[Libreoffice-qa] Gerrit auto-merge

[Bjoern Michaelsen]

On Sat, Jun 02, 2012 at 08:54:42AM +0200, Lionel Elie Mamane wrote:
> On Fri, Jun 01, 2012 at 03:08:55PM +0200, Michael Stahl wrote:
> > On 01/06/12 15:01, Norbert Thiebaud wrote:
> >> On Fri, Jun 1, 2012 at 5:24 AM, Michael Meeks <michael.meeks at suse.com> wrote:
> 
> >>> * Quality of round-robin patch review (Markus)
> >>>        + often generalists review specific patches for master
> >>>        + gerrit should be able to help (Norbert)
> >>>                + queueing and checking them
> >>>                + one-day timeout from gerrit: if no protest,
> >>>                  auto-merge to master.
> 
> >> For the record: I did not suggest, nor do I support that last point.
> 
> > iirc that auto-merge was only for commits from authors who have commit
> > rights anyway, not for patch submissions from new and inexperienced
> > people which is what the topic was about.
> 
> Ah, that's very different. So essentially people that now have
> immediate gratification^W commit rights would "go back" to "one day
> delay"?

AFAIK we so far only talk about master. I think there is a common understaning
the  review process for backports to the release branch is good and works and
we will continue to use it as-is on gerrit (saving us some mail fuddling and
manual patch tweaking).
People who had commit rights now will have commit rights to master in gerrit
too. So they can push directly to master bypassing review just like before.
Just remember that everyone will look very angry at you if your
one-line-change-that-could-not-possibly-break-anything breaks something and it
causes annoyances to everyone because you were impatient to give it a change
for review in one day.
But I dont think "immediate gratification" is that important for people who are
longer on the project as for them it is business as usual anyway. It the first
time contributors, who send their patch and then check every 2 minutes if its
in yet.
In addition, people with commit rights usually knwo their peers well -- if you
pair up with somebody who is also working in your area of code reviewing can be
really smooth (the requested reviewer gets an email: If you make a deal to
review each others commits faster, you can be even quicker).

So -- people with commit rights are not the issue:
- can commit directly to master on their own responsibility
  (this should be discouraged, except for urgent buildbreaker fixes that save
   everyone pain)
- can submit patches for review. Since we intend to assign a "patch pilot" each
  week, his patch will be on master after <24 hours, unless there are questions
  about it (in which care it is sane to hold back until this is clarified)
- if you want to be faster, team up with someone for mutual review and you can
  be as fast as you want

people without commit right:
- will have a reply by the patch pilot in <24 hours and his patch in when it
  looks innocent in the same time. If there are questions about it integration
  can take a bit longer, but I think to keep contributors motivated, the first
  response time is critical, not so much the time to integration.

Best,

Bjoern