Relay-Version: version B 2.10 5/3/83; site utzoo.UUCP
Path: utzoo!mnetor!uunet!husc6!cmcl2!rutgers!clyde!watmath!watdragon!jmsellens
From: jmsellens@watdragon.waterloo.edu (John M. Sellens)
Newsgroups: comp.windows.x
Subject: Re: X V10 XOpenDisplay - should it assume hostname:0??
Message-ID: <3688@watdragon.waterloo.edu>
Date: Tue, 6-Oct-87 00:01:16 EDT
Article-I.D.: watdrago.3688
Posted: Tue Oct  6 00:01:16 1987
Date-Received: Fri, 9-Oct-87 05:14:49 EDT
Reply-To: jmsellens@watdragon.waterloo.edu (John M. Sellens)
Organization: U. of Waterloo, Ontario
Lines: 99

I (John Sellens) originally asked:
>	Currently, it seems that the X V10.4 XOpenDisplay requires
>	either an argument or an environment variable, and if both of
>	of them are missing, or if a display number is not given,
>	then XOpenDisplay will fail.  My question is: isn't it
>	reasonable for it to assume the current host and display 0
>	if either or both are missing?
>	

Jim Gettys answered:
>	Not really; there are people who run more than server from a single
>	machine.  This was true for VS100's, and is true today for some
>	VAXstation II/GPX machines (dual headed) ...
>	
>	It is more than slightly irritating if other people's mistakes pop
>	up windows on your display. ... It seemed best to return an error
>	rather than blindly continue.

and Paul Raveling added:
>    In my opinion this is eminently reasonable.  I'd rather have Xlib
>    take a reasonable default than to have to code the defaulting into
>    each client we write.
>
>    We haven't, to my knowledge, had that sort of threat yet in using
>    X in our corner of ISI.  In our environment the benefits of this
>    defaulting would outweigh the risks.
>
>    BTW, I'm using a two-headed Bobcat; others on our project have
>    private single-headed Bobcats at this time.


So - I figured what the heck, and added it to our copy of Xlib anyway.
It seems to work, so I thought that I would throw it out into the net
in case anyone else was interested.  Just a couple of trivial changes
to Xlib/XDisplayName.c, the revised version of which is included here
in all it's splendor - the changes are marked with #ifdef waterloo.

John

John Sellens -- Logic Programming and Artificial Intelligence Group

{decvax|utzoo|ihnp4|allegra|clyde}!watmath!watdragon!jmsellens
jmsellens@dragon.waterloo.{edu,CDN} jsellens@watmta.BITNET
jmsellens%dragon@waterloo.csnet



/* $Header: XDisplayName.c,v 10.1 86/11/19 18:17:19 jg Rel $ */

/* XDisplayName.c */
/* 
 * Returns the name of the display XOpenDisplay would use.  This is better
 * than just printing the "display" variable in a program because that
 * could be NULL and/or there could be an environment variable set.
 * This makes it easier for programmers to provide meaningful error
 * messages. 
 *
 * 
 * For example, this is used in XOpenDisplay() as
 *	strncpy( displaybuf, XDisplayName( display ), sizeof(displaybuf) );
 *      if ( *displaybuf == '\0' ) return( NULL );
 *  This check is actually unnecessary because the next thing is an index()
 *  call looking for a ':' which will fail and we'll return(NULL).
 */
/* Written at Waterloo - JMSellens */

#include <stdio.h>
#ifdef waterloo
#include <sys/param.h>	/* for MAXHOSTNAMELEN */
#endif

char *getenv();


char *
XDisplayName( display )
char *display;
{
#ifdef waterloo
    /* has to be static since we return it */
    static char hname[MAXHOSTNAMELEN+5];	/* xtra space for :0 */
#endif
    char *d;
    if ( display != (char *)NULL && *display != '\0' )
	return( display );
    if ( (d = getenv( "DISPLAY" )) != (char *)NULL )
	return( d );
#ifdef waterloo
    /* This causes XOpenDisplay to assume the current host for display if
       no other guess is available - Jim Gettys said it would be a
       bad thing and lead to more people popping windows on your
       terminal by accident - JMS Oct 5/87 */
    if ( gethostname(hname,sizeof(hname))==0 ) {
	(void) strcat( hname, ":0" );
	return( hname );
    }
#endif
    return( "" );
}