Horton

Re: Bugs in TOAST handling, OID assignment and redo recovery

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

View thread View on postgr.es
From: Tom Lane <tgl@sss.pgh.pa.us>
To: Michael Paquier <michael@paquier.xyz>
Cc: Pavan Deolasee <pavan.deolasee@gmail.com>, Heikki Linnakangas <hlinnaka@iki.fi>, pgsql-hackers <pgsql-hackers@postgresql.org>
Date: 2018-04-11T23:51:35Z
Lists: pgsql-hackers 
Michael Paquier <michael@paquier.xyz> writes:
> I have not really checked this thread in details, but one thing that
> strikes me is that it would be rather easy to add a TAP test based on
> the initial script that Pavan has sent.  Would that be worth testing
> cycles or not?

I doubt it --- that test is so specialized that it'd be unlikely to
catch any other bug.

An example of the sort of bug that could exist in the same area would
be if we'd failed to note the places in tuptoaster.c that also need to
be switched to SnapshotAny.  Only a test covering heap-rewrite could
have found that, which this test case surely doesn't.

What I think would actually be worth spending some time on is taking
a hard look at the remaining uses of SnapshotDirty to see if any of
them look squishy.

			regards, tom lane

Commits

  1. Ignore nextOid when replaying an ONLINE checkpoint.

  2. Do not select new object OIDs that match recently-dead entries.