Postgresql XA prepare() method behaviour

Hello all.
A couple of weeks ago I stumbled upon a strange behaviour of postgresql
jdbc driver that I think was initially identified almost 5 years ago.
The problem i'm referring to is this:

http://archives.postgresql.org/message-id/Pine.OSF.4.61.0602121957130.244416@kosh.hut.fi

It seems to me that the patch never made it into trunk;
the current source from CVS still returns XA_OK from prepare() when the
"PREPARE TRANSACTION" command returns "ROLLBACK" (as is the case when a
statement like "SELECT * FROM table_that_does_not_exists" was executed
in transaction).

I contacted the original author, Heikki Linnakangas, and he told me that
the patch was a little bit messy and that it probably simply got
forgotten, and urged me to raise the question here again, since one way
or another it should be fixed.

Regards

--
Francesco Degrassi

Emaze Networks S.p.A
www.emaze.net

Can anyone provide feedback on this ?
I'm adding a shortened URL to the original thread to make it easily accessible
from the web archives of the ML.
 
http://bit.ly/elrpFE
 
Regards
 
--
Francesco Degrassi
Emaze Networks S.p.A.
www.emaze.net