[emuckenhuber]

So an update on this topic. Basically i removed the duplicate annotation processing
and added a jsf injection handler which delegates the injection to the TomcatInjectionContainer,
which seems to work pretty ok now. For jsp's we check if the classloader of the object is the JasperLoader (thanks to Remy for the hint)
to determine whether it's an 'dynamic bean' or not. I've added simple annotated jsp test, based on the ENCServlet.

Although i think that in this case we have the same issue we had with servlets once. That the 'Resolution should not happen via InjectionContainer'.
Which is basically that the references are not processed by the MappedReferenceMetaDataDeployer.
But i guess we can postpone this specific injection handling a bit ? :)

[starksm64]

"emuckenhuber" wrote:

Although i think that in this case we have the same issue we had with servlets once. That the 'Resolution should not happen via InjectionContainer'.
Which is basically that the references are not processed by the MappedReferenceMetaDataDeployer.
But i guess we can postpone this specific injection handling a bit ? :)

Yes. If we do need late runtime resolution it should be based on the MappedReferenceMetaDataResolverDeployer DeploymentEndpointResolver output.

[emuckenhuber]

Yes. If we do need late runtime resolution it should be based on the MappedReferenceMetaDataResolverDeployer DeploymentEndpointResolver output.

Yes, this should be done now - based on the WebEjbHandler you committed a while ago.

Furthermore i noticed that there is a name mismatch in what the InjectionUtil and jboss-metadata is doing:

String encName = getEncName(ref, field);

// is doing

getEncName(field.getDeclaringClass()) + "/" + field.getName();

So the InjectionUtil use the class + / + fieldName whereas jboss-metadata is just using the fieldName.
I guess we should change that either in metadata or in the InjectionUtil?