Strange behaviour of try..finally

[Laurent Bossavit]

 Hello all,

 Me again. Still trying to get somewhere with Kaffe on FreeBSD. I've
gotten past the 'classes.zip' problem; removed the hard-coded call to
"initializeSystemClass" (I'm using a classes.zip which doesn't have
it; a good fix would be to only call that initializer if it actually
exists). Passed all the tests in 'make test'. Now getting ready to
use it on a real-world application, Sun's Java Web Server. 

 I'm running into a very weird behaviour with a try..finally construct
which catches an exception in the finally block. The problem boils
down to the following test case :

import java.io.*;

public class FinalTest {

 public static void main(String[] args) {
  new FinalTest();		
 }

 public FinalTest() {
  System.out.println(tryfinally());
 }

 public String tryfinally()
 {
  String yuck = null;
  String good = "Perfect";
  try {
   try {
    return good.toLowerCase();
   }
   catch (Exception e) {
   }
   return "I can handle that";
  }
  finally {
   try {
    String x = yuck.toLowerCase();
   }
   catch (Exception e) {}
  }
 }

 Compile and run under JRE or JDK; this outputs "perfect". Compile
and run under Kaffe; this dies with a NullPointerException which the
try..catch construct in the finally block does *not* catch.

 I'll be diving into the code forthwith; naturally though, I'll be
glad if anybody can come up with some idea of why this happens.

 Thanks,


=====================================================
Laurent Bossavit
NetDIVE Paris offices                      (o o)
--------------------------------     -oOO--( )--OOo-
http://www.netdive.com/
NetDIVE: the leader in virtual meeting technology
=====================================================