Message-Id: <3.0.16.19980212200915.358fdc8c@hem1.passagen.se>
Date: Thu, 12 Feb 1998 20:09:18 -0500
To: Eli Zaretskii <eliz AT is DOT elta DOT co DOT il>
From: Peter Palotas <blizzar AT hem1 DOT passagen DOT se>
Subject: Re: Bash crashing!
Cc: djgpp AT delorie DOT com
Mime-Version: 1.0
Content-Type: text/plain; charset="us-ascii"
Precedence: bulk

At 13.39 1998-02-12 +0200, you wrote:
>
>On Thu, 12 Feb 1998, Peter Palotas wrote:
>
>> Enough space IS available on the directory pointed to by TEMP (both in
>> environment and in DJGPP.ENV).
>
>TEMP or TMPDIR?  Bash uses the latter.  Exactly how much free space is on 
>that disk?  Is it RAM drive or a real disk?

Well, both of them are set and points to the same directory, which is not a
ramdrive, but a physical harddrive with about 900MB free (shouldn't that be
enough!? ;)

>> Okay, now from the GNU Hello 1.3 package (as downloaded from a the simtel
>> mirror at ftp.sunet.se today) when trying to run configure from within bash
>> in a DOS box from withing windows I get the following:
>> 
>> bash$ ./configure
>
>That's not how you are supposed to configure hello for DJGPP!  There is 
>config.bat in the djgpp/ directory, and a patched configure script in the 
>same directory that the batch file runs.  Please read the file README.dos 
>in the distribution and do what it tells.

Oh, I know that, but I didn't a while ago when I downloaded it, and then
bash crashed. You asked for a congfigure script where bash crashed, and I
mentioned one! (It has crashed on other configure scripts too, I tried with
the script provided with the GCC 2.8.0 sources too, and it crashed there as
well. (Not that I thought it would compile anyway ;))


>> Exiting due to signal SIGSEGV
>> Page fault at eip=0000a1ba, error=0004
>> eax=00181900 ebx=00181900 ecx=ffffffff edx=00000000 esi=00000000
edi=00000000
>> ebp=0016b6e8 esp=0016b6d4 program=D:\CODING\DJGPP\BIN/bash.exe
>
>This means a null pointer dereference.  Please insert "set -x" at the 
>beginning of the script and see what line does it execute when it 
>crashes.

Okay, I did this, and guess what... It didn't crash!


>> >      - does Bash crash every time you run that script, or only
>> >        sometimes?
>> I guess the answer to this must be NO, since it doesn't CRASH under windows
>> 3.11.
>
>But does it always crash on DOS with that script?  (The fact that it 
>doesn't crash on Windows is irrelevant: Windows doesn't catch null 
>pointer dereferences like CWSDPMI does.  However, the message printed by 
>Bash under Windows clearly tells that some function gets a null pointer, 
>so the same problem exists on both Windows and DOS.)

If it always crashes... Well, from what I've experienced, yes. (Btw, does
win 3.11 also provide a DPMI server? I thought it was only win32).


>> Yes, maybe the script won't work anyway, but I still don't think bash
>> should crash!
>
>It doesn't unless you give it a reason that is good enough ;-).

Hehehe... Yeah, guess I didn't do that. BUT... I discovered something...
You told me to insert set -x at the beginning, and I did so, and it didn't
crash! Then I removed set -x, and it crashed, then I inserted a blank line
before the first comment (#!/bin/sh), and it worked again.

Is there a reason that you can see for this, and should bash really crash
at anytime anyway!?


-- Peter Palotas alias Blizzar -- blizzar AT hem1 DOT passagen DOT se --

******************************************************
* A brief description of DJGPP:                      *
* NEVER BEFORE HAVE SO FEW DONE SO MUCH FOR SO MANY! *
******************************************************