hadoop-yarn-dev mailing list archives

Site index · List index
Message view « Date » · « Thread »
Top « Date » · « Thread »
From Andrew Wang <andrew.w...@cloudera.com>
Subject Re: DISCUSSION: Patch commit criteria.
Date Fri, 27 Feb 2015 22:27:27 GMT
I have the same interpretation as Konst on this. +1 from at least one
committer other than the author, no -1s.

I don't think there should be an exclusion for trivial patches, since the
definition of "trivial" is subjective. The exception here is CHANGES.txt,
which is something we really should get rid of.

Non-committers are still strongly encouraged to review patches even if
their +1 is not binding. Additional reviews improve code quality. Also,
when it comes to choosing new committers, one of the primary things I look
for is a history of quality code reviews.


On Fri, Feb 27, 2015 at 1:04 PM, Konstantin Shvachko <shv.hadoop@gmail.com>

> There were discussions on several jiras and threads recently about how RTC
> actually works in Hadoop.
> My opinion has always been that for a patch to be committed it needs an
> approval  (+1) of at least one committer other than the author and no -1s.
> The Bylaws seem to be stating just that:
> "Consensus approval of active committers, but with a minimum of one +1."
> See the full version under Actions / Code Change
> <http://hadoop.apache.org/bylaws.html#Decision+Making>
> Turned out people have different readings of that part of Bylaws, and
> different opinions on how RTC should work in different cases. Some of the
> questions that were raised include:
>  - Should we clarify the Code Change decision making clause in Bylaws?
>  - Should there be a relaxed criteria for "trivial" changes?
>  - Can a patch be committed if approved only by a non committer?
>  - Can a patch be committed based on self-review by a committer?
>  - What is the point for a non-committer to review the patch?
> Creating this thread to discuss these (and other that I sure missed) issues
> and to combine multiple discussions into one.
> My personal opinion we should just stick to the tradition. Good or bad, it
> worked for this community so far.
> I think most of the discrepancies arise from the fact that reviewers are
> hard to find. May be this should be the focus of improvements rather than
> the RTC rules.
> Thanks,
> --Konst

  • Unnamed multipart/alternative (inline, None, 0 bytes)
View raw message