ISPN-7097 Fix random failures in FlagsReplicationTest

[rvansa]

https://issues.jboss.org/browse/ISPN-7097

Make sure that the CommitThread runs the transaction started in LockingThread.

[danberindei]

@rvansa I'd like a JIRA, for one I'd like to see an explanation why this doesn't fail every single time :)

[danberindei]

Looking at the code, I'm not sure why the test is using separate threads at all. haveSecondaryThreadTakeLock could just as well start a tx, acquire the lock, suspend the tx and return it, with haveSecondaryThreadReleaseLock resuming the tx and committing it.

[rvansa]

@danberindei I think that it does not fail all the time because threads are recycled and there is a good chance that the FJP will assign the operation the same thread. If it does not, there won't be any transaction to commit.
TBH I forgot if I've managed to trace the actual behaviour, I have it in my commit pool for quite some time :) But changing that to single-threaded test makes sense, I'll do that.

[rvansa]

@danberindei here you come....

[danberindei]

@rvansa That's why it's good to create a JIRA, it means you have it written down and you don't have to remember everything. E.g. AbstractInfinispanTest.fork() actually uses a ThreadPoolExecutor, not FJP :)

[danberindei]

Integrated, thanks Radim!