openstack team mailing list archive

[Mark McLoughlin <markmc@xxxxxxxxxx>]

Hi Dave,

On Thu, 2011-11-10 at 17:33 +0000, Dave Walker wrote:
> On Thu, Nov 10, 2011 at 08:02:23AM -0800, James E. Blair wrote:
> <SNIP>
> > 
> > > But wait! Vish +2ed a stable branch patch yesterday:
> > >
> > >   https://review.openstack.org/328
> > >
> > > James, help a poor confused soul out here, would you? :)
> > >
> > > Right, that makes sense. Only folks that understand the stable branch
> > > policy[1] should be allowed to +2 on the stable branch.
> > >
> > > Basically, a stable branch reviewer should only +2 if:
> > >
> > >   - It fixes a significant issue, seen, or potentially seen, by someone
> > >     during real life use
> > >
> > >   - The fix, or equivalent, must be in master already
> > >
> > >   - The fix was either a fairly trivial cherry-pick that looks 
> > >     equally correct for the stable branch, or that the fix has 
> > >     sufficient technical review (e.g. a +1 from another stable 
> > >     reviewer if it's fairly straightforward, or one or more +1s from 
> > >     folks on core it it's really gnarly)
> > >
> > >   - If this reviewer proposed the patch originally, another stable
> > >     branch reviewer should have +1ed it 
> > >
> > > All we need is an understanding of the policy and reasonable judgement,
> > > it's not rocket science. I'd encourage folks to apply to the team for
> > > membership after reviewing a few patches.
> > 
> > It sounds like the best way to implement this policy is to give
> > openstack-stable-maint exclusive approval authority on stable branches,
> > and then make sure people understand those rules when adding them to
> > that team.  If that's the consensus, I can make the change.
> 
> Hi,
> 
> Thanks for helping to add clarification to this.  From our
> perspective, I have confidence that ~*-core members know the
> difference between trunk and stable policy.  Therefore for the short
> term, it makes sense to have more eyes - especially those which are
> likely to have good knowledge of the internals.
> 
> Therefore, I am happy for ~*-core to still have +2 access; especially
> if it helps seed the maint team.
> 
> Going forward, it probably will make sense to have a distinction, but
> I feel it might be quite early for that to be a requirement.

I basically said the same thing initially to Thierry on irc, but he
turned me around.

I'm not actually sure all folks on core do grok (or even want to grok)
the subtleties of the stable branch policy and the tradeoffs you need to
make when deciding whether to +2 some on the stable branch. Thierry has
had some similar experience with the milestone-proposed branch, I guess.

Also, I'm not even sure all folks on core always notice that a patch is
being submitted against stable, not master :)

But, of course, if anyone in core wanted to help with +2ing on the
stable branch, we'd add them to stable-maint in a flash.

Cheers,
Mark.