Message-ID: <34CA3F71.CD6AD0B1@gmx.net>
Date: Sat, 24 Jan 1998 20:22:25 +0100
From: Robert Hoehne <robert DOT hoehne AT gmx DOT net>
Organization: none provided
MIME-Version: 1.0
To: andrewc AT rosemail DOT rose DOT hp DOT com, djgpp-workers AT delorie DOT com
Subject: Re: -g problem
References: <199801240047 DOT AA252462849 AT typhoon DOT rose DOT hp DOT com>
Content-Type: text/plain; charset=us-ascii
Content-Transfer-Encoding: 7bit
Precedence: bulk

Because Andrew posted this also to the workers, I do cc it
too to the workers, just for information.

Andrew Crabtree wrote :
> 
> Perhaps I worded it wrong, but I think this is still in the gcc sources.
> I do use the egcs sources now, but I have had this problem for almost
> 10 months, before egcs existed.  Just copy the 6 lines of code below
> to a file, and try compiling with -g, -gstabs, and no -g.  This
> code should be in the file libgcc1-test.c at the very bottom, but the
> problem will manifest itself if you just make a foo.c with these lines ...
> 
> extern void start() __asm__("start");
> extern void _start() __asm__("_start");
> extern void __start() __asm__("__start");
> 
> void start() {}
> void _start() {}
> void __start() {}
> 
> I have tried to debug it and sent message to dj-workers but it is
> beyond my knowledge.  For some reason the assembler errors out with
> -g, even though no additional symbols are defined, just debug stuff.

I checked it now (got also the mail from the workers list) and found the
following: This problem is to 99.99% a problem of the gcc you are
using. I came to this after checking your posted assembler output.

Comparing it with my generated output (from gcc 2.7.2.1 and gcc 2.8.0)
there is (beside other diffs) one important diff. You have for instance
between the two functions start and _start

	.def	.ef;	.val	.;	.scl	101;	.line	1;	.endef
	.def	_start;	.val	.;	.scl	-1;	.endef
	.def	__start;	.val	_start;	.scl	2;	.type	041;	.endef


but I have there

	.def	.ef;	.val	.;	.scl	101;	.line	1;	.endef
	.def	_start;	.val	.;	.scl	-1;	.endef
	.p2align 2
	.def	__start;	.val	_start;	.scl	2;	.type	041;	.endef

(with gcc 2.7.2.1 there is a .align 2 instead of .p2align).

That means now, you have modified gcc in a way, that it produces wrong
(at least
for gas wrong) code. You can simply check it, by inserting the align
directives
and it will compile OK.

BTW: I don't know, why gas complains this, but it _IS_ so and I haven't
the
time to debug this problem since it does not happen with a normal gcc
:-)

Robert
-- 
******************************************************
* email:   Robert Hoehne <robert DOT hoehne AT gmx DOT net>     *
* Post:    Am Berg 3, D-09573 Dittmannsdorf, Germany *
* WWW:     http://www.tu-chemnitz.de/~sho/rho        *
******************************************************