Horton

Re: postgres_fdw bug in 9.6

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

View thread View on postgr.es
From: Tom Lane <tgl@sss.pgh.pa.us>
To: Jeff Janes <jeff.janes@gmail.com>
Cc: pgsql-hackers <pgsql-hackers@postgresql.org>
Date: 2016-12-08T17:50:50Z
Lists: pgsql-hackers 
Jeff Janes <jeff.janes@gmail.com> writes:
> I have a DML statement which triggers the error:
> ERROR:  XX000: outer pathkeys do not match mergeclauses
> LOCATION:  create_mergejoin_plan, createplan.c:3722

Hmm.

> Any tips on investigating this further in situ?  Or is the best option just
> to work harder on a minimal and disclosable test case?

I think we need a test case --- not minimal necessarily, but something
other people can reproduce.  You might find that setting enable_hashjoin
and/or enable_nestloop to false makes it easier to provoke the error,
since evidently this requires that we (a) generate a faulty mergejoin Path
and then (b) choose it as the cheapest one, since the error occurs while
converting it to a Plan.

BTW, if you're not doing this in a debug (--enable-cassert) build, it'd
be useful to try it in one.  I'm a little suspicious that the root cause
might be a memory-stomp type of problem, ie somebody scribbling on a
pathkey data structure without accounting for it being shared with another
path.  It's possible that cassert memory checking would help catch that.

			regards, tom lane

Commits

  1. Fix test case for 'outer pathkeys do not match mergeclauses' fix.

  2. postgres_fdw: Avoid 'outer pathkeys do not match mergeclauses' error.

  3. postgres_fdw: Consider foreign joining and foreign sorting together.

  4. Allow foreign and custom joins to handle EvalPlanQual rechecks.

  5. Allow FDWs to push down quals without breaking EvalPlanQual rechecks.