Xref: utzoo comp.windows.x:2669 comp.bugs.misc:90
Path: utzoo!mnetor!uunet!husc6!tut.cis.ohio-state.edu!bloom-beacon!gatech!mcnc!decvax!purdue!spaf
From: spaf@cs.purdue.EDU (Gene Spafford)
Newsgroups: comp.windows.x,comp.bugs.misc
Subject: Re: X11 & ksh? (+Workaround)
Message-ID: <3465@medusa.cs.purdue.edu>
Date: 8 Mar 88 19:19:06 GMT
References: <3429@medusa.cs.purdue.edu>
Sender: news@cs.purdue.EDU
Organization: Department of Computer Science, Purdue University
Lines: 40
Summary: Problem found & partial fix enclosed.


A few days ago, I posted a request for help.  The problem I had was
that if I ran "xinit" under the ksh, the server just hung and never
finished starting up.  I narrowed the problem down further to find that
only if I had the monitor (jobs facility) option set did the problem
occur.

Well, after an afternoon and evening of intense frustration debugging,
I found the problem.  The following is a description of the problem and
a workaround; the bug and some approaches to fixes have been submitted
to the official buglist, but this may help others soon to encounter
similar problems.

When "xinit" starts the server, it resets the process group of the
process running the X server (Xsun in my case) to 0 so it is in a
different process group than the xinit process.   This is to prevent a
problem that will not occur on a Sun -- that of the companion client
process being spawned with the -L flag, and having a subsequent
vhangup() call blow the parent processes away.  Unfortunately, one of
the first things done by the new X server process is a
"write(2, buffer, 0)" to see if the fd is open for error messages.
Since the process no longer belongs to the distinguished process group
of the terminal where unit 2 is open, it gets a SIGTTOU and blocks.

Workaround is simple:  run xinit with stderr redirected to a file or
pipe.  I now run ksh out of my .profile with:
(xinit $HOME/bin/xstart; kbd_mode -a; clear) 2>&1 | cat -u
Ugly maybe, but it works just fine.

Mark Moraes @ University of Toronto sent me a patch he developed to get
newcsh and tcsh running under X11.1.  He suggests I "#ifndef sun" the
setpgrp causing the problem.  Unfortunately, the killpg mechanism in
use causes the parent ksh to die when xinit terminates if this change
is applied.  Hopefully the people supporting X11 will come up with a
better, "official" fix.

-- 
Gene Spafford
Dept. of Computer Sciences, Purdue University, W. Lafayette IN 47907-2004
Internet:  spaf@cs.purdue.edu	uucp:	...!{decwrl,gatech,ucbvax}!purdue!spaf