Horton

Re: Buildfarm failure and dubious coding in predicate.c

Tom Lane <tgl@sss.pgh.pa.us>

View thread View on postgr.es
From: Tom Lane <tgl@sss.pgh.pa.us>
To: Thomas Munro <thomas.munro@enterprisedb.com>
Cc: Pg Hackers <pgsql-hackers@postgresql.org>
Date: 2017-07-24T19:24:01Z
Lists: pgsql-hackers 
Thomas Munro <thomas.munro@enterprisedb.com> writes:
> Ahh, I think I see it.  This is an EXEC_BACKEND build farm animal.
> Theory: After the backend we see had removed the scratch entry and
> before it had restored it, another backend started up and ran
> InitPredicateLocks(), which inserted a new scratch entry without
> interlocking.

Ouch.  Yes, I think you're probably right.  It needs to skip that if
IsUnderPostmaster.  Seems like there ought to be an Assert(!found)
there, too.  And I don't think I entirely like the fact that there's
no assertions about the found/not found cases below, either.

Will fix, unless you're already on it?

			regards, tom lane

Commits

  1. Fix race condition in predicate-lock init code in EXEC_BACKEND builds.

  2. Improve comments about partitioned hash table freelists.