NPE in InvalidationsTxGrouper.registerInvalidationSynchroniz

[chrisxsb]

When I try to access a CMP bean from a session bean without a trantaction, I get the following error (this is on jboss-3.2.1, but the problem seems to still exist on later versions):

19:05:13,249 ERROR [LogInterceptor] RuntimeException:
java.lang.NullPointerException
at org.jboss.cache.invalidation.InvalidationsTxGrouper.registerInvalidationSynchronization(InvalidationsTxGrouper.java:76)
at org.jboss.cache.invalidation.triggers.EntityBeanCacheBatchInvalidatorInterceptor.invoke(EntityBeanCacheBatchInvalidatorInterceptor.java:106)
at org.jboss.ejb.plugins.EntitySynchronizationInterceptor.invoke(EntitySynchronizationInterceptor.java:328)
at org.jboss.resource.connectionmanager.CachedConnectionInterceptor.invoke(CachedConnectionInterceptor.java:186)
at org.jboss.ejb.plugins.EntityReentranceInterceptor.invoke(EntityReentranceInterceptor.java:82)
at org.jboss.ejb.plugins.EntityInstanceInterceptor.invoke(EntityInstanceInterceptor.java:174)
at org.jboss.ejb.plugins.EntityLockInterceptor.invoke(EntityLockInterceptor.java:89)
at org.jboss.ejb.plugins.EntityCreationInterceptor.invoke(EntityCreationInterceptor.java:53)
at org.jboss.ejb.plugins.AbstractTxInterceptor.invokeNext(AbstractTxInterceptor.java:84)
at org.jboss.ejb.plugins.TxInterceptorCMT.runWithTransactions(TxInterceptorCMT.java:216)
at org.jboss.ejb.plugins.TxInterceptorCMT.invoke(TxInterceptorCMT.java:104)
at org.jboss.ejb.plugins.SecurityInterceptor.invoke(SecurityInterceptor.java:117)
at org.jboss.ejb.plugins.LogInterceptor.invoke(LogInterceptor.java:191)
at org.jboss.ejb.plugins.ProxyFactoryFinderInterceptor.invoke(ProxyFactoryFinderInterceptor.java:122)
at org.jboss.ejb.EntityContainer.internalInvoke(EntityContainer.java:483)
at org.jboss.ejb.Container.invoke(Container.java:674)
at org.jboss.ejb.plugins.local.BaseLocalProxyFactory.invoke(BaseLocalProxyFactory.java:353)
at org.jboss.ejb.plugins.local.EntityProxy.invoke(EntityProxy.java:38)
at $Proxy147.getXpdlPackage(Unknown Source)
at org.obe.server.j2ee.repository.EJBProcessRepository.findWorkflowProcess(EJBProcessRepository.java:247)
at org.obe.engine.WorkflowEngine.findWorkflow(WorkflowEngine.java:2892)
at org.obe.engine.WorkflowEngine.startProcess(WorkflowEngine.java:625)
at org.obe.engine.WorkflowEngine.startProcess(WorkflowEngine.java:612)
at org.obe.server.j2ee.J2EEServerBean.startProcess(J2EEServerBean.java:187)
at sun.reflect.NativeMethodAccessorImpl.invoke0(Native Method)

[snip]...

Looking at the source, I see that InvalidationsTxGrouper seems to assume that tx cannot be null, even though
EntityBeanCacheBatchInvalidatorInterceptor may pass it a null value (see line 106).

Am I doing something wrong, or is this a bug?

Thanks

--Chris

[acoliver]

JBossCMP requires a transaction.

[chrisxsb]

In that case, perhaps more graceful failure is in order?
Rather than throwing a NPE, maybe logging an error "JBossCMP requires a transaction", and throwing an IllegalStateException (or something similar) would be nicer?

[acoliver]

Sure. Please suppy a patch.

[acoliver]

Oh, I suggest you submit your patch to the Deployment verifier. JBoss should just refuse to deploy it if you try and run CMP with no transaction. (Of course, I recall that happening at one time...)

[chrisxsb]

Here's a patch that does what I suggested once the situation that would cause the NPE is detected
...I just now read your message about the Deployment verifier...
That sounds a little tougher...
I'll see if I can come up with something.

===================================================================
RCS file: /cvsroot/jboss/jboss/src/main/org/jboss/cache/invalidation/InvalidationsTxGrouper.java,v
retrieving revision 1.1.2.1
diff -u -r1.1.2.1 InvalidationsTxGrouper.java
--- InvalidationsTxGrouper.java 8 Nov 2002 16:35:34 -0000 1.1.2.1
+++ InvalidationsTxGrouper.java 15 Feb 2004 20:01:18 -0000
@@ -66,6 +66,11 @@

public static void registerInvalidationSynchronization(Transaction tx, InvalidationGroup group, Serializable key) throws Exception
{
+ if (tx == null)
+ {
+ log.warn("JBossCMP requires a transaction.");
+ throw new IllegalStateException("JBossCMP requires a transaction.");
+ }
InvalidatorSynchronization synch = null;
synchronized(synchronizations)
{

[acoliver]

If no one replies with a reason why this would hurt anything, I'll apply it. That being said, if I do, someone will say "Why do I only get this at runtime, wouldn't it make more sense to throw an error at runtime?"

It would seem sensible to me that if a CMP bean doesn't have Requires, RequiresNew or Manditory that the DV could kick it...

[chrisxsb]

Agreed. However, having a sensible error message indicating the real error is better than nothing :-).

If I get time, I'll look into the DV fix (this is my first foray into the JBoss source, so it might take some time for me to get acquainted).

Thanks for the replies.

[chrisxsb]

For posterity:

The original reason I tried to use CMP beans outside of a transaction was that I was having some *serious* performance issues. This article has all the answers I needed:

http://www.onjava.com/lpt/a/3871

[acoliver]

You may also find this helpful: http://jboss.org/jbossBlog/blog/bburke/?permalink=Optimizing+JBoss%3A+Experiences+with+SPECj2002.html

[chrisxsb]

I just noticed that there probably should be one small adjustment to the patch I posted. Instead of using log.warn, I should have used log.error:

===================================================================
RCS file: /cvsroot/jboss/jboss/src/main/org/jboss/cache/invalidation/InvalidationsTxGrouper.java,v
retrieving revision 1.1.2.1
diff -u -r1.1.2.1 InvalidationsTxGrouper.java
--- InvalidationsTxGrouper.java 8 Nov 2002 16:35:34 -0000 1.1.2.1
+++ InvalidationsTxGrouper.java 15 Feb 2004 20:01:18 -0000
@@ -66,6 +66,11 @@

public static void registerInvalidationSynchronization(Transaction tx, InvalidationGroup group, Serializable key) throws Exception
{
+ if (tx == null)
+ {
+ log.error("JBossCMP requires a transaction.");
+ throw new IllegalStateException("JBossCMP requires a transaction.");
+ }
InvalidatorSynchronization synch = null;
synchronized(synchronizations)
{

--Chris

[marc.fleury]

yes, this is a spec violation I think, the nice message is a good idea.