delorie.com/archives/browse.cgi   search  
Mail Archives: djgpp-workers/2001/07/30/04:22:44

From: "Nimrod A. Abing" <abing AT redberger DOT com>
Organization: Kulitpuro Heavy Industries - Ltd. Roxas City Branch
To: Eli Zaretskii <eliz AT is DOT elta DOT co DOT il>
Subject: Re: DJGPP core dumper
Date: Mon, 30 Jul 2001 16:17:34 +0800
X-Mailer: KMail [version 1.2]
References: <3 DOT 0 DOT 1 DOT 32 DOT 20010521101259 DOT 0071c0d0 AT wingate> <200107281642 DOT MAA23764 AT delorie DOT com>
In-Reply-To: <200107281642.MAA23764@delorie.com>
Cc: djgpp-workers AT delorie DOT com
MIME-Version: 1.0
Message-Id: <01073016173401.00846@freeman>
Reply-To: djgpp-workers AT delorie DOT com

Noong Sunday 29 July 2001 12:42 am, sinulat ni Eli Zaretskii ang mga 
sumusunod:
> First, you don't dump FP registers, which we need to debug
> floating-point programs.

Okay.
 
> Next, the names of the sections that you've chosen seem to be
> different from what other platforms use in ELF core files; see
> bfd/elf.c in the Binutils or GDB distribution.  (I'd like to be as
> compatible with other platforms as possible, to minimize
> DJGPP-specific code in both Binutils and GDB's core file support.)
> For example, the section with registers is called ".notes", not
> "regs", the sections with memory blocks are called ".mem", etc.  Is
> there any reason for the particular choice of names in coredump.c?

I would like to add something to the readme.txt file included with the 
package. It's been a long time since I've taken a look at this myself so all 
of this is "off the top of my head". IIRC the core file layout I made is 
divided into the Elf Headers and the Elf Segments. There can be at the very 
least 2 Elf segments; one is the notes segment, the other is the memory 
segments containing blocks of memory pointed to by the handles in the 
__djgpp_memory_handle_list array.

The notes segment is subdivided into sections, since the notes segment is 
essentially a general purpose segment that can contain anything. I had the 
sections named as they were in George's old code. This could easily be 
changed in the next release that I'll be posting in a few days. As for the 
names of the sections, they are like that so that they can fit into a 32-byte 
integer. And because there were no standard specs for the core file format 
and its layout, I followed George's old layout.

> Finally, I'm not sure what to do with the environment and backtrace
> sections: GDB doesn't use them.  Given that the stack is dumped as
> well, do we really need the backtrace?  As for the environment, it's
> part of the normal address space, so why should we dump it once more,
> into a separate section?

Er, it was just for convenience. You're right though Eli, and I will remove 
them in the next release.

-- 
_nimrod_a_abing_


`/usr/games/fortune -s` says:
When the candles are out all women are fair.
		-- Plutarch

- Raw text -


  webmaster     delorie software   privacy  
  Copyright © 2019   by DJ Delorie     Updated Jul 2019