> I find the text of the jta 1.0.1B spec somewhat unclear on this but I > notice on p. 63 in the change history for 1.0.1B they changed the > description of when to call Synchronization.beforeCompletion from > "before the start of the transaction completion process" to "before > the start of the two-phase transaction commit process" so I think > there's at least an implication that we shouldn't call > beforeCompletion when we know we're going to rollback. > > I imagine the problem is that hibernate tries to use something that > objects because it's set rollbackOnly? > > Which version of geronimo is your version of jencks using? > > Would you like to file a jira issue for this? > > thanks > david jencks >
Hi David - Thanks for your quick reply. I think that it might be worth filing a JIRA for this - I will post this info into it. Interestingly, the test code fails with NPE when using JOTM in the afterCompletion method, as the TM will no longer supply a transaction to getStatus() on, and you need to use the supplied result code. Its interesting that activemq works though; i'll forward this to logicblaze. Best Regards, James
