I wrote:
> What I suspect is we've broken the logic that manages REINDEX applied to
> pg_class itself.
Hm, this is a HOT bug. (Unless the OP found it on something older
than 8.3, in which case we might have more than one issue involved.)
What I'm seeing is that when heap_update checks to see if the new
pg_class tuple is HOT-updatable, RelationGetIndexAttrBitmap tells it
that only the OID is an indexed attribute, and so it does a HOT update,
leaving the wrong index entry in place.
Apparently, rd_indexattr needs to get cleared someplace where it
isn't being cleared right now. I'm sure this is related to the
fancy dance we do for pg_class reindexing, but not sure yet where
to fix it.
regards, tom lane
--
Sent via pgsql-bugs mailing list (pgsql-bugs@postgresql.org)
To make changes to your subscription:
http://www.postgresql.org/mailpref/pgsql-bugs
