Re: [9fans] isatty

[presotto@closedmind.org]

[-- Attachment #1: Type: text/plain, Size: 588 bytes --]

The version comparison is an overkill and potentially wrong since
whatever is implementing /dev/cons may decide to update the
version each write, as is our practice in other places.

That should work in general as long as the namespace hasn't
been changed twixt the opening of the fd and the dirstat.
Otherwise /dev/cons may be under a different mount point
number and hence a different c.dev.

The question isatty is usually trying to answer is
"does the fd has an interacting user behind it?".  A
hard question to answer in a world where the geography
can shift so easily.

[-- Attachment #2: Type: message/rfc822, Size: 2165 bytes --]

From: Boyd Roberts <boyd@strakt.com>
To: "9fans@cse.psu.edu" <9fans@cse.psu.edu>
Subject: [9fans] isatty
Date: Wed, 23 Jan 2002 12:56:44 +0100
Message-ID: <3C4EA4FC.A777CA4D@strakt.com>

Given I had to implement this 'thing' on Plan 9 I decided on:

#include <u.h>
#include <libc.h>

static char *cons = "/dev/cons";

/*
 * Compare fd to cons should get it right.
 */
int
isatty(int fd)
{
	Dir c;
	Dir f;

	if (dirstat(cons, &c) != 0)
		return 0;

	if (dirfstat(fd, &f) != 0)
		return 0;

	return f.type == c.type &&
		f.dev == c.dev &&
		f.qid.path == c.qid.path &&
		f.qid.vers == c.qid.vers;
}

Seemed obvious to me, but is this a good plan?

The overkill on the multiple dirstat's on /dev/cons and
qid comparison I thought would ensure that I wouldn't
get fooled [again].