Re: [PATCH v3 29/30] gitlab: limit re-builds of the containers

[Daniel P . Berrangé]

On Mon, Jun 29, 2020 at 01:42:19PM +0200, Thomas Huth wrote:
> On 29/06/2020 12.11, Alex Bennée wrote:
> > 
> > Daniel P. Berrangé <berrange@redhat.com> writes:
> > 
> > > On Fri, Jun 26, 2020 at 07:13:56PM +0100, Alex Bennée wrote:
> > > > Most of the time we are just rebuilding the same things. We can skip
> > > > this although currently there is no mechanism for picking up new
> > > > distro releases.
> > > > 
> > > > Rather than try to be too fine grained allow any change to trigger all
> > > > the images being rebuilt.
> > > > 
> > > > Signed-off-by: Alex Bennée <alex.bennee@linaro.org>
> > > > ---
> > > >   .gitlab-ci.d/containers.yml | 4 ++++
> > > >   1 file changed, 4 insertions(+)
> > > > 
> > > > diff --git a/.gitlab-ci.d/containers.yml b/.gitlab-ci.d/containers.yml
> > > > index 20b2fb1de5d..f56aa44f711 100644
> > > > --- a/.gitlab-ci.d/containers.yml
> > > > +++ b/.gitlab-ci.d/containers.yml
> > > > @@ -19,6 +19,10 @@
> > > >       - docker push "$TAG"
> > > >     after_script:
> > > >       - docker logout
> > > > +  only:
> > > > +    changes:
> > > > +      - .gitlab-ci.d/containers.yml
> > > > +      - tests/docker/*
> > > 
> > > How does this work for a person who forks the QEMU git repo and pushes
> > > a change which doesn't touch the containers.yml file ?  AFAICT, all
> > > their jobs will fail due to not having previously built any container
> > > images in their brand new fork.
> > 
> > Hmm what we really need is a condition check to see if there is a local
> > registry with images in it.
> 
> Maybe it would be good enough to do something like this instead:
> 
>   rules:
>     - if: '$CI_COMMIT_BRANCH == "master"'
>     - changes:
>       - .gitlab-ci.d/containers.yml
>       - tests/docker/*
> 
> ... so that the pipeline is certainly run the first time when the master
> branch gets initialized? (I did not try it out, just had the idea after
> looking at the gitlab CI docs)

IIUC, the rules: can only reduce scenarios where a job is run, vs the
baseline config. AFAIK, forking a repo doesn't cause any jobs to be
run, so adding extra rules: can't make that happen.

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 :|