[hibernate-dev] Regression in AS7 (master) cluster cache invalidation test after moving from Hibernate ORM 4.1.6 to 4.1.9

[Scott Marlow]

http://hudson.jboss.org/hudson/job/as7-param-pull/5123/ shows a failure 
with the second level cache not being invalidated as expected across a 
two node cluster.

I'm able to reproduce locally, which is good.  We will continue to use 
Hibernate 4.1.6 until we get past this regression (hopefully we can 
understand the test failure cause soon).

https://github.com/jbossas/jboss-as/blob/master/testsuite/integration/clust/src/test/java/org/jboss/as/test/clustering/cluster/ejb3/xpc/StatefulWithXPCFailoverTestCase.java#L276 
is the line that is failing (yeah, the arguments to assertEquals are 
backwards but that isn't the root issue ;)

It is possible that its a bug in the test also.

The cluster has two nodes, each with their own in-memory database.  The 
test assumes that it can create the same entity on both nodes (same 
primary key) and that for the purpose of testing the second level cache, 
we can expect cache invalidation to occur on both nodes when deleting 
the entity on one of the nodes.

I believe that the cache invalidation should be synchronous but I'll 
probably try adding a sleep to see if it could be an async/timing issue.

I hope to have more information soon (probably in the morning on Friday).

Scott