Charles-François Natali added the comment: > It appears that Linux's "spurious readiness notifications" are a deliberate > deviation from the POSIX standard. (They are mentioned in the BUGS section > of the man page for select.)
I don't think it's a deliberate deviation, but really bugs/limitations (I can remember at least one occurrence case where a UDP segment would be received, which triggered a notification, but the segment was subsequently discarded because of an invalid checksum). AFAICT kernel developers tried to fix those spurious notifications, but some of them were quite tricky (see e.g. http://lwn.net/Articles/318264/ for epoll() patches, and http://lists.schmorp.de/pipermail/libev/2009q1/000627.html for an example spurious epoll() notification scenario). That's something we have to live with (like pthread condition spurious wakeups), select()/poll()/epoll() are mere hints that the FD is readable/writable... Also, in real code you have to be prepared to catch EAGAIN regardless of spurious notifications: when a FD is reported as read ready, it just means that there are some data to read. Depending on the watermark, it could mean that only one byte is available. So if you want to receive e.g. a large amount of data and the FD is non-blocking, you can do something like: """ buffer = [] while True: try: data = s.recv(8096) except BlockingIOError: break if data is None: break buffer += data """ Otherwise, you'd have to read() only one byte at a time, and go back to the select()/poll() syscall. (For write ready, you can obviously have "spurious" notifications if you try to write more than what is available in the output socket buffer). > Should I just apply the following patch to the default branch? LGTM. ---------- _______________________________________ Python tracker <rep...@bugs.python.org> <http://bugs.python.org/issue16507> _______________________________________ _______________________________________________ Python-bugs-list mailing list Unsubscribe: http://mail.python.org/mailman/options/python-bugs-list/archive%40mail-archive.com
