2007/3/5, Martin Marques <martin@bugs.unl.edu.ar>:
> mikie wrote:
> > 2007/3/2, Martin Marques <martin@bugs.unl.edu.ar>:
> >> mikie wrote:
> >> > 2007/3/1, Robert Treat <xzilla@users.sourceforge.net>:
> >> >> Yeah, do the above recommendation to see if all of your statements are
> >> >> making
> >> >> thier way into pg
> >> >
> >> > OK, I have checked the logs. I can see both COMMIT statements in the
> >> > log - the first commit from my transaction and then the other that I
> >> > send to workaround the problem.
> >>
> >> There must be some other problem. Do you have trasaction ID on the logs?
> >> See if both commits go in the same transaction.
> >
> > Could you please advise me how to check the transaction ID in the log?
>
> Edit postgresql.conf:
>
> log_line_prefix = '<%t %x>'
>
> And restart postmaster.
Thanks, it works. I can see the transaction IDs -- I described my
solution in my previous post.
> >> Any errors during the transaction?
> >
> > As I wrote in my first post, I make this error intentionally to check
> > if transaction works (if something fails, then the entire transaction
> > should be rolled back). I intentionally give wrong data in the "date"
> > field for example.
> > Perhaps I should ask again: is it my responsibility to check if the
> > transaction failed and issue a ROLLBACK command, or will the PG server
> > do it automatically?
>
> One question: Why dont you but each query in a diferent pg_exec() command?
It does not make any difference if I put single query in each
pg_query() command.
--