On Mon, Apr 23, 2012 at 09:09:56AM -0600, Kevin Fenzi wrote:
On Mon, 23 Apr 2012 11:05:59 -0400
Josh Boyer <jwboyer(a)redhat.com> wrote:
> On Mon, Apr 23, 2012 at 10:58:21AM -0400, Tom Callaway wrote:
> > On 04/23/2012 10:52 AM, Niels de Vos wrote:
> >
> > > Well, it is something that all fedora packages could use, it is
> > > not restricted to tke kernel package.
> > >
> > > Maybe fedpkg should create a tag when building or filing an
> > > update. CC'ing jkeating who seems to be the fedpkg contact (see
> > >
https://fedorahosted.org/fedpkg/) for his thoughts.
> >
> > I'm not sure it can do that safely, because only a _successful_
> > build merits a tag, and fedpkg is done once it hands off a build to
> > koji.
>
> Why is that exactly? If an official build is submitted to koji and it
> fails, it's still an official build. You cannot submit a new build
> with the same NVR as koji will disallow that. A tag is just a
> reflection of an official build (or could be construed as such
> anyway).
You cannot submit a new build that completed successfully.
You can resubmit a build with the same N-V-R until it works.
ie,
submit 3.3.3-1, it fails, you tweak some things, commit, don't change
the n-v-r and resubmit it works.
Hm... then my memory is getting fuzzy. I was almost certain it would
not let you do that. If it does, I'm kinda disappointed.
Which git hash has the 3.3.3-1 tag? It should be the one that
finished,
but if you tagged when you first submitted that N-V-R it would be
wrong.
> A build submitted and canceled does run into trouble though. Maybe
> that is what you were referring to?
Thats another case too.
Right, and that is why it probably has to be a manual post-build
operation to tag it anyway. Before someone mentions a koji plugin, I
think there is a strong desire to NOT let the buildsys be able to write
to the SCM, so that is also out.
josh