Well, I have not been able to track it down fully, but I modified the
locktest to be a bit more agressive. First I simply up'd the thread and
iteration count (which could cause a NullPointerException in the entity
syncrhonization interceptor... but not all the time). It will also hang
some times, and then sometimes it will just run fine.
I was not sure if the commit-option had anything todo with it, so I expanded
the test to run over each commit option. With the thread count low (10 or
so), option a and b work, but when it comes to c it never makes it through,
the test just hangs. I have no idea what happens with option d.
All the while it produces a lot (alot of transactional contention on
contextXXX) messages, which makes sence.
I also changed the test to actually assert if the test failed.
Unfortunatly, the test will timeout and not be included in the report... not
really sure why that is, but it is.
*sigh*
I was hoping to reproduce what I am seeing with this test, but it has not
turned out that way. I can not even tune the test to fail each time... I
really need a drink. It does fail though... with a hang or null pointer.
Anyways, I commented out the c & d tests for now since they cause the report
not to show up and I turned the thread count down to 10 and iter to 20.
Someone who knows a little more about the locking mechanism (marc?) might
want to play with this test a little to see where it hangs and where the
null is thrown.
I ended up using the RMI IL for now, but I think that I still get a
contention/hang after a while =(
--jason
_______________________________________________
Jboss-development mailing list
[EMAIL PROTECTED]
http://lists.sourceforge.net/lists/listinfo/jboss-development
