I can't quite see how this could make a difference but you might try
setting the ThreadContextClassloader on the threads you create to the
same as the gbeans TCCL.
thanks
david jencks
On Apr 10, 2007, at 1:39 PM, Oli Kessler wrote:
It might be a threading issue.
Our GBean spawns some more threads, each instrumented with a new
connection to the JMS. In this scenario, all threads fail with the
NCDFE.
However, if the GBean accesses the JMS on it's own (in the same
thread), everything works.
The broker implementation starts up, the logging is indicating
success (Journal open, channel at vm://localhost open, client gets
added)
The test code is the following:
logger.info("uuuhh - trying direct JMS access");
TopicConnection con = ((TopicConnectionFactory)
this.managedConnectionFactoryWrapper.$getResource
()).createTopicConnection();
con.setClientID("customGBean-main");
con.start();
javax.jms.TopicSession s = con.createTopicSession(false,
Session.AUTO_ACKNOWLEDGE);
javax.jms.Topic t = s.createTopic(this.jmsTopicName);
TopicPublisher pub = s.createPublisher(t);
logger.info("Publisher is " + pub);
If however I run this code in a separate thread spawned by the
GBean, the startup fails with the NCDFE.
Do I have to use thread pools or any other special means in this case?
Thanks in advance
-ok
PS: now on geronimo-1.2beta
On 10.04.2007, at 02:25, David Jencks wrote:
Generally if you use dependencies on jars that are also present in
server classloaders you will get into trouble, although I'm a
little surprised at NCDFE.
Try replacing all the geronimo dependencies with one on the
connector configuration
<dependencies>
<dependency>
<groupId>geronimo</groupId>
<artifactId>j2ee-server</artifactId>
<type>car</type>
</dependency>
(I think that's correct for 1.1.1, I'm a lot more familiar with
1.2 and trunk at the moment).
thanks
david jencks
