Considering that I decided to go in a different way. I did some tests
and I'm able to create a separate classloader programmatically inside
websphere and load classes without any colision with app server and app
classes. It is dangerous to do this specially inside a container like
websphere but it think that this is the cleanest solution. I would like to
thank you guys form CXF. It is really rare to find such a good support even
when you have to pay.

    Ats,
    Osvaldo Pina.


On Tue, Aug 6, 2013 at 10:13 AM, Daniel Kulp <dk...@apache.org> wrote:

>
> On Aug 6, 2013, at 8:41 AM, Sergey Beryozkin <sberyoz...@gmail.com> wrote:
>
> > As it happens Dan has just enhanced CXF JAXBUtils to support the Moxy
> JAXB implementation directly, so that may help in your case as well.
>
> Kind of….  Every time I try updating the code in CXF to work with Moxy, I
> run into another bug in Moxy that I need to log with them and wait for a
> fix.   :-(   Hit another one yesterday.   The good news is that they HAVE
> been fixing the issues fairly quickly.
>
> The other thing I keep running into are tests that use direct XML ->
> String and assert information about text in that String.   Moxy uses
> completely different default namespace prefixes so all those tests end up
> failing.   Kind of a slow going side project.
>
>
> > I can see from the Dan's comments that Moxy supports a Map directly as a
> namespace mapper.
> > So I'd like to ask, would it help if I also added a custom
> NamespaceMapper object setter, assuming that in your case, you have nether
> RI or Moxy JAXB implementations used and also have to use some custom
> NamespaceMapper representation, something different than a Map ?
>
> The best option is to use a Map<String, String> in the implementation and
> use the JAXBUtils.setNamespaceMapper(Map, Marshaller) to handle adding it
> to the context.   That will automatically handle the two Sun
> implementations and Eclipse depending on the implementation it finds.
>
> Dan
>
>
>
> > Or may be even Map but not with Moxy ?
> >
> > Let me know please
> > Thanks, Sergey
> >
> > On 01/08/13 22:06, Sergey Beryozkin wrote:
> >> Hi
> >> On 01/08/13 19:03, Osvaldo Pina wrote:
> >>>   Sergey,
> >>>
> >>>   Sorry to bother you again with this problem but I did some research
> in
> >>> the cxf source code and I think that changing the property will not
> solve
> >>> the problem because the AbstractJAXBProvider will always try to create
> >>> a NamespaceMapper (inside protected void setNamespaceMapper) which is a
> >>> child of jaxb-impl com.sun.xml.bind.marshaller.NamespacePrefixMapper
> (and
> >>> in vm runtime the class is
> >>> com.sun.xml.internal.bind.marshaller.NamespacePrefixMapper). So, I
> think
> >>> that the best solution for that is to create a child
> >>> of JAXBElementProvider, overriding setNamespaceMapper in order to
> >>> create a
> >>> custom NamespaceMapper. Am I correct? If so how can I instruct
> >>> the JAXRSClientFactoryBean to programmatically create an instance of my
> >>> custom JAXBElementProvider instead of the default one?
> >>>
> >> well, it is me who is sorry for not addressing this issue properly :-).
> >> Thanks for your investigation. Yes, at the moment, registering a
> >> JAXBElementProvider extension will do, I will also get rid of
> >> "namespaceMapperProperty" and see what else might be improved
> >>
> >> Thanks, Sergey
> >>>    Ats,
> >>>    Osvaldo Pina.
> >>>
> >>>
> >>> On Wed, Jul 31, 2013 at 12:07 PM, Osvaldo Pina
> >>> <osvaldo.p...@gmail.com>wrote:
> >>>
> >>>> Thanks a lot for the information!
> >>>>
> >>>>
> >>>> On Wed, Jul 31, 2013 at 11:55 AM, Sergey Beryozkin
> >>>> <sberyoz...@gmail.com>wrote:
> >>>>
> >>>>> On 31/07/13 15:52, Osvaldo Pina wrote:
> >>>>>
> >>>>>>   Sergey,
> >>>>>>
> >>>>>>    You mean 2.6.7? because I could not find the 2.7.7 version.
> >>>>>>
> >>>>>>  Actually, I got confused, I'm telling everyone CXF 2.7.7 is the
> >>>>>> latest
> >>>>> :-) I meant 2.7.6, sorry about it. CXF 2.6.9 should also have those
> >>>>> properties supported
> >>>>>
> >>>>> Cheers, Sergey
> >>>>>
> >>>>>
> >>>>>      Ats,
> >>>>>>     Osvaldo Pina.
> >>>>>>
> >>>>>>
> >>>>>> On Tue, Jul 30, 2013 at 6:09 AM, Sergey Beryozkin
> >>>>>> <sberyoz...@gmail.com
> >>>>>>> wrote:
> >>>>>>
> >>>>>>  Hi
> >>>>>>>
> >>>>>>> On 29/07/13 21:32, Osvaldo Pina wrote:
> >>>>>>>
> >>>>>>>       Hi all,
> >>>>>>>>
> >>>>>>>>      I have a very strange scenario. I need to use CXF 2.7.1 for
> >>>>>>>> jax-rs (I
> >>>>>>>> can change to another version) in a IBM JVM (build 2.6, JRE 1.6.0)
> >>>>>>>> without
> >>>>>>>> including the jaxb-impl that cxf requires. I have a lot of
> >>>>>>>> serialization
> >>>>>>>> tests that I executed under IBM Jvm and all passed. In my
> scenario I
> >>>>>>>> noticed only a problem with namespaces. Which is not a problem
> >>>>>>>> for me
> >>>>>>>> to
> >>>>>>>> remove them. Does anyone knows other compatibility issues?
> >>>>>>>>
> >>>>>>>>  One thing I'm aware of is that up to CXF 2.6.6 (inclusive) XML or
> >>>>>>> Jettison
> >>>>>>> (JAXB-based) providers will fail if namespace prefixes have to be
> >>>>>>> customized due to these provides using JAXB RI specific properties
> to
> >>>>>>> set
> >>>>>>> namespace mappers. The same issue would apply to JAXBProvider
> >>>>>>> trying to
> >>>>>>> set
> >>>>>>> XML Processing Instructions.
> >>>>>>> In CXF 2.7.7 I added properties to these providers (ex,
> >>>>>>> "namespaceMapperProperty") which can be used to set a property
> >>>>>>> recognized
> >>>>>>> by non JAXB-RI implementations and which will be used to set
> >>>>>>> namespace
> >>>>>>> mappers if needed.
> >>>>>>>
> >>>>>>> May be that can help ?
> >>>>>>>
> >>>>>>> Sergey
> >>>>>>>
> >>>>>>>
> >>>>>>>>      Ats,
> >>>>>>>>      Osvaldo Pina.
> >>>>>>>>
> >>>>>>>>
> >>>>>>>>
> >>>>>>>
> >>>>>>>
> >>>>>>
> >>>>>
> >>>>> --
> >>>>> Sergey Beryozkin
> >>>>>
> >>>>> Talend Community Coders
> >>>>> http://coders.talend.com/
> >>>>>
> >>>>> Blog: http://sberyozkin.blogspot.com
> >>>>>
> >>>>
> >>>>
> >>>
> >>
> >>
> >
> >
>
> --
> Daniel Kulp
> dk...@apache.org - http://dankulp.com/blog
> Talend Community Coder - http://coders.talend.com
>
>

Reply via email to