From: Ivo Janssen <ivo AT ricardis DOT tudelft DOT nl>
Newsgroups: comp.os.msdos.djgpp
Subject: Re: SIGSEGV error meanings??
Date: Fri, 12 Sep 1997 10:11:45 +0200
Organization: TNO-FEL
Lines: 36
Message-ID: <3418F941.CCB821B8@ricardis.tudelft.nl>
References: <Pine DOT SUN DOT 3 DOT 91 DOT 970910192855 DOT 10908H-100000 AT is>
NNTP-Posting-Host: dewey.fel.tno.nl
Mime-Version: 1.0
Content-Type: text/plain; charset=us-ascii
Content-Transfer-Encoding: 7bit
To: djgpp AT delorie DOT com
DJ-Gateway: from newsgroup comp.os.msdos.djgpp
Precedence: bulk

Eli Zaretskii wrote:
> > Exiting due to signal SIGSEGV
> > Page fault at eip=000056ba, error=0004
> > eax=69696b6a ebx=0002a7e4 ecx=00b68004 edx=00000000 esi=69696b6a
> 
> This is a null pointer dereference, if you ask me.  For Page Faults,
> error=0004 means you are writing into an address that's not present, and
> EDX is zero, which is probably the pointer.
> 
> Can it be that your friend uses CWSDPMI while you use something else,
> like Windows?

When I compile a program, in a dosbox in win95, using win95's DPMI, and
then run it in the same box, can I expect SIGSEGV's due to the DPMI
host? I have this program (ld.exe from binutils) which compiles
correctly on a unix-box, but gives SIGSEGV in dos. I gdb'd ld.exe and it
seems that dos filles a pointer with some random address, while all
unix's fill it with (the right value) NULL, so in the next line, where
it says: if (!pointer) break; else do_some_pointer_stuff; unix does the
right thing, but dos causes a sigsegv due to a bad address.

Would changing DPMI server help?

Ivo
-- 
--------------------+--------------------------------+-+     +---+
| IVO JANSSEN       | ivo AT ricardis DOT tudelft DOT nl        | |     |   |
| Ricardishof 1/428 | janssen AT cardit DOT et DOT tudelft DOT nl   | |   |-+   |
| 2614 JE Delft     | http://www.et.tudelft.nl/~ivo/ | +->-|    (|)
| the Netherlands   | Dept of Electrical Engineering |     |-+   |
| +31-15-2855973    | Delft University of Technology |       |   |
+-------------------+--------------------------------+      --- ---

  Do you want to win $1000 by using you computer's idle time AND
  prove that 56-bit encryption isn't _that_ safe?
  Try http://rc5.distributed.net/