Re: "Idle in Transaction" revisited.

On Fri, 17 Sep 2004, John R Pierce wrote:

> > Yeah, that's from prematurely issuing BEGIN.  I believe there's a
> > solution for this in the most recent JDBC driver versions --- but
> > I dunno if it's in any official releases yet.
> 
> Its my understanding from reading the prior threads on this that somewhere 
> after jdbc 213 they moved the initial BEGIN to the first SQL Execute rather 
> than immediately upon Connection.
> 

This change only occured in the 8.0 development branch, and not on the 7.4 
branch.  I produced a patch and then current versions of the 7.4 jar files 
for people to try this out on their 7.4 servers, but no one ever got back 
to me about it, so I didn't apply it because it was a pretty ugly fix that 
no one seemed to need.  See

http://www.ejurka.com/pgsql/jars/transaction_state/

> When an app does a Commit or Rollback, this generates an END;
> After this END; does a new BEGIN happen immediately, or is it postponed 
> until the next Execute ?

Yes, in the 7.4 branch a BEGIN is issued immediately.

Kris Jurka