On Wed, Feb 28, 2018 at 12:49:09PM -0600, Richard Shaw wrote:
On Wed, Feb 28, 2018 at 12:38 PM, Daniel P. Berrangé
<berrange(a)redhat.com>
wrote:
> On Wed, Feb 28, 2018 at 01:12:03PM -0500, Matthew Miller wrote:
> > On Wed, Feb 28, 2018 at 05:57:53PM +0000, Daniel P. Berrangé wrote:
> > > mistake that caused files to go missing, and was never detected by the
> person
> > > making the change, because of the use of globs. So I agree it is good
> practice
> > > to explicitly list files without globs whereever it is practical todo
> so. I'd
> > > make an exception for files which don't have functional impact eg
> don't list
> > > 1000 HTML files individually, but it is always worth listing
> everything in
> > > /usr/bin, and /usr/lib(64) explicitly without globs.
> >
> > I used to agree with this, but I've come around to thinking that spec
> > files should be smaller, less complicated, and more automatable. I
> > think we'd be better having a post-build test warning that this package
> > has files missing from the previous build. That could be advisory, or
> > it could even gate, with the packager clearing the gate by updating the
> > file list in the test, rather than in the spec file.
>
> The further down the workflow a problem is detected the more time expensive
> / disruptive it is to fix it. So while having post-build tests to validate
> lots of things is great (and I wish we had more of it in Fedora), I see it
> as complementary to anything that we can do to detect problems earlier. I
> rather see failures right away when I test the new RPM build locally, than
> waiting to push it through koji and wait again for post-build tests to find
> the problem, as by that time I've context switched my mind away to a
> different bit of work.
I don't have the infra experience to implement, but what about adding
adding a pkgdiff option to fedpkg where it would complete a scratch build
(--srpm if necessary) and then run pkgdiff against it and the current
packages in the repository and putting the report somewhere accessible?
My typical scenario is to use a combination of "fedpkg local"
and "fedpkg scratch-build". So if there's a way to run tests on
the results of either or both of those, that could be a useful
thing.
Regards,
Daniel
--
|:
https://berrange.com -o-
https://www.flickr.com/photos/dberrange :|
|:
https://libvirt.org -o-
https://fstop138.berrange.com :|
|:
https://entangle-photo.org -o-
https://www.instagram.com/dberrange :|