nontrapping math in Java

[Bryce McKinlay]

Richard Henderson wrote:

>On Mon, Mar 25, 2002 at 12:18:07PM +1200, Bryce McKinlay wrote:
>
>>My testing with the small "Sieve.java" on PowerPC @ -O2 showed no 
>>difference to generated code as a result of -fnon-call-exceptions, 
>>except for more unwind info being generated.
>>
>
>Err.. Java already sets flag_noncall_exceptions, so you wouldn't
>expect to see any changes by adding the flag again.
>

I modified java/lang.c to not set it (and changed libgcj.spec).

Bryce.