Geert Bevin wrote:
Hi Eddy,
not that I recall. Care to expand on this?
Best regards,
Geert
Geert,
After our online conversation yesterday, I did some more tests to try
and pinpoint the exact source of the problem. I think I've managed to
establish that the problem is related to the RIFE classloader in some way.
But first, here is some background information for the others on the
mailing list.
I am using JPOX, a JDO implementation, to persist objects to a database.
In JDO, JDOHelper.getPersistenceManagerFactory(Properties properties) is
used to instantiate the PersistenceManagerFactory implementation that is
specified in the properties (that is, introspection is used to locate
and instantiate the correct class).
In my tests, when that method is called from a console application or a
JSP page (RIFE not involved), no exception is thrown; all classes are
loaded and instantiated successfully. However, when the same method is
called from a RIFE Element, a LinkageError exception is thrown for the
Level class from the Log4j package. That exception triggers a chain of
other exceptions that prevent the PersistenceManagerFactory from being
instantiated.
After much fiddling around (until 2am to be exact), I tried
instantiating the PersistenceManagerFactory class with a classloader
specified. I called the variant method
JDOHelper.getPersistenceManagerFactory(<properties>, <classloader>),
passing 1) the current thread context classloader and 2) the parent
classloader of the current thread context classloader.
In case 1) the classloader is the RIFE classloader. That caused the
LinkageError exception to be thrown.
In case 2) the classloader is the Catalina WebappClassloader (I am using
Tomcat), and with that, the PersistenceManagerFactory class was
instantiated successfully.
This is as far as I reached in my investigation. I wanted to share this
here in order to get your feedback on it. If you have encountered this
problem in a differnet scenario, please let me know. As you rightly
pointed out, such conflict errors are very hard to trace, so I would be
very grateful if you could help in any way.
If my analysis is wrong (again!), I apologise; as I said, I stayed up
until quite late to try and sort this out.
Regards,
Eddy
--
http://coding.mu
http://priscimon.com/blog
_______________________________________________
Rife-users mailing list
[email protected]
http://lists.uwyn.com/mailman/listinfo/rife-users
