divide overflow

[Jason Schonberg]

>Jason Schonberg <schon@mips.com> writes:

>> _____
>> #include <limits.h>
>> 
>> int main()
>> {
>>  long int A, B, C;
>>  A = INT_MIN;
>>  B = -1;
>>  C = A/B;
>> }
>> _____

Ulrich Drepper writes:

> This is known and the result is correct.  The hardware simply signals
> that an invalid operation is performed since the result cannot be
> represented.  If other platforms do not check this and you depend on
> this the program is simply not portable.

I understand that in two's compliment notation, the lowest negative integer has 
an absolute value one larger than the maximum positive number.  Thus I would 
expect an exception.

What puzzles me more is that the following code _DOES NOT_ display similar 
behavior.  (again egcs 1.1.2 - X86 Linux Redhat 6.0)

#include <limits.h>

int main()
{
 long int A, B, C;
 A = INT_MIN;
 B = -1;
 C = A*B;
} 

-- 
Jason Schonberg
schon@mips.com