delorie.com/archives/browse.cgi   search  
Mail Archives: djgpp/2000/07/08/02:52:07

Message-ID: <3966CFBC.3686882B@austin.rr.com>
Date: Sat, 08 Jul 2000 01:52:44 -0500
From: Brogdon <brogdon AT austin DOT rr DOT com>
X-Mailer: Mozilla 4.05 [en] (Win95; U)
MIME-Version: 1.0
To: Eli Zaretskii <eliz AT is DOT elta DOT co DOT il>
CC: djgpp AT delorie DOT com
Subject: Re: ANNOUNCE: GDB stub for remote debugging of DJGPP targets
References: <39669600 DOT 896FFBB6 AT austin DOT rr DOT com> <200007080630 DOT JAA17055 AT mailgw1 DOT netvision DOT net DOT il>
Reply-To: djgpp AT delorie DOT com

Eli:


> Hmm...  Browsing through the sources, it looks like the implementation
> doesn't use the DJGPP debug support routines (in libdbg.a from
> djdev203.zip) at all, and instead does everything on its own.  Is that
> true?
>

I didn't use the libdbg.a library, mainly because I didn't have any
information to understand how this code worked (other than digging through
it).  Perhaps, this information exists, but I didn't know where to find it.

> If so, perhaps it would be better either to use libdbg.a functions, at
> least those in dbgcom.c which load the debuggee and allow it to run,
> or borrow some of the code from there.  The code in dbgcom.c has
> several useful features that might make sense in the stub case, such
> as true exception handling, hardware breakpoints/watchpoints support,
> the ability to restart the debuggee without exiting the debugging
> session, support for FP registers, etc.
>

I would be interested in understanding how this works.  Point me to more
information, and I will look at converting to this.

> Also, I see that the debugged program needs to be linked with
> libgdb.a.  Why is this required?  libgdb.a is quite a large library,
> so this would make for a very large executable, I think, whereas
> remote debugging is supposed to keep the target part lean and mean...

The libgdb.a really only contains 2 objects:  i386-stub.o (the stub
functionality called out by the GDB manual) and i386-supp.o (the support
functions which the user must provide for the stub).  That's it.  Creating
the library was my idea.  I don't believe that the GDB distrubution will
cause a library to be built for this.  This may not be the most efficient
code (actually, I can think of at least one thing that I could have done to
cut down on register copies and updates--always room for improvement), but
it is not a lot of code.

Regards,
Jon

- Raw text -


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