> -----Original Message-----
> From: Stephen McConnell [mailto:[EMAIL PROTECTED]
> Sent: Tuesday, December 23, 2003 4:35 PM
>
...
>
> I'm trying to think why this would happen.  Basically if I understand
> correctly, Merlin has created a classloader and populated it with the
> content you have specified in you block.  The scanner then kicks in to
> establsh the available components by read the metainfo from the .xinfo
> resources.  The TypeBuilder is supplied the component class from which
> it gets the class's classloader to read the xinfo.  In you the classloader
> is not finding the xinfo.
>
> This means one thing - the class in question already exists in a parent
> classloader - but the xinfo is not present in that parent.  This suggests
> that maybe you have a duplication of classes referenced in the jar file
> component-impl-0.1.jar and classes in target/classes.
>
> >
> >classLoader is the classloader that loaded the class itself, it
> should also
> >load the .xinfo file in the same package. Or is it a classloader issue?
> >
>
> There is clearly a conflict going on at the classloader level.  The class
> exists but its not being loaded from the jar file. To find out
> what is going
> on, set the following properties in merlin.properties
>
>   merlin.debug = true
>   merlin.info = true
>
> Then run maven with debug enabled:
>
>   maven -X test
>
> As the test cases kicks you will see merlin system debug messages and you
> will also see junit messages.  From this you should be able to figer out
> where the class is comming from.
>

Thanks, Steve, for your tireless efforts. FYI: I am using eclipse and ran
all my testcases directly from the workbench. Following your advice I tried
to switch to maven for running the testcase in question, yet found it
somehow difficult to get the setup right. To be more specific: I wasn't able
to tell AbstractMerlinTestCase to use my local maven repo as one of its
hosts. Yes, I specified a custom kernel.xml containing my maven repo as a
<host> entry via "merlin.kernel", but this wouldn't work either. Could be a
good idea to make the list of hosts to be used configurable.

In the end, I rewrote AbstractMerlinTestCase in order to have more control.
I let my testcase derive from this new base class, kicked in maven, et
voila: all's well. I run the same testcase from the workbench: it fails with
the usual exception, i.e. missing xinfo file. That's just not fair but goes
to show that it's indeed very likely a classloader issue. Anyway, I'll stick
to maven.

Cheers,

Olaf


---------------------------------------------------------------------
To unsubscribe, e-mail: [EMAIL PROTECTED]
For additional commands, e-mail: [EMAIL PROTECTED]

Reply via email to