[jboss-jira] [JBoss JIRA] Commented: (JGRP-652) ReplicatedHashMapTest fails

ReplicatedHashMapTest fails --------------------------- Key: JGRP-652 URL: http://jira.jboss.com/jira/browse/JGRP-652 Project: JGroups Issue Type: Bug Reporter: Bela Ban Assigned To: Vladimir Blagojevic Fix For: 2.7 Attachments: ReplicatedMapDemo.java [email from Robert Newson] The ViewDeliveryDemo was testing a new feature, it was not the code I sent to Vlad or the code we discussed for the atlanta test. The ViewDeliveryDemo works fine for me once I suppress sending messages after block() as per your change. The test I'm struggling with is the ReplicatedMapDemo which I gave to Vlad several days before writing the ViewDeliveryDemo, sorry for any confusion, but the atlanta lab were testing before I wrote the VDD, so I'm not sure how we got muddled. I've reattached the demo in question. Also, I can confirm that I've written a separate test to demonstrate the losslessness of MessageDispatcher with a zero timeout. It seems to work, though on occasion a VM will block indefinitely on the cast (even though the other boxes are still receiving replies from it), probably a coding error on my part, but fyi. For clarity, the only test the Atlanta lab should run is ReplicatedMapDemo and with these arguments (5 N 50) where N is the total number of nodes. For more stress, increase the first and last parameter (the first is number of maps, the second the number of entries). All boxes should print "PASSED" within some reasonable number of retries (the retry number indicates how long it took for all the puts everywhere to arrive, it polls). It should be started on all boxes concurrently and needs to be run a number of times. I find that it seldom reaches "PASSED" on all boxes though, curiously can reach passed on some but not others (even though map.put() is synchronous)... I have a workaround where I multiplex my own maps over a single, non-multiplexed channel, which works but is far from ideal.