them? :)
How much CPU power do you have there, anyway?
> Hi,
>
> I'm interested in this port (as well as I'm interested in a CentOS
> PowerPC port), but I don't have too much spare time left to be able to
> do some work here.
>
> However, I do have some hardware (both an IBM Power 285 and 275) as well
> as internet connectivity to give, as long as it is within an acceptable
> scale (I work for an ISP and have some colo space).
>
> So, if I can be of any help hosting stuff on x86 machines, please tell
> me. I could furthermore set up one of the Power machines, however it'd
> surely take some time.
The issue is getting the build horsepower and the hosting in the same
place, I think. I'm not sure if those two boxes would suffice to do the
work that jwb's cluster is doing... but if so, perhaps we could cope if
you were to just do a basic install to each of them and connect them.
from my experience in Fedora/s390x most bandwidth is needed for the hub,
the builders (or a group of builders) can use a cache and be in a
different location, it works quite well for us
Dan