Re: Patches that don't apply or don't compile: 2017-09-12

Aleksander Alekseev wrote:

> Agree, especially regarding build logs. All of this currently is only an
> experiment. For some reason I got a weird feeling that at this time it
> will be not quite successful one. If there will be too many false
> positives I'll just return the patches back to "Needs Review" as it was
> before. But I strongly believe that after a few iterations we will find
> a solution that will be good enough and this slight inconveniences will
> worth it in a long run.

Thinking further, I think changing patch status automatically may never
be a good idea; there's always going to be some amount of common sense
applied beforehand (such as if a conflict is just an Oid catalog
collision, or a typo fix in a recent commit, etc). Such conflicts will
always raise errors with a tool, but would never cause a (decent) human
being from changing the patch status to WoA.

On the other hand, sending an email to the patch author (possibly CCing
the mailing list, incl. In-Reply-To headers as appropriate) notifying
them that there is a conflict would be very useful. I think it would be
perfectly reasonable to automate that. Include exactly what went wrong,
and of course the date where the problem was detected (in the email
headers) so that reviewers can see "this patch's author received a
notice and didn't act on it for two weeks" and take a decision to set it
WoA.

--
Álvaro Herrera https://www.2ndQuadrant.com/
PostgreSQL Development, 24x7 Support, Remote DBA, Training & Services