Thread: Endiana Jones and The Last Crash-ade

How about something like:

Code:

itsme@dreams:~/C$ cat endiantest.c
#include <stdio.h>
#include <netinet/in.h>

#define IS_BIG_ENDIAN (0x12345678 == htonl(0x12345678))

int main(void)
{
  printf("This is a %s endian machine.\n", IS_BIG_ENDIAN?"big":"little");

  return 0;
}

On my spiffy little Intel box I get:

Code:

itsme@dreams:~/C$ ./endiantest
This is a little endian machine.

What do you mean by endian-independent? Do you have a networked application and you want it to work if the client is big-endian and the server is little endian? Have you looked at using the standard conversion macros, i.e. htonl, htons, ntohl, ntohs?

In general the 'h' stands for host and the 'n' stands for network, so you can read the macros as "htonl: host to network long", "htons: host to network short", etc. Host is whatever the native platform byte-order is, network-byte-order is always big-endian.

So, on a SUN machine (big-endian) htonl is a no-op, while on a windows little-endian machine, it will convert a 32-bit number from big endian to little endian format.

So when you are packing your data structures to send, you call either htons or htonl, and when you are unpacking your data structures in the receiving code, you call ntohs or ntohl.

Originally Posted by SMurf

It's not a network app

So you're storing integers or floats in a file? Why not use text?

Originally Posted by SMurf

It's not a network app, so I don't see the necessity in having to include the sockets library just to get htonl and htons. I've got my own functions to do that job (Which amount to a few bit shifts).

I mean, I'll do it if it's the only way, but isn't there anything else?

And your own functions know whether the host system is big or little endian somehow? Why not use that same method then? See, on big endian machines all those htonl()-ish functions really do nothing at all. They only do something on little endian machines. So if your functions already can tell what type of architecture it is then I don't understand the point of this thread.

I used htonl() because it will do different things on different architectures in just the right way for you.

And also, at least on my system, there is no additional library that needs to be added. All you have to do is #include the header file. I don't know if gcc silently adds -lnet to the compile line or if those network functions are tucked away in libc, but I don't have to explicitly link in another library.

Why bother at all then? Save everything a byte at a time, using a format that you come up with. Need RGB sets? Fine, save the first byte for R, the second for G, the third for B, or whatever. Your image format shouldn't care what endian anyone who uses it is.

Quzah.