Question

When writing values to a file in C under Windows, I’m getting something which seems to be wrong, and the result differs from the same program run in Cygwin.

In this case, I’m writing a float to a file:

#include <fcntl.h>
#include <stdio.h>
#include <stdlib.h>
#include <io.h>

int         main( void )
{
    int     fd;
    float   f;

    fd = open("file.a", O_RDWR | O_CREAT);
    f = (float)atof("-0.1352237");
    printf("%.7f\n", f);

    write(fd, (void *)&f, sizeof(float));
    close(fd);
    printf("sizeof(float): %d\n", sizeof(float));

    return (EXIT_SUCCESS);
}

I’ve compiled and run this file on both Windows and in Cygwin (without the include), and I’m not getting the same result. Since I’m writing a float, I would expect the output file to have 4 bytes written to it

However, the output when compiled with cl.exe in the command line, seems wrong:

1578 0d0a be

If I read from the file to a float, I’m not getting the correct value, obviously. And the amount of bytes written to the file is wrong to, it should only be 4 bytes, not 5.

This is what I get when running in Cygwin:

1578 0abe 

This is correct. If I read this from the file to a float I get the correct value (-0.1352237).

Am I missing something here ? Are the windows implementation of those functions flawed ?
Or maybe I’m not writing the data properly ?

Answer 1

You have opened the file in text mode (the default), so the 0a ("\n") is being translated to 0d0a ("\r\n").

You need to open the file in binary mode, using the O_BINARY flag.