[freenet-devl] Re: Yet another GCJ update. (a/k/a PR 1615 followup)

Fri Feb 23 13:31:00 GMT 2001

On 23 Feb 2001, Tom Tromey wrote:
> >>>>> "Mark" == Mark J Roberts <mjr@statesmean.com> writes:
>> Mark> However, I've observed the problem Warren was talking about,
> Mark> where mod will sometimes return negative values, in modPow. This
> Mark> is almost certainly the sole cause of the bogus results. Here's
> Mark> one example (run with a fresh checkout from two weeks ago,
> Mark> because the current tree won't build):
>> My best guess is that it will take us (meaning Java hackers at RH) a
> while before we can get to this one. If you're motivated to write a
> patch, that would be great ...

We fixed this, but I forgot to CC your list. Per's already committed the
patch:
Index: BigInteger.java
===================================================================
RCS file: /cvs/gcc/gcc/libjava/java/math/BigInteger.java,v
retrieving revision 1.10
diff -r1.10 BigInteger.java
857c857,865
< remainder.set(ywords, rlen);
---
> if (ywords[rlen-1] < 0)
> {
> ywords[rlen] = 0;
> remainder.set(ywords, rlen+1);
> }
> else
> {
> remainder.set(ywords, rlen);
> }

But don't worry, we might have another bug yet...
-- 
Mark Roberts
mjr@statesmean.com