Transaction leak?

[epbernard]

If someone outside Bill has a clue on that...

Not that I know of. Maybe Clebert or somebody was trying to fix a
memory leak somewhere. Doesn't make sense to me.

Emmanuel Bernard wrote:
> Apparently, Hibernate tries to register a work on a transaction that is
> no longer active. But to me the latest code cannot fall into this case.
> Plus the user solve his issue by increasing the heap size.
>
> Could it be that JBoss keep track of the transaction or something
> related to the Tx in some kind of weak hashmap making it candidate for
> GC when memory is low. This could explain the behavior the user have.
>

[bill.burke]

have you been able to verify this with a test case? I don't use any WeakHashMaps anywhere really.

Are they sure that a transaction timeout was not reached and the tx was marked for rollback?

[epbernard]

No, no test case.
This may be a possibility