Horton

Re: failures in t/031_recovery_conflict.pl on CI

Andres Freund <andres@anarazel.de>

View thread View on postgr.es
From: Andres Freund <andres@anarazel.de>
To: Tom Lane <tgl@sss.pgh.pa.us>
Cc: Robert Haas <robertmhaas@gmail.com>, pgsql-hackers@postgresql.org, Thomas Munro <thomas.munro@gmail.com>
Date: 2022-05-03T19:13:05Z
Lists: pgsql-hackers

Attachments

Hi,

On 2022-05-03 14:23:23 -0400, Tom Lane wrote:
> Andres Freund <andres@anarazel.de> writes:
> >> So it's almost surely a timing issue, and your theory here seems plausible.
> 
> > Unfortunately I don't think my theory holds, because I actually had added a
> > defense against this into the test that I forgot about momentarily...
> 
> Oh, hm.  I can try harder to repro it.

I've now reproduced it a couple times here running under rr, so it's probably
not worth putting too much effort into that.

Attached is a fix for the test that I think should avoid the problem. Couldn't
repro it with it applied, under both rr and valgrind.


My current problem is that I'm running into some IPC::Run issues (bug?). I get
"ack Broken pipe:" iff I add "SELECT pg_sleep(1);" after
"-- wait for lock held by prepared transaction"

It doesn't happen without that debugging thing, but it makes me worried that
it's something that'll come up in random situations.

It looks to me like it's a bug in IPC::Run - with a few changes I get the
failure to happen inside pump_nb(), which seems like it shouldn't error out
just because the child process exited...


I *think* it might not happen without the sleep. But I'm not at all confident.

In general I'm kinda worried on how much effectively unmaintained perl stuff
we're depending :(

Greetings,

Andres Freund

Commits

  1. Disable 031_recovery_conflict.pl until after minor releases.

  2. Temporarily skip recovery deadlock test in back branches.

  3. Backpatch addition of pump_until() more completely.

  4. Fix timing issue in deadlock recovery conflict test.

  5. Backpatch 031_recovery_conflict.pl.

  6. Fix possibility of self-deadlock in ResolveRecoveryConflictWithBufferPin().

  7. Backpatch addition of wait_for_log(), pump_until().

  8. Add tests for recovery deadlock conflicts.

  9. Phase 3 of pgindent updates.