Date:	Wed, 3 Jan 2001 11:03:50 +0200 (IST)
From:	Eli Zaretskii <eliz AT is DOT elta DOT co DOT il>
X-Sender:	eliz AT is
To:	DJ Delorie <dj AT delorie DOT com>
cc:	djgpp-workers AT delorie DOT com
Subject:	Re: config/mh-djgpp in Cygnus tree
In-Reply-To:	<200101022142.QAA11774@envy.delorie.com>
Message-ID:	<Pine.SUN.3.91.1010103110328.1129S-100000@is>
MIME-Version:	1.0
Reply-To:	djgpp-workers AT delorie DOT com
Errors-To:	nobody AT delorie DOT com
X-Mailing-List:	djgpp-workers AT delorie DOT com
X-Unsubscribes-To:	listserv AT delorie DOT com

On Tue, 2 Jan 2001, DJ Delorie wrote:

> > It may *look* like one, but it may just as well be an essential
> > piece to the DJGPP port of the program that just didn't get
> > commented understandibly. IIRC, you can not build GCC with -g
> > active.  At least at the time of DJGPP 1.x, when DJ says this change
> > was conceived, that was the case. You would hit the hard limit of
> > 64K symbols in a COFF object file, otherwise. A debuggable GCC needs
> > '-gstabs', instead.
> 
> Doh!  I knew there was something about that...
> 
> Well, I took it out.  If anyone wants to submit a version with a
> better comment, I'll humbly put it back in.

I think the gravity of not using -g is much less these days, since
Binutils 2.8 and later don't treat the line info overflow as a fatal
problem.  So I'd expect that a GCC build with -g _is_ possible.

Note that GDB is compiled with -g by default, even though mh-djgpp
says otherwise.  I get line number overflow in GDB as well, but the
executable runs and is debuggable.

- Raw text -