I have a commandline php script, which opens a nonblocking socket to a server. Inside the message loop, where it polls the socket to determine whether there is any data to read/write, I've been running into problems, where the application takes as much CPU time as it can (80%, 90%, sometimes 99%, etc)... Normally, using socket_select() with a timeout should allow allow it to poll the socket without using all the cpu time, but it isn't working. I've tried with timeouts of 15ms, 100ms, 200ms, 500ms, 1sec, and it's still using 100% cpu time. Even adding sleep() and usleep() calls will not relieve the system load that the script causes. For now, I've fixed the problem temporarily by not setting the socket in nonblocking mode (so it blocks when reading), but the reason I need to use nonblocking is that I'm adding more to it that will require that the script can respond to events *between* reading from the socket. Can anyone explain why socket_select and other timeout function calls would still cause it to use 100% cpu time? If it makes a difference, the script is on gentoo linux, and it includes a fork() call, to fork into the background (though this shouldn't make a difference).
Is it possible that socket_select() ignores the timeout values, if the sockets are nonblocking? Or would I be able to use socket_select() on a blocking socket, to simulate the effect of a nonblocking socket (so that I would only call socket_read() IF socket_select() indicates that there is data waiting.) Thanks Joe -- PHP General Mailing List (http://www.php.net/) To unsubscribe, visit: http://www.php.net/unsub.php
