exras build system

[Dan Williams]

On Mon, 2005-10-24 at 17:11 +0200, Eric TANGUY wrote:
> > On Mon, 2005-10-24 at 15:24 +0200, Eric TANGUY wrote:
> >> Is there any problem on the build system ?
> >
> > Hmm, mock wasn't noticing that its child process had terminated on one
> > of the builders...  kicked now.
> >
> > Dan
> >
> Ok thanks but there is some things i don't understand :
> http://buildsys.fedoraproject.org/build-status/job.psp?uid=208 this job is
> said failed on x86_64 but if i look at the logs the jobs finished fine in
> i386 and x86_64 arch but not in ppc ...

I've requeued all the jobs that said they failed because they couldn't
download the RPMS from hammer3.  Other jobs that used hammer3 look like
they got through fine after I kicked stuff earlier, so I think we're OK.

> If i look at http://buildsys.fedoraproject.org/build-status/success.psp i
> don't understand why there is package in needsign status and others in
> add-to_repo status.

Job 25 is an anomoly, it needs to be bumped and pushed through again...
Not sure why that one didn't go through last week.  The rest should be
fine.  Jobs can't move from add_to_repo -> needsign until the
'createrepo' command has been run on the build server.  'add_to_repo'
means that the jobs are in the queue for the createrepo run.

It takes a while to run createrepo because we need to wait for all
builders to leave the 'prepping' state before we can run createrepo,
because createrepo touches the repo (of course) and the builders are
reading from the repo during their 'prepping' phase.  If we allow the
write+read at the same time, jobs fail because the repo is inconsistent.

Dan