It's definitely due to improper/non-existent teardowns and the 
ContextClassloader.   I'll have a patch ready shortly.

For everyones information, do NOT ever do:

Setup:
Thread.currentThread().setContextClassLoader(getClass().getClassLoader());
Teardown:
Thread.currentThread().setContextClassLoader(null);

That is wrong.    The original classloader may not be null.   Thus, after 
the teardown, the VM is not back to it's original state.   You need to do 
something like:

Setup:
origLoader = Thread.currentThread().getContextClassLoader();
Thread.currentThread().setContextClassLoader(getClass().getClassLoader());
Teardown:
Thread.currentThread().setContextClassLoader(origLoader);

Enjoy!
Dan



On Thursday 23 March 2006 09:46, Daniel Kulp wrote:
> Ick...   OK.  Another set of tests that only fail on my machine.  
> (like the rhino ones)
>
> I just did a fresh checkout into a completely new directory, deleted my
> ~/.m2/repository dir, and rebuilt and I'm still seeing the failures.
> Probably some more teardown methods that need to be implemented to
> reset things back to pre-test state.   I'll investigate more.
>
> Thanks for confirming.   I really wonder what it is with my machine
> that is exposing these problem.   Is anyone else using JDK 1.5.0_06 on
> Linux? I'm using ext3 filesystem which I expect everyone else is using
> as well.
>
> Dan
>
> On Thursday 23 March 2006 09:04, ant elder wrote:
> > svn st and svn up say I'm current and mvn clean, mvn for
> > container.java are successful for me.
> >
> >    ...ant
> >
> > On 3/23/06, Frank Budinsky <[EMAIL PROTECTED]> wrote:
> > > Hi Dan,
> > >
> > > Are the container.java tests run as part of "mvn" from the root? I
> > > can run "mvn clean" and then "mvn" from the tuscany/java directory
> > > and everything works fine for me.
> > >
> > > Could somebody else please confirm that they are also having
> > > problems like Dan?
> > >
> > > Thanks,
> > > Frank.
> > >
> > >
> > >
> > >
> > > Daniel Kulp <[EMAIL PROTECTED]>
> > > 03/22/2006 10:57 PM
> > > Please respond to
> > > tuscany-dev
> > >
> > >
> > > To
> > > [email protected]
> > > cc
> > > Frank Budinsky/Toronto/[EMAIL PROTECTED]
> > > Subject
> > > Re: new XSD to SDO mapping
> > >
> > >
> > >
> > >
> > >
> > >
> > >
> > > Frank,
> > >
> > > This commit seams to have REALLY broken the sca builds.   The
> > > container.java tests are failing pretty badly.   Are you (or
> > > someone else) working on fixing them?
> > >
> > > For now, I'm going to revert my SDO dir to -r 387955.
> > >
> > > Thanks!
> > > Dan
> > >
> > > On Wednesday 22 March 2006 17:55, Frank Budinsky wrote:
> > > > I just committed a change to the SDO XSDHelper.define() method
> > > > which changes the behavior significantly. It used to mangle names
> > > > (using EMF's mangling algorithm) and it didn't map simple types
> > > > (like xsd:int) to the proper SDO Types as specified in the SDO 2
> > > > spec.
> > > >
> > > > With this change, many of the sca schemas no longer generated
> > > > valid Java classes. For example things like <element
> > > > name="binding.java" ...> now produces a get method named
> > > > getBinding.java(), which doesn't compile because of the "."
> > > > character. Don't blame me, that's what the SDO spec says it
> > > > should do :-)
> > > >
> > > > The solution is to add sdo annotations (e.g.,
> > > > sdo:name="bindingJava") in the schema with the bad names. I fixed
> > > > all of the existing places with this problem ... seemed like
> > > > hundreds of them :-) ... so everything is still working. Please
> > > > beware of this issue, though, when adding new schemas.
> > > >
> > > > Longer term, we might want to think about adding some mangling to
> > > > the Tuscany SDO generator as an option (value-add feature).
> > > >
> > > > Frank.
> > >
> > > --
> > > J. Daniel Kulp
> > > Principal Engineer
> > > IONA
> > > P: 781-902-8727  C: 508-380-7194
> > > [EMAIL PROTECTED]

-- 
J. Daniel Kulp
Principal Engineer
IONA
P: 781-902-8727  C: 508-380-7194
[EMAIL PROTECTED]

Reply via email to