Code review required for commits - Discuss.

Jeremy Allison jra at
Thu Oct 11 14:43:10 MDT 2012

On Thu, Oct 11, 2012 at 01:33:43PM -0700, Matthieu Patou wrote:

> To my french mind the wording seems a bit confusing, are you saying that:
> 1) every author could (should?) add their own sign-off
> 2) in total 2 sign-off are need, which means at least 1 non-author
> review (but can be two if the author decide not to sign-off it's
> patch)


> >Release branches work as they do today, though the + should be treated
> >as a sign-off in bugzilla.
> Why-not but should it be automatic ?

I don't think we need any change to what we're currently
doing in release branches - this is just writing up what
we already have. The reason it can't be automatic I think
is that a patch gets posted to bugzilla, other people +1
it then Karolin commits. For a + to add a signoff people
would have to re-upload the patch with their signoff added,
and I don't think we need that extra step.

> >All bug commits should now contain their
> >bug number in them, so we can track back what happened.
> Agreed.

We've mostly been doing that anyway.

> In general I'm mostly ok with this proposal but we have to aware
> that will increase the workload on some of us.

Yes, this is true.

> Also I want to have precision on how to handle the failure in
> autobuild and I'd like to see a kind of watchdog so that patches
> can't wait more that xx days for review.

Hopefully peer pressure will act as a watchdog here, plus
pressure from the patch author.

> Finally as we did so far we can try this rule on a voluntary basis
> (it was the case for autobuild too) and see how it flies for 1 or 2
> months.

I'm not sure this is enforcible by automation (git experts please chime in
here :-) so it'll be by Team convention anyway.


More information about the samba-technical mailing list