Horton

Substantial bloat in postgres_fdw regression test runtime

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

View thread View on postgr.es
From: Tom Lane <tgl@sss.pgh.pa.us>
To: pgsql-hackers@postgreSQL.org
Date: 2016-11-02T16:39:25Z
Lists: pgsql-hackers 
In 9.6, "make installcheck" in contrib/postgres_fdw takes a shade
under 3 seconds on my machine.  In HEAD, it's taking 10 seconds.
I am not happy, especially not since there's no parallelization
of the contrib regression tests.  That's a direct consumption of
my time and all other developers' time too.  This evidently came
in with commit 7012b132d (Push down aggregates to remote servers),
and while that's a laudable feature, I cannot help thinking that
it does not deserve this much of an imposition on every make check
that's ever done for the rest of eternity.

			regards, tom lane

Commits

  1. Improve speed of contrib/postgres_fdw regression tests.

  2. Don't make FK-based selectivity estimates in inheritance situations.

  3. postgres_fdw: Push down aggregates to remote servers.