Tonight I've made some commits to these tests to:
* Fix compilation in JDK5
* A Remove concurrency problem which lead to BrokenBarrierException
* Add a test that Removal is not blocked by Passivation for a given Session
So I think our test coverage is now at a point where we can say that SimpleStatefulCache
is better than in previous releases.
However, a lot bothers me will this whole setup:
1) Poor documentation for concurrency. What objects must be guarded by what locks?
2) StatefulBeanContext should be Thread-safe. We should be encapsulating its locking
policies internally.
3) The EJBTHREE-1549 tests are tightly bound to the implementation of SimpleStatefulCache.
Instead, we should be coding the tests to an SPI (from ejb3-cache) and able to plug in
whichever Cache implementations we'd like.
Just wanted to get these observations down so we can revisit them when integrating
ejb3-cache and Brian's work. I do think we have something good enough now that we can
proceed w/ EJB3 Core releases again. And then we'll have Andy run his performance
tests on there to reveal we squashed the "stops the world" problem during
passivation.
S,
ALR
View the original post :
http://www.jboss.com/index.html?module=bb&op=viewtopic&p=4186212#...
Reply to the post :
http://www.jboss.com/index.html?module=bb&op=posting&mode=reply&a...