All right. What createBindingInfo() does is, among other things, to define the transport factory linked to a namespace. When a namespace is read for the very first time, CXF read a configuration file named bus-extensions.xml. At that point, Jetty registers as a factory for some namespaces. So what I did was simply to override the settings read by CXF right after the file reading. createEndPointInfo() calls createBindingInfo() and then actually retrieves the transport factory linked to the input namespace. As Jetty was overriden by our transport factory, createEndPointInfo() is then able to call our own transport factory.
I'll provide a patch with cleaner code as soon as all my tests are done. David Blevins wrote: > > > On May 19, 2009, at 2:44 AM, Jean-Sébastien Scrève wrote: > >> I began to migrate to CXF 2.2 revision. I was able to get a first >> revision >> working : https://issues.apache.org/jira/browse/OPENEJB-977 >> >> There's still a piece of code I don't like that I try to remove. See >> posts >> on CXF : >> http://www.nabble.com/Migrating-to-CXF-2.2-in-OpenEJB-td23518058.html#a23612789 > > In that thread you mention that overriding createBindingInfo() solves > the problem. Not very CXF aware, can you tell me more about that? > >> My question : it seems OpenEJB team decided not to use Spring to >> configure >> CXF (Spring libraries not included from CXF and the >> ExtensionManagerBus >> usage). Does anybody know the reason for that ? I guess we wanted to >> avoid >> Spring dependencies. > > We've so far tried to get by without Spring as a requirement for > OpenEJB to run. The first library we encountered that used Spring was > ActiveMQ, but even they don't require Spring to run and it was just an > option. So for our ActiveMQ integration we support both with the > default being to not use Spring. > > When we did the CXF integration, we took the same route as they also > didn't require Spring to run and it was just an option; tough I'm not > sure in our integration if we have a way to turn "spring > configuration" support on for CXF like we can for ActiveMQ. > > There's the history. > > If we can continue to get by without it as a requirement, that'd be > the best option in my opinion at least. > > -David > > > -- View this message in context: http://www.nabble.com/CXF-migration-to-version-2.2-tp23612958p23704340.html Sent from the OpenEJB Dev mailing list archive at Nabble.com.
