I've been giving +2s for stuff I think is ready to commit, but I always
assumed that Googlers would have no superpowers here, and that Steering
Committee members and other designated people with specialized knowledge
could do so as well. Google doesn't automatically import changes without
reviewing them, and so our internal heuristic is something like "if
importing changes from gerrit, if there isn't a +2 from a Google employee,
we need to take a deeper look."

For example, John giving +2 to I18N or OOPHM changes seems ideal to me. I
actually wouldn't feel comfortable giving more than a +1 on I18N unless
John was unavailable and it needed to get landed because of a critical
issue.



On Mon, Jun 10, 2013 at 9:35 AM, Thomas Broyer <t.bro...@gmail.com> wrote:

>
> On Monday, June 10, 2013 5:12:34 AM UTC+2, Stephen Haberman wrote:
>>
>> Hey,
>>
>> I wasn't quite sure, so figured I would ask: what's our policy on who
>> can give +2s? Is it any committer?
>>
>
> Technically, yes:
> https://gwt-review.googlesource.com/#/admin/projects/gwt,access (if you
> can see it; restricted access)
> I don't know who exactly is a "committer" though. Of non-Googlers, at
> least John (jat) and me.
>
>
>> I assume so, but perhaps out of tradition, I've been assuming Googlers
>> would give +2s.
>>
>> Well, and even if we are/have moved past the Google-only phase of GWT,
>> for now/awhile they will still have the most knowledge about the
>> codebase (perhaps sans Thomas).
>>
>
> I try to not use +2 too much (or at least not submit reviews) because the
> CI server currently doesn't run tests against real browsers, and doesn't
> run tests at all "pre submit".
> Maybe we should do something like OpenStack:
> http://ci.openstack.org/gerrit.html They have a third category "Approved"
> in addition to "Verified" and "CodeReview", when set to "approved" then the
> CI server is triggered to do a build, and updates the Verified score. This
> is so that they can control the charge put on their CI server (only code
> that "looks OK" is built; but it's not worth scrutinizing the code if it'd
> break, so they don't wait the CodeReview+2).
> I think we could have our first "checkstyle" check done automatically for
> all reviews (like today, when it works ;-) ) but only put Verified-1 if it
> fails, and trigger a full build with tests only with an Approved+1 (or
> CodeReview+2 if that's easier to setup). Ideally, the build could just run
> JVM tests, and we (committers) could trigger browser tests (HtmlUnit only,
> or in real browsers) if needed.
> But that's not easy to setup (and someone would have to do it), so for now
> I "submit" sparingly (unless Googlers tell us to go ahead): for now, tests
> are run by Google when master is merged on google/pu, and they do reverts
> when something breaks.
>
> --
> http://groups.google.com/group/Google-Web-Toolkit-Contributors
> ---
> You received this message because you are subscribed to the Google Groups
> "GWT Contributors" group.
> To unsubscribe from this group and stop receiving emails from it, send an
> email to google-web-toolkit-contributors+unsubscr...@googlegroups.com.
> For more options, visit https://groups.google.com/groups/opt_out.
>
>
>

-- 
http://groups.google.com/group/Google-Web-Toolkit-Contributors
--- 
You received this message because you are subscribed to the Google Groups "GWT 
Contributors" group.
To unsubscribe from this group and stop receiving emails from it, send an email 
to google-web-toolkit-contributors+unsubscr...@googlegroups.com.
For more options, visit https://groups.google.com/groups/opt_out.


Reply via email to