Path: utzoo!news-server.csri.toronto.edu!cs.utexas.edu!sdd.hp.com!usc!apple!agate!usenet.ins.cwru.edu!eagle!snuffy.lerc.nasa.gov!mckim
From: mckim@snuffy.lerc.nasa.gov (Jim McKim)
Newsgroups: comp.unix.sysv386
Subject: Roell X11r4 Intel 3.2/2.1 fix
Message-ID: <1991Mar5.151157.23574@eagle.lerc.nasa.gov>
Date: 5 Mar 91 15:11:57 GMT
Sender: news@eagle.lerc.nasa.gov
Reply-To: mckim@snuffy.lerc.nasa.gov (Jim McKim)
Organization: NASA/Lewis Research Center, Cleveland
Lines: 34

I am compiling Mr. Roell's X11R4 for Intel 3.2/2.1 with Lachman TCP.
I am using a Swan ET4000 type graphics card.  I had a problem which I
think some other people here mentioned they had also.

The symptoms are: xinit causes the screen to flicker briefly.  Then
you are dumped out with a "0 events processed..." message.  -debug
shows ET4000Save as the last routine entered.  X run by itself shows
a memory fault in that routine.

The problem on my system was: my unix wasn't letting X write to one of
the bona fide SVGA ports despite all the finagling that is done in
vgaScreenInit().  I think my version of unix accepts only the EGA
range of I/O addresses.  My fix was to include the line:

	ioctl(consoleFd, KDADDIO, (unsigned short)0x3bf);

somewhere prior to attempting I/O on that port [ET4000Save(), line
163].  I put it in vgaScreenInit() around line 232.

So far this is the only address I have had to do this for.  I don't
have any documentation for VGAs or SVGAs - there might be other
candidates for this treatment.

A quick fix (I haven't tried this - I will tonight) would be to make a
program consisting solely of this ioctl call.  You might have to run
it setuid root.  If the setting of this "attribute" sticks around then
the original binary should work afterwords.

- Jim
--
----------------
Jim McKim  /  Internet: mckim@snuffy.lerc.nasa.gov		k-crash
Phone: +1 216 891 2982  /  Packet: kb8dcr@kb8dcr.ampr.org	is coming!
----------------