GCJ 3.3 crash with string handling reproduceable case

[Ranjit Mathew]

> I've put together a small program that crashes eventually. This
> is not a livelock, but I created the app when I was trying
> to create a usable bugreport on the livelock problem.

"eventually" is an understatement - it takes VERY long on my
machine to crash (but it does crash). I can reduce the crashing
time by using "gij -mx=???". 

The really frustrating part is that I get a Windows GPF every 
time, but the stack trace shown by the "drmingw" JIT-windows-debugger 
is always never from the thread that really caused the fault 
and is always quite different.

Just once, I got a "SuspendThread failed" dialog box from the GC...

I really suspect that it is some problem with the GC but I do not
know how to trace it. :-(

BTW, as Jeff had pointed out earlier, the "gctest" from the GC
always crashes immediately for me as well with a GPF. :-(

Ranjit.