On 2014-07-24 18:32:09 -0400, Tom Lane wrote:
> Andres Freund <andres@2ndquadrant.com> writes:
> > On 2014-07-24 17:46:13 -0400, Tom Lane wrote:
> >> It looks to me like this never worked: the code around recoveryStopsHere
> >> has never paid any attention to COMMIT_PREPARED (or ABORT_PREPARED).
>
> > I wonder if this is possibly influenced by
> > bb38fb0d43c8d7ff54072bfd8bd63154e536b384. IIRC before it committing a
> > prepared xact caused a plain commit record for the xid doing the COMMIT
> > PREPARED.
>
> Hm, I see nothing in that commit that looks like it would've changed
> what gets written to WAL (and if it did, we have a serious minor version
> compatibility issue). OTOH, if it did, that would explain the
> complaint of a behavioral change since the last minor releases.
Afaics before the commit a LockGXact() did a GetTopTransactionId(),
afterwards it doesn't anymore. That'd fit, right?
Greetings,
Andres Freund
--
Andres Freund http://www.2ndQuadrant.com/
PostgreSQL Development, 24x7 Support, Training & Services