Re: postgres_fdw: perform UPDATE/DELETE .. RETURNING on a join directly

On Thu, Feb 8, 2018 at 11:05 AM, Tom Lane <tgl@sss.pgh.pa.us> wrote:
> Etsuro Fujita <fujita.etsuro@lab.ntt.co.jp> writes:
>> (2018/02/08 10:40), Robert Haas wrote:
>>> Uggh, I missed the fact that they were doing that.  It's probably
>>> actually useful test coverage, but it's not surprising that it isn't
>>> stable.
>
>> That was my purpose, but I agree with the instability.  Thanks again,
>> Robert!
>
> According to
>
> https://buildfarm.postgresql.org/cgi-bin/show_log.pl?nm=rhinoceros&dt=2018-02-08%2001%3A45%3A01
>
> there's still an intermittent issue.  I ran "make installcheck" in
> contrib/postgres_fdw in a loop, and got a similar failure on the
> 47th try --- my result duplicates the second plan change shown by
> rhinoceros, but not the first one.  I speculate that the plan change
> is a result of autovacuum kicking in partway through the run.

Hmm.  Maybe inserting an ANALYZE command in the right place would fix it?

-- 
Robert Haas
EnterpriseDB: http://www.enterprisedb.com
The Enterprise PostgreSQL Company