Horton

Re: Tuple sort is broken. It crashes on simple test.

Peter Geoghegan <pg@heroku.com>

View thread View on postgr.es
From: Peter Geoghegan <pg@heroku.com>
To: Pavel Stehule <pavel.stehule@gmail.com>
Cc: Michael Paquier <michael.paquier@gmail.com>, Mithun Cy <mithun.cy@enterprisedb.com>, pgsql-hackers <pgsql-hackers@postgresql.org>, Robert Haas <robertmhaas@gmail.com>, Heikki Linnakangas <hlinnaka@iki.fi>
Date: 2017-01-16T18:43:59Z
Lists: pgsql-hackers 
On Mon, Jan 16, 2017 at 10:38 AM, Pavel Stehule <pavel.stehule@gmail.com> wrote:
> Should not be enhanced regress tests too?

We already have coverage of multi-pass external tuplesorts, as of a
few months back. That didn't catch this bug only because it was a
pass-by-value datum tuplesort. The relevant RELEASE_SLAB_SLOT() call
does have line coverage already.

I wouldn't object to adding a test case that would have exercised this
bug, too. It took me a while to talk Tom into the test that was added
several months back, which discouraged me from adding another test
case here. (There were concerns about the overhead of an external sort
test on slower buildfarm animals.)

-- 
Peter Geoghegan

Commits

  1. Fix NULL pointer dereference in tuplesort.c.

  2. Change the way pre-reading in external sort's merge phase works.