1. Sublcassed ObjectInputStream to allow me to set the classloader I want.

2. Created a sublcass of EJBRequest that also implements EJBInvocation
   and reads the data in a lazy fasion.  By the time the arguments are
   needed, the right classloader has been set on the ObjectInputStream by
   the EJBRequestHandler and all is good.

-David

On Mon, Apr 12, 2004 at 12:11:35PM -0500, Dain Sundstrom wrote:
> How did you get this to run?
> 
> -dain
> 
> /*************************
>  * Dain Sundstrom
>  * Partner
>  * Core Developers Network
>  *************************/
> 
> On Apr 12, 2004, at 3:51 AM, David Blevins wrote:
> 
> >Still quite a few failures, but minimal remote server functionality is
> >working.  I whipped up an EJBInvocation implementation that reads the
> >data off the stream as it passes through the interceptor stack in the
> >container.  As soon as we are in a post-release cycle, I'm going to
> >stream-line that even further.
> >
> >Test results are:
> >Tests run: 112,  Failures: 27,  Errors: 6
> >
> >Couple things causing problems:
> >
> > - DatabaseBean can't get a datasource, so no BMP tests run. (open)
> >
> > - BeanPolicy$2.invoke throwing "Not yet implemented" is preventing
> >   nearly all the SFSB tests from running. (open)
> >
> > - Still having classloader issues on outbound proxy replacements,
> >   so many RMI-IIOP tests are failing. (working on it)
> >
> >All in all the test suite is doing it's job.
> >
> >-David

Reply via email to