Yes, you're absolutely right - let's concentrate us on one issue at time. Just forget #3 (I've created a separated topic at WS forum - so far no answer) and #4 (even if we require this to work in our application, it is not the actual problem right now and there is already a JIRA for that). So the question is if we should go for #1 or for #2? 1) Deployment ordering issue - are you sure that the initial problem (NullPointerException in JavaEEComponentHelper) is caused by the deployment order? This is actually what I'd like to solve. 2) Usage of mappedName should be just a work-around to #1 above. Actually we don't want to use mappedName at all as it is also application server specific. I have a smaller application, but still containing around 100 different jar files and 5-10 EJBs. And also this version of JBoss I've mentioned previously that, according to Red Hat support people, should give more debugging information. Do you want me to test this smaller application on this particular JBoss version as well? Shall I continue with mappedName, i.e. trying to resolve #2 first or regenerate code without mappedName and go for #1 once again? View the original post : http://www.jboss.org/index.html?module=bb&op=viewtopic&p=4226506#... Reply to the post : http://www.jboss.org/index.html?module=bb&op=posting&mode=reply&a...