[Topic: transaction aborted after error, no commit possible]
"Daniel Serodio" <daniel@checkforte.com.br> wrote:
> While I also dislike this behaviour, it's the backend's fault, not the
> driver's. Try the same thing in psql (or any other frontend) and you'll
> get the same results.
>
> If you do try to repost this question on another list (pgsql-hackers
> etc) I'd be interested in following the discussion, if you could CC: me
> I'd be grateful.
There is no sense in re-posting to hackers. This has been discussed before
and is well known behaviour. Postgres' transactions are atomic, either the
whole transaction will commit or nothing at all. Look at the archives to see
discussion about exactly the same topic.
AFAIK, Bruce is working on nested transactions for 7.4 or later. That will
open the possibility for the behaviour Marcos expected.
Best regards,
Michael Paesold