[jboss-jira] [JBoss JIRA] Assigned: (JBCACHE-914) Adapt JBCACHE-913 fix for JBC 2.0

Adapt JBCACHE-913 fix for JBC 2.0 --------------------------------- Key: JBCACHE-914 URL: http://jira.jboss.com/jira/browse/JBCACHE-914 Project: JBoss Cache Issue Type: Sub-task Security Level: Public(Everyone can see) Reporter: Brian Stansberry Assigned To: Brian Stansberry Fix For: 2.0.0.GA Need to port JBCACHE-913 fix to HEAD, where the code works differently. Vladimir, I'm assigning this to you, because you are doing heavy lifting on state transfer and I don't want to step on you. I experimented with adding the following to StateTransferManager and calling it in setState right after the call to integrator.integrateState. That made the unit test (o.j.c.eviction.StateTransferIntegrationTest) pass. One concern I have is calling getChildren() on a node will end up pulling non-in-memory persisted nodes into RAM. A look at NodeImpl makes me think that's not the case, but that could change. /** * Generates EvictedEventNode notifications for all nodes of the tree. This is * called whenever the tree is initially retrieved (state transfer) */ protected void notifyEvictionPolicyOfNodesCreated(Node curr) { Node n; Set children; if (curr == null) return; Fqn fqn = curr.getFqn(); // JBCACHE-913 Region region = cache.getCacheSPI().getRegion(fqn, false); if (region != null) { region.putNodeEvent(new EvictedEventNode(fqn, NodeEventType.ADD_NODE_EVENT)); } if ((children = curr.getChildren()) != null) { for (Iterator it = children.iterator(); it.hasNext();) { n = (Node) it.next(); notifyEvictionPolicyOfNodesCreated(n); } } } Another approach would be to not bypass the interceptor chain when the integrator puts the node in the cache. That's very clean. But then we have to make sure no unwanted things are done by any interceptors.