Horton

Re: Testing LISTEN/NOTIFY more effectively

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: 2019-07-27T22:20:52Z
Lists: pgsql-hackers

Attachments

I wrote:
> Andres Freund <andres@anarazel.de> writes:
>> Perhaps we could just have isolationtester check to which
>> isolationtester session the backend pid belongs? And then print the
>> session name instead of the pid? That should be fairly easy, and would
>> probably give us all we need?

> Oh, that's a good idea -- it's already tracking all the backend PIDs,
> so probably not much extra work to do it like that.

I found out that to avoid confusion, one really wants the message to
identify both the sending and receiving sessions.  Here's a patch
that does it that way and extends the async-notify.spec test to
perform basic end-to-end checks on LISTEN/NOTIFY.

I intentionally made the test show the lack of NOTIFY de-deduplication
that currently happens with subtransactions.  If we change this as I
proposed in <17822.1564186806@sss.pgh.pa.us>, this test output will
change.

			regards, tom lane

Commits

  1. Improve test coverage for LISTEN/NOTIFY.

  2. Don't drop NOTICE messages in isolation tests.

  3. Fix isolationtester race condition for notices sent before blocking.