> I don't have the Knuth books at hand now, but if we are implementing
> that method then we have to use his rule of election. Can someone
> check?
Well, I have read a little about this issue. Our i64 implementation defines
the
quotient as truncate(a/b), like many C implementations. For that reason the
remainder has the same sign as the dividend (like C99). I wrote a patch to
fix
the mod result sign considering that and some tests to prove it.
Thanks for investigating and fixing this.
OTOH, could someone test this patch on a win32 platform and evaluate the
results
as for FS#105 ? I believe they are related.
I will try tomorrow in a Windows machine.
Patch follows,
Applied.
--
Jose E. Marchesi <[email protected]>
http://www.jemarch.net
GNU Project http://www.gnu.org
