see http://caml.inria.fr/mantis/view.php?id=6749

On Sat, Apr 18, 2015 at 7:14 PM, Tao Stein <taostein@gmail.com> wrote:

>
> I'm getting a strange result using ocamlopt that seems to violate the
> basic axioms of integer division.
>
> Some context first:
> taostein@~/b2/f2: which ocamlopt
> /Users/taostein/.opam/4.02.1/bin/ocamlopt
> taostein@~/b2/f2: ocamlopt -o foo.opt foo.ml
> taostein@~/b2/f2: which ocamlc
> /Users/taostein/.opam/4.02.1/bin/ocamlc
> taostein@~/artcode/b2/f2: ocamlc -o foo foo.ml
>
> Built with ocamlopt, the following code throws an exception. Built with
> ocamlc it does not.
>
> let k1 = 1;;
> let rec f i k =
>   if (((i / k1) * k1 + (i mod k1)) <> i)
>   then (raise (Failure "violation of integer division axiom"))
>   else (if (i < 10) then (f (i + 1) k) else ()) in
> f 0 1;;
>
> Now, if I change the above only on line 3 changing k1 to k, then it works
> fine, no problem.
>
> let k1 = 1;;
> let rec f i k =
>   if (((i / k) * k + (i mod k)) <> i)
>   then (raise (Failure "violation of integer division axiom"))
>   else (if (i < 10) then (f (i + 1) k) else ()) in
> f 0 1;;
>
> No exception in either ocamlc or ocamlopt.
>
> Using ocamlopt, it seems 1 mod 1 = 1. This is problematic.
>
> Tao Stein / 石涛
>
>