OK, I admit. I have lied. A little bit. I have not executed the test on a BE
platform, I have simply typed what I knew the output would be. But I got
away with it, heh heh.
But now I did it for real, on both architecture types, because I got curious
about the floats...
So:
#include <fcntl.h>
#ifndef _T
#define _T(x) x
#endif
struct endian_test_tag
{
char abc8[4];
short abc16[4];
long abc32[4];
float fpie;
} endian_struct =
{
{'A', 'B', 'C', 0},
{65, 66, 67, 0},
{65, 66, 67, 0},
1.234
};
main()
{
int fh;
int nbytes;
printf(_T("sizeof(endian_struct) = %d\n"), sizeof(endian_struct));
fh = open(_T("abc.out"), O_CREAT | O_TRUNC | O_RDWR, 0666);
nbytes = write(fh, &endian_struct, sizeof(endian_struct));
printf(_T("%d bytes written.\n"), nbytes);
close(fh);
}
results in:
* on LE:
0000:0000 41 42 43 00 41 00 42 00-43 00 00 00 41 00 00 00 ABC A B C A
0000:0010 42 00 00 00 43 00 00 00-00 00 00 00 B6 F3 9D 3F B C
??�?
* on BE:
0000:0000 41 42 43 00 00 41 00 42-00 43 00 00 00 00 00 41 ABC A B C
A
0000:0010 00 00 00 42 00 00 00 43-00 00 00 00 3F 9D F3 B6 B C
?�??
So, the floats are 'upside down' too. Makes sense. IEEE defines them in
terms of bits. And basically one can cast a float and consider it a double
word for transmission purposes. Fortunately, otherwise the whole thing would
be even a bigger mess than it already is.
Anyway, this should cover the floating point part. And thus give no excuse
to anyone to revive this thread. Unless for a better cause... ;)
Lars
