Hi Andreas
The issue I see is with my dual core, where a thread updating the latch is
not seen from another thread. The concurrency of 4 is reached if you put a
sout, but the latch update is not visible to the main thread. This is due to
the Java memory model AFAIK
This would mean that either CountDownLatch is not used correctly in
MinConcurrencyTest or that there is a bug in CountDownLatch, which is
very improbable. Can you exclude the possibility that
concurrencyReachedLatch.await simply times out too early?
This infact is strange.. I looked into this again, and it seems like the
original code has some unknown impact on the rest of the behavior.. the
test consistently passes with the new code, whereas it intermittently
fails with the original code. If you use the old code and just take out
the use of latches, and use a sout, you always see 4 messages being
received immidiately
asankha
--
Asankha C. Perera
http://adroitlogic.org
http://esbmagic.blogspot.com
