worrying comment from John McKay

[Daniel Allcock <allcock@math.utexas.edu>]

Hi all,

John McKay told me something unlikely but theoretically worrying, and I'd like to check into it.  Namely, that some integer-to-integer functions use floating point ops for speed, and that the reconversion to integers at the end of the calculation isn't or wasn't "properly" justified.  Not quite sure what this means, but of course I do want to trust my results.  What's the story?  (If it matters, I'm using 2.4.3 on a mac, with gmp).

thanks,
Daniel