Horton

Re: pgsql: Add support for coordinating record typmods among parallel worke

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, Andres Freund <andres@anarazel.de>
Cc: Thomas Munro <thomas.munro@enterprisedb.com>, pgsql-committers <pgsql-committers@postgresql.org>
Date: 2017-09-15T04:29:48Z
Lists: pgsql-hackers 
Andres Freund <andres@anarazel.de> writes:
> Sorry for missing that during review - unfortunately I don't have a computer with me now - so I won't get around to this till tomorrow...

It turns out that this breaks my local build, too, so I went ahead and
pushed a fix.

At least two buildfarm members think there's something deeper broken
here:

2017-09-15 00:08:34.061 EDT [97092:75] LOG:  worker process: parallel worker for PID 101175 (PID 101256) was terminated by signal 11: Segmentation fault
2017-09-15 00:08:34.061 EDT [97092:76] DETAIL:  Failed process was running: select * from pg_get_object_address('operator of access method', '{btree,integer_ops,1}', '{int4,bool}');

I wonder if this indicates you forgot to consider transmission of state
to parallel worker processes.

			regards, tom lane

Commits

  1. Get rid of shared_record_typmod_registry_worker_detach; it doesn't work.

  2. Add support for coordinating record typmods among parallel workers.

  3. Add minimal regression test for blessed record type transfer.