bug in cl (c compiler) by Dmitry Volkinshtein

The following program gives incorrect output when compiled with optimization on x86 machine


#include <stdio.h>
#include <limits.h>

int main(int argc, char *argv[])
{
    volatile int i10 = 10;
    printf("%d\n", -(INT_MIN/i10));
    getchar();
}

The problem seems to be the following:
During the optimization process, the compiler tries to save the '-' at the end of the calculation by changing the sign of the constant dividend first instead. It is ok for all constant ints except INT_MIN because of the 2's complement feature: -INT_MIN==INT_MIN and the result of (-INT_MIN)/10 is negative. The output of the program is correct for other constants, and when the optimization if off (Debug)