Setting FLT_EVAL_METHOD not working

Question

#include <stdio.h>
#include <float.h>

#ifdef FLT_EVAL_METHOD
    #undef FLT_EVAL_METHOD
    #define FLT_EVAL_METHOD 1
#else
    #define FLT_EVAL_METHOD 1
#endif

int main(void)
{
    printf("%f\n%f\n", FLT_MAX, DBL_MAX);
    float fp = FLT_MAX;
    printf("%f\n", fp + 1);
    printf("%f\n", FLT_MAX + 1);
    printf("%f\n", (float) FLT_MAX + 1);
}

The expected output is:

340282346638528859811704183484516925440.000000
179769313486231570814527423731704356798070567525844996598917476803157260780028538760589558632766878171540458953514382464234321326889464182768467546703537516986049910576551282076245490090389328944075868508455133942304583236903222948165808559332123348274797826204144723168738177180919299881250404026184124858368.000000
340282346638528859811704183484516925441.000000
340282346638528859811704183484516925441.000000
340282346638528859811704183484516925441.000000

But the actual output is:

340282346638528859811704183484516925440.000000
179769313486231570814527423731704356798070567525844996598917476803157260780028538760589558632766878171540458953514382464234321326889464182768467546703537516986049910576551282076245490090389328944075868508455133942304583236903222948165808559332123348274797826204144723168738177180919299881250404026184124858368.000000
340282346638528859811704183484516925440.000000
340282346638528859811704183484516925440.000000
340282346638528859811704183484516925440.000000

Since FLT_EVAL_METHOD is set to 1, shouldn't all float expressions evaluated as double s? What does "evaluate float and double as double, and long double as long double." mean? Are float s treated identical to double s?

Answer 1

FLT_EVAL_METHOD is not a switch that the programmer uses to control the compiler. It's a means for the compiler/platform to report to the programmer how floating-point expressions will be evaluated. You're not supposed to change its value, and doing so is not supposed to have any effect on anything.