On Tuesday 19 April 2011, Daniel Kulp wrote: > On Tuesday 19 April 2011 8:38:31 AM Vassilis Virvilis wrote: > > Hi > > > > I am using the cxf 2.3.6 (we haven't upgraded yet to 2.4.0) and we have the > > following exception on deploy (not in runtime) > > There isn't a 2.3.6. I assume 2.3.3?
Yes I meant 2.3.3. Sorry. > In either case, if it worked with 2.2.x, I'm actually quite surprised and > would have been completely by accident. For both JAXB and Aegis, the > interpretation of the generics is really just done for two cases (well, one > case for JAXB): > > 1) List<..> - both JAXB and Aegis will handle this to determine the "type" > of > the list. For the most part, it's treated the same as an [] of the type. > > 2) Map<.. , ..> - Aegis will also look at this and create a pseudo map type > for it with key/value entries. So lists and maps are special. > > That's pretty much it at this point. It MAY have worked with 2.2.x as I > think 2.2.x really didn't look at any of the generics much at all. For the > most part, it would have seen the "getFirst"/"setFirst" methods on the Pair > and just treated them as the "Object" version and mapped them to xsd:any, > which is kind of a crappy schema, but it may have worked for you. Again, > purely by accident and not by design. In 2.2.6 we had some heavy generic servlets that were working for java2java but they were failing for java to js. I have filed a JIRA https://issues.apache.org/jira/browse/CXF-1758 for this but since we were just experimenting with js it was a toy problem. I think this confirms the works by accident theory. > That all said, an NPE is definitely not what should occur. A better > exception should be generated. If you could create a small test case and > attach to a JIRA, that would be great. > > Dan > Ok I will file a jira in wednesday after easter holdays :-) Thanks for the detailed and helpful answer. It is much appreciated. Vassilis
