Hi,
Again the comments should rather self-explanatory; sometimes it happens that
a socket is closed due to an error, in which case it is still erroneously in
the fd_set, causing possible confusion if a new socket is opened and happens
to get the same fd. This patch simply clears out the "ready" bits when
processing the fds.
By the way, I should probably note that while many of these patches have
lived on a long time internally, and as such should be quite stable, there
might crop up new bugs in the forward-porting. Please let me know if there's
something that looks wrong or breaks -- I didn't write these patches, but I
could probably get in touch with those that did.
/* Steinar */
-- Software Engineer, Google Norway