On Mon, 2008-03-03 at 07:33 -0500, Rafael Schloming wrote: > I suspect the potential for the deadlock to occur is actually still > there on the 0-8 codepath, but it is not occuring for some reason. I > don't know if this is timing related or some other reason. The stack > trace suggests that the dispatcher thread is running as we attempt to > cancel the consumer. Stopping the thread should prevent the deadlock, > I > don't know if the 0-8 code does that or not. >
It is not occurring when running MessageListenerMultiConsumerTest as all the tests pass. With 0.10 there are some tests that we know should fail until we have a way of turning pre-fetching off. I believe this is the reason why we will not see this happening when using the 0.8 code path. Wit the 0.10 code path messages are still in the process of being dispatched when the test fails and call tearDown that closes the connection.
