Hi,
I'm trying to move an iterative server (a small multiplayer card game)
from using select() to poll() (BTW is it a good idea at all?)
I have to use iterative design instead of a forking one, because it is
easier to move players between tables and the chatroom this way.
So it is important for me that none of the reads/writes/accepts block.
My question is, what events should I poll() ? With select() it is easy -
read/write won't block if FD_ISSET is set for that socket and:
1) read/write returns > 0 means ok
2) read/write returns = 0 means connection closed
3) read/write returns < 0 means connection interrupted
But with poll() it is unclear. The Stevens' book says it is implementation-
specific, but the OpenBSD manpage doesn't describe, what is considered
high-priority data (for example, I listen on a non-blocking socket. If a new
client connects, is it high-priority or not? Will POLLIN suffice here?)
If I poll() for POLLIN/POLLOUT, will the read/write ever block?
Do I need to poll() for POLLHUP and/or POLLERR to detect
closed or interrupted connections or is POLLIN enough and it
is same as 2) and 3) above?
Regards
Alex
