Swaroop Sridhar wrote:

>But, getting a run-time exception is a little too late.
>...
>
>Probably, the best solution in the case of non_mod_+
>and non_mod_- is that if the compiler (to be understood as
>compiler-prover combination) is unable to trivially (for
>large values of trivially, which is decidable) show that
>the expression is guaranteed not to overflow, then it
>is a compile time error, and the programmer should choose
>the mod_+ and mod_- operators instead.
>
It should be noted that I was not against having an
operator that throws a run-time exception.
The point is that there should be a facility to statically
(and conservatively) detect all possible overflows,
for applications in which it matters.

Swaroop.


_______________________________________________
bitc-dev mailing list
[email protected]
http://www.coyotos.org/mailman/listinfo/bitc-dev

Reply via email to