I'm just investigating a Lingo test case (LINGO-22) submitted by Jim Beattie which tests that we leave no threads around after shutting down lingo & ActiveMQ - which sounds good to me.
I've noticed that the Executor for all connections is a static and shared across connections. I was gonna make it so that after all connections are closed all threads are cleanly shutdown - I've a refactor of Jim's test case to test for this. I just wondered before I go ahead and change things - was there a good reason to keep the TaskRunnerFactory / Executor a singleton? I was thinking of creating one per connection. Failing that we could keep it a singleton and just reference count the connections so we can close it when the last connection goes away Am wondering if this was an optimisation to minimise context switching when using lots of sessions - or is it just an oversight? -- James ------- http://radio.weblogs.com/0112098/
