Horton

Re: Improving the isolationtester: fewer failures, less delay

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

View thread View on postgr.es
From: Tom Lane <tgl@sss.pgh.pa.us>
To: Andres Freund <andres@anarazel.de>
Cc: pgsql-hackers@lists.postgresql.org
Date: 2021-06-15T21:09:00Z
Lists: pgsql-hackers

Attachments

I wrote:
> I'll have a v2 in a little bit --- the cfbot pointed out that there
> were some contrib tests I'd missed fixing, and I found a couple of
> other improvements.

Here 'tis.  This passes check-world, unlike v1 (mea culpa for not
checking that).  I also cleaned up the variant expected-files,
so it's now no worse than HEAD as far as failures in serializable
mode go.

I played a bit more with insert-conflict-specconflict.spec, too.
It now seems proof against delays inserted anywhere in the
lock-acquiring subroutines.

			regards, tom lane

Commits

  1. Use annotations to reduce instability of isolation-test results.