Tom Lane wrote:
> "Heikki Linnakangas" <heikki@enterprisedb.com> writes:
>> I just noticed that freezing a tuple (in vacuumlazy.c) is not
>> WAL-logged.
>
> The theory is that this doesn't matter because the tuple is committed
> either way ... it's equivalent to a hint-bit update which we don't
> WAL-log either.
Because the relminxid-update is WAL-logged, you can get into situation
where relminxid > the real smallest xid of the table, as demonstrated by
the example I gave. relminxid is used to determine the safe clog cut-off
point, so that's not harmless.
Granted, the chances of getting data corruption from this are small, but
it's possible.
--
Heikki Linnakangas
EnterpriseDB http://www.enterprisedb.com