> Seems Mandrake Linux is not the only platform where roundoff behavior is
> less IEEE-perfect than Thomas would like it to be. Perhaps we need a
> slightly more robust approach to controlling roundoff error.
Go ahead. istm that asking modulo, trunc, etc to Do The Right Thing is
not a big deal, and it would be better to understand how to build
executables that can do math.
Certainly better than writing a bunch of extra checking code to work
around the inability of a compiler (or compiler options) to do IEEE
math. It *is* a standard, ya know ;)
- Thomas