I wrote a 2 versions of a battleship game. I wrote a text-base game with socket code. So client and server talk to each other though the socket. So the way I lay out my code is that, I constantly check the socket to see if anything come in, if there is, then I record it as an attack, then I attack back. Now I try to do the same thing but in graphics-base with openGL. I stumble on deciding where to put the piece the client code. I cant put it in main() because, it will hang on to the while loop that I check to see if anything come into the socket. Therefore It will never get to glutMainLoop(), which process all my graphics event. I guess my question is that, is there a callback function that constantly check the socket, if not, how do I implement socket in this situation. Thank you.

Its simply a case of not hanging on the socket listen. Look up the function select(). This examines a set of 1 or more sockets and tells you their status for reading and writing.

Also, a more complicated (and in this case rather unnecessary) alternative solution is to put the opengl renderer in its own thread.

Can you elaborate more on how to use select() to detect things that coming into the socket. Because I use select(), but not to detect things from socket, but to see if any new client show up. So I use select() in my server code. The server detect clients showing up, and pair up 2 clients. when client attack, the server just pass the attack from 1 client to another, so the client wait to see if any thing come in from the socket. That is how I code it. I think it more like a client base game rather than a server base game.

select() just tells you if there is any activity on the sockets you supply it with. You only need to read from sockets where select() has told you there is information to be read. Just add all relevant sockets to the FD_SET, pass that in as the readfds parameter of select, and examine the output to see which sockets need reading from.

If your client application is running a frequent loop anyway, I believe that polling the (nonblocking) socket is just as good an approach as any.

So just call a socket polling function inside your main loop, which reads as much data as there is on the socket, breaks it up / recombines it into distinct messages and processes as many (complete) messages as it has through your message processing function.

This isn't really terribly difficult, certainly much more straightforward than faffing around with select() etc (although that's not hard either).

Likewise, keeping your game entirely single threaded will make it easier to write (albeit may not use multiple CPUs / cores efficiently).

If you're using a fixed-timestep-with-interpolation system for rendering, then you probably want to do network polling inside your timestep processing function.

Mark