Xref: utzoo comp.unix.wizards:12218 comp.bugs.sys5:678
Path: utzoo!attcan!uunet!lll-winken!lll-tis!helios.ee.lbl.gov!pasteur!agate!labrea!rutgers!mailrus!tut.cis.ohio-state.edu!bloom-beacon!mit-eddie!uw-beaver!uw-june!ka
From: ka@june.cs.washington.edu (Kenneth Almquist)
Newsgroups: comp.unix.wizards,comp.bugs.sys5
Subject: Re: empty() for USG UNIX ?
Summary: No such thing
Message-ID: <6373@june.cs.washington.edu>
Date: 9 Nov 88 01:28:56 GMT
References: <292@rna.UUCP>
Organization: U of Washington, Computer Science, Seattle
Lines: 29

Dan Ts'o <dan@rna.rockefeller.edu> asks for a  System V routine to test
whether data is available to be read on a file descriptor referring to a
slow device (such as a tty).  I doubt that such a routine can be written.
Depending on what you are doing, you may be able to use FNDELAY mode to
get the same effect.  Code like:

	ioctl(fd, FIONREAD, (char *)&nchars);
	if (nchars) {
		count = read(fd, buf, sizeof(buf));
		Code to process input.
	} else {
		Do something else.
	}

becomes:

	fcntl(fd, F_SETFL, FNDELAY);
	if ((count = read(fd, buf, sizeof(buf))) != 0) {
		Code to process input.
	} else {
		Do something else.
	}

These two pieces of code are approximately equivalent, but the former
contains a race condition.  (The fact that characters are available to
be read when the FIONREAD ioctl is performed doesn't mean that these
characters will still be around by the time the code gets around to
performing the read.)  So the latter version of the code is preferable
on systems that support both.
				Kenneth Almquist