gcc 4.0.3 double math/conversion error

[Ben Bridgwater]

Binary package hint: gcc

Running gcc (4.0.3) from Ubuntu 6.10 on a Pentium 4:

double d = 10.1;
int i = (int) (d * 10); // = 100

double t = d * 10;
int j = (int) t; // = 101

I would expect to get the same result (whatever it may be) in both cases.

To reproduce:

#include <stdio.h>

int main()
{
  float f = 10.1;
  double d = 10.1;

  double t;

  printf("%i %i %i\n", (int) (f * 10), (int) (d * 10), (int) (t = (d * 10)));

  // gcc 4.0.3 (Ubuntu 6.10) output is: 101 100 101

}

gcc test.c

a.out

[Matthias Klose (doko)]

fixed in edgy and newer releases; (gcc-4.1 and gcc-4.2)