> so far I couldn't build a simple sample application that shows this
> behaviour. But I found out that commenting line #285 in
> AbstractPingablePooledConnection.java seems to "solve" the problem.
>
> Maybe "someone" in the JBoss environment is closing an already closed
> connection. Maybe it's possible to make the code a bit more robust for
> scenarios like this? What do you think? Maybe via some configuration value!?

Well, if we would definitely know that JBoss closes the physical connection,
then yes - we could create something similar to what we did for the Sun
AppServer - own pool classes with overridden behavior. No problem with that.

Only that in the moment I am not sure about it. But I will check if we can live
without that checkValidity() call.

Could you please put this in tracker, so I do not forget?

Roman