Kepler

BTW - the purpose of the shutdown hook is:

Added a shutdown hook so that model is gracefully stopped when the JVM shuts down due to control-C, the user logging out, etc. This allows 30 seconds for the shutdown, then shuts down anyway.

About leaks in Manager concerning the shutdown hook, see https://chess.eecs.berkeley.edu/ptexternal/wiki/Main/MemoryLeaks#ManagerShutdownHooks

Edward suggested:

The solution is that Manager should have a static code body and register exactly one shutdown hook.
Then Manager._registerShutdownHook should create a static list of weak references to Managers that have been created.
The one shutdown hook should check if the weak reference is valid and whether the thread is running, and if it is, call finish() and attempt a Thread.join().

I'd accept a patch that implements a fix, but I probably won't get around to fixing this myself any time soon.

Ignoring the memory leak issue, the shutdown hook does not appear to be correctly implemented since it is unregistered after the first execution (sometimes) and control-c during subsequent executions won't call it. A simple solution is to register the hook at the beginning of each model execution, and unregister when execution finishes.... this also fixes the memory leak.

Good point. I'll keep this bug in my queue.

I can take a look at this next week, after the dust settles from the other clone() bug.

I've made adjustments for how we create and delete the shutdown hook. Let's wait a bit to see what happens before closing this one.