Relay-Version: version B 2.10 5/3/83; site utzoo.UUCP
Posting-Version: version B 2.10 5/3/83; site cvl.UUCP
Path: utzoo!linus!philabs!seismo!rlgvax!cvl!joe
From: joe@cvl.UUCP (Joseph I. Pallas)
Newsgroups: net.emacs
Subject: Re: guido mchan fix (LONG MSG) - (nf)
Message-ID: <566@cvl.UUCP>
Date: Fri, 19-Aug-83 15:54:35 EDT
Article-I.D.: cvl.566
Posted: Fri Aug 19 15:54:35 1983
Date-Received: Sun, 21-Aug-83 02:55:32 EDT
References: <2597@uiucdcs.UUCP>
Organization: U. of Md. Computer Vision Lab
Lines: 24

Purtilo@uiuccsb.UUCP claims that the select call should be changed from

the old line ...
  nfds = select (20, &readfds, &writefds, (int*)0, (int*)0);

the new line ...
  nfds = select (20, &readfds, &writefds, (int*)100000);

to keep emacs from doing a poll instead of blocking (I assume the
missing argument in the new line is an oversight).  The key point is
that this reveals a failure to understand the following critical line
in the manual section for select(2):

"If timeout is a zero pointer, the select blocks indefinitely.  To
affect [sic] a poll, the timeout argument should be non-zero, pointing
to a zero valued timeval structure."

A zero value passed to select does NOT indicate a poll, and a value of
100000 is meaningless--*((int *)100000) is definitely NOT 100000.

Joe Pallas
joe@cvl.uucp	(seismo!rlgvax!cvl!joe)
joe@cvl.arpa