[kaffe] Bug Report: in particular for rmi

[Dalibor Topic]

Salut Jeremy,

Jeremy Buisson wrote:
> In kaffe 1.1.0 and also in cvs: I don't exactly know where the true 
> problem resides, but I can describe the symptoms...
> 
> This concerns the kaffe setup shell script, the internal class loader, 
> and it appeared when I used RMI. For some reason, the 
> kaffe.rmi.rmic.RMIC has a null ClassLoader. Then, when it tries to load 
> the class for which stub and skeleton should be generated, the VM falls 
> back to the internal class loader. Because the kaffe setup shell script 
> (jre/bin/kaffe) sets the BOOTCLASSPATH environment variable to only 
> reference runtime jars, rmic is only able to handle the classes of the 
> runtime library. It is thus unusable.

thanks for the bug report, and the excellent analysis. I didn't react to 
it, as my plan was to switch over to the better maintained RMI 
implementation from GNU Classpath, and then address the issues together 
with Classpath developers.

Now that the merge of GNU Classpath's code is over, it should work 
again. I've modified a Class.forName call to use the SystemClassLoader 
to avoid problems with the BOOTCLASSPATH not containing compiler classes.

cheers,
dalibor topic