This content has been marked as final.
Show 4 replies
-
1. Re: Great but...
belaban Sep 12, 2005 7:09 AM (in response to ccsaxton)#1 I leave this to Ben...
#2 Can you check out the CallbackTest unit tests ? Please provide us with unit tests that show what IYO are incorrect semantics. Note that we will move all callback-related logic into its own interceptor in 1.3.
#3 I assume you mean for replication. A new thread is used (but not created). Essentially the listener thread which receives the replication message. *However*, we use the same global transaction, so the 2PC works. -
2. Re: Great but...
ben.wang Sep 12, 2005 5:18 PM (in response to ccsaxton)You can't create *Region* (essentially a sub-tree) dynamically now.
-Ben -
3. Re: Great but...
ccsaxton Sep 12, 2005 5:58 PM (in response to ccsaxton)"bela@jboss.com" wrote:
#3 I assume you mean for replication. A new thread is used (but not created). Essentially the listener thread which receives the replication message. *However*, we use the same global transaction, so the 2PC works.
I was assuming that the cache uses a thread for the eviction policy even if it is not a replicated cache...(is that not so...I haven't looked at the code yet). This assumption was made on the premise that I have to call startService and stopService methods.
#2 is easy to replicate...create a listener and add it to a cache...add a node to the cache and then in the listener logic that catches the node add event just try and lookup the newly added node from the cache using the passed FQN...returns null. -
4. Re: Great but...
belaban Sep 13, 2005 4:55 AM (in response to ccsaxton)Look at CallbackTest:
testLocalGetCallbackSameFqnWithoutTransaction()
does exactly what you describe, and passes. Can you check whether this works for you ?