Horton

Re: Failing assertions in indxpath.c, placeholder.c and brin_minmax.c

Michael Paquier <michael.paquier@gmail.com>

View thread View on postgr.es
From: Michael Paquier <michael.paquier@gmail.com>
To: Andreas Seltenreich <seltenreich@gmx.de>
Cc: PostgreSQL mailing lists <pgsql-hackers@postgresql.org>
Date: 2015-07-27T01:50:17Z
Lists: pgsql-hackers 
On Sun, Jul 26, 2015 at 10:32 PM, Andreas Seltenreich <seltenreich@gmx.de>
wrote:

> Michael Paquier writes:
>
> >> Footnotes:
> >> [1]  https://github.com/anse1/sqlsmith
> >
> > This is really interesting stuff. I think that it would be possible to
> > extract self-contained test cases from your tool and those queries to
> > reproduce the failures. It is written that this tools connects to a
> > database to retrieve the schema, what is it exactly in the case of
> > those failures?
>
> I used the database "regression" that pg_regress leaves behind when you
> remove the --temp-install from it's default invocation through make
> check.  Sorry about not being explicit about that.
>
> So, dropping one of the queries into src/test/regress/sql/smith.sql and
> invoking
>
>     make check EXTRA_TESTS=smith
>
> was all that was needed to integrate them.  I was then able to perform
> "git bisect run" on this command.  Er, plus consing the expected output
> file.
>
> I'm using the regression db a lot when hacking on sqlsmith, as it
> contains much more nasty things than your average database.
>

Ah, OK. Thanks. The code is licensed as GPL, has a dependency on libpqxx
and is written in C++, so it cannot be integrated into core as a test
module in this state, but I think that it would be definitely worth having
something like that in the code tree that runs on the buildfarm. We could
have caught up those problems earlier.  Now I imagine that this is a costly
run, so we had better have a switch to control if it is run or not, like a
configure option or a flag. Thoughts?
-- 
Michael

Commits

  1. Fix broken assertion in BRIN code

  2. Support flattening of empty-FROM subqueries and one-row VALUES tables.

  3. Fix resource leak pointed out by Coverity.

  4. Add missing volatile qualifier.

  5. Remove dependency on wsock32.lib in favor of ws2_32

  6. Check interrupts during logical decoding more frequently.

  7. Fix contrib/pg_upgrade/test.sh for $PWD containing spaces.

  8. Secure Unix-domain sockets of "make check" temporary clusters.

  9. Adjust 9.4 release notes.

  10. Improve predtest.c's ability to reason about operator expressions.

  11. Fix pg_restore's processing of old-style BLOB COMMENTS data.

  12. Improve tuplestore's error messages for I/O failures.

  13. Remove unnecessary output expressions from unflattened subqueries.

  14. Stamp 9.4beta1.