I am using the select function for communication between my sockets. I have a while loop and I do -
while(!done) {
FD_ZERO(&read_flags);
FD_ZERO(&write_flags);
FD_SET(comm_fd1, &read_flags);
FD_SET(comm_fd2, &read_flags);
FD_SET(STDIN_FILENO, &read_flags);
FD_SET(comm_fd1, &write_flags);
FD_SET(comm_fd2, &write_flags);
FD_SET(STDIN_FILENO, &write_flags);
//call select
sel = select(comm_fd1+comm_fd2+1, &read_flags, &write_flags, (fd_set*)0, &waitd);
and the same with different variables on the client side. I got this basic technique from a tutorial online and just went with it. Then it hit me - why do I clear the set and add file descriptors each time I loop? If they are already added, why clear them and add again? So I tried only doing this once before the while, and the code does not work the same anymore. Can someone explain why? Is it just because select modifies the contents of the set? Any help and/or insight is appreciated.
When select returns, it has updated the sets to show which file descriptors have become ready for read/write/exception. All other flags have been cleared.
It's important that you re-enable the file descriptors that were cleared prior to starting another select, otherwise, you will no longer be waiting on those file descriptors.
As for re-clearing, it can be a good habit to get into, since if you need to change the set of file descriptors (such as adding a newly opened socket to the read set), you'll want to clear it and rebuilt it every time, so that it's correct as the state of the program changes.
Is it just because select modifies the contents of the set?
Yes, after select returns, only ready descriptors are left within the sets.
来源：https://stackoverflow.com/questions/7637765/why-fd-set-fd-zero-for-select-inside-of-loop