*BSD News Article 12378


Return to BSD News archive

Newsgroups: comp.os.386bsd.questions
Path: sserve!manuel.anu.edu.au!munnari.oz.au!news.Hawaii.Edu!ames!pacbell.com!att-out!cbnewsj!dwex
From: dwex@cbnewsj.cb.att.com (david.e.wexelblat)
Subject: Re: XFree-1.2 crashes w/ xv weather map
Organization: AT&T
Date: Fri, 5 Mar 1993 17:04:39 GMT
Message-ID: <1993Mar5.170439.5847@cbnewsj.cb.att.com>
References: <1mc917INN6db@usenet.INS.CWRU.Edu> <1993Feb25.214202.11624@dsuvax.dsu.edu> 
Lines: 45

>In <1mc917INN6db@usenet.INS.CWRU.Edu> david@jake.EEAP.CWRU.Edu (David Nerenberg) writes:
>
>>I was trying to remote display a weather map on my machine and it did
>>not like it at all.  Bacially, I run xv from another machine authorized
>>to display on my screen and X crashes locally, all the way out.  At 
>>least it does re-set the vidio mode back to text!
>
>I've done some more investigating into this problem and now I'm thoroughly
>puzzled.  I get the same crash from the XFree86 1.2 color server on my 8MB
>40Mhz AMD 80386 at home -- it appears the server gets a signal 4 when xv
>is starting up (although I've only been running xv locally).  I tried the
>running xv under XFree86's server on a 16MB 25Mhz intel 80386 w/ 80387,
>and it works fine!  The only differences that should count between the two
>machines is the amount of RAM and the 80387 chip that the working machine
>has; both were using the same ET4000 card and have almost the same amount
>of swap.
>
>I haven't been able to use gdb (either the one supplied with 386BSD or the
>recent 4.8 release) on the XFree86 server to get a stack backtrace from
>the coredump -- gdb freaks out with an internal error or complains that it
>can't access some odd memory location.
>
>I just had an idea -- could the 80387 emulator have caused the illegal
>instruction trap and killed the X server?
>-- 
>Guy Helmer, Dakota State University Computing Services - ghelmer@dsuvax.dsu.edu
>

I believe that the 387 emulater is in fact the cause.  Rich Murphey is
pursuing this angle, I believe.  Here's the reason I reach this conclusion:

	1) The problem seems to only be occuring on 386BSD
	2) The problem seems to only be on 387-less machines
	3) The problem started occurring with XFree86 1.2
	4) GCC 2.3 introduced some changes in 387 handling
	5) XFree86 1.2 was build with GCC 2.3.3, and XFree86 1.1 was not

I may be wrong, but the evidence points to the FP emulator as the problem.

--
David Wexelblat <dwex@mtgzfs3.att.com>  (908) 957-5871
AT&T Bell Laboratories, 200 Laurel Ave - 3F-428, Middletown, NJ  07748

"In and around the lake, mountains come out of the sky.  They stand there."
	-- Yes, Roundabout