GCC bug report

["Yeh Hong-ming" <hongming at cs dot utexas dot edu>]

Hi,

I believe that I have found a bug in GCC.  The report is enclosed below.

Thanks,
Yeh Hong-ming
Doctoral Student
Department of Computer Science
The University of Texas at Austin
U.S.A.

Version of GCC: gcc version 3.3.1 (cygming special)
Input file:
-------------
#include <stdio.h>
#include <netinet/in.h>

static const float p = 0.9;

void f(long* v) {
    long q = htonl(*(long*)&p);
    *v = *(long*)&q;
}

void g(char* v) {
    long q = htonl(*(long*)&p);
    *v = *(char*)&q;
}

int main(){
    int x;

    f((long*)&x);
    x = ntohl(x);
    printf("%s\n", *(float*)&x == p ? "OK" : "FAIL");

    g((char*)&x);
    x = ntohl(x);
    /* Expected OK but got FAIL */
    printf("%s\n", *(float*)&x == p ? "OK" : "FAIL"); 
}
-------------
GCC command arguemnts: none.  (Simply gcc test.)

Type of machine: Cygwin (DLL version 1.5.3) over Windows XP on Intel Pentium
III

GCC configure/modification: None (Cygwin binary release without change)

Bug behavior: htonl() and ntohl() do not work correctly under certain
combination typecasts.  The program is supposed to print two "OK OK", but it
prints "OK FAIL:.