(j3.2006) Fwd: (SC22WG14.11572) What is INT_MIN % -1?

[Craig Rasmussen]

FYI, here is the original question originating on the WG14 list.

Regards,
Craig

Begin forwarded message:

> From: David Svoboda <svoboda at cert.org>
> Date: October 23, 2008 11:32:44 AM MDT
> To: "sc22wg14 at open-std.org" <sc22wg14 at open-std.org>
> Cc: David Svoboda <svoboda at cert.org>
> Subject: (SC22WG14.11572) What is INT_MIN % -1?
>
> Does the C standard dictate what INT_MIN % -1 should evaluate to?
>
> The best citation I can find is section 6.5.5, paragraph 6:
>
>    When integers are divided, the result of the / operator is the
>    algebraic quotient with any fractional part discarded.) If the
>    quotient a/b is representable, the expression (a/b)*b + a%b shall
>    equal a.
>
> So a reasonable interpretation would be that INT_MIN % -1 = 0.
>
> But I think another reasonable interpration exists. On
> two's-complement machines, INT_MIN / -1 is mathematically equivalent
> to INT_MAX + 1, which cannot be represented as a signed
> int. Consequently the condition in the second statement in the above
> quotation fails, and so the above quotation leaves INT_MIN % -1 in the
> realm of undefined behavior.
>
> On MSVC++, taking the modulo of INT_MIN by -1 yields the value 0.  But
> on gcc/Linux, taking the modulo of INT_MIN by -1 produces a
> floating-point exception.
>
>
> So what is going on here? Is there another section of the C standard
> that defines INT_MIN % -1? Or is one of the above interpretations
> better than the other?  Or are both reasonable, and INT_MIN % -1
> really is undefined for two's-complement machines?
>
> Put another way, is MSVC++ non-complaiant with the standard in this
> facet? Is GCC? Or are they both standards-compliant here?
>
> -- 
> David Svoboda <svoboda at cert.org>
> Software Security Engineer
> CERT Coordination Center
> (412) 268-3965

-------------- next part --------------
An HTML attachment was scrubbed...
URL: http://j3-fortran.org/pipermail/j3/attachments/20081029/a5adacc1/attachment.html