Kathey Marsden wrote:
Following the possible codepaths through the BrokeredStatement constructor we could get to XAStatementControl.getRealStatement() which might have a null xaconnection. I could try fixing DERBY-2142, if you think it might be related. Seems a little like a shot in the dark since they say they have only a single connection going at the time, but I suppose it would be good to have that fixed anyway.
Even with a single connection there's a chance that multiple threads could be involved. There could be a thread that is performing the connection pool management, it gets the notification the connection has been closed by the application and then it executes statements to reset the physical connection to a known state. That kind of behaviour would hit DERBY-2142. No idea if that is actually happening in this case though.
Dan.
