X-Authentication-Warning:	acp3bf.physik.rwth-aachen.de: broeker owned process doing -bs
Date:	Tue, 31 Jul 2001 11:08:27 +0200 (MET DST)
From:	Hans-Bernhard Broeker <broeker AT physik DOT rwth-aachen DOT de>
X-Sender:	broeker AT acp3bf
To:	djgpp-workers AT delorie DOT com
Subject:	Re: Bug in COFF debug info
In-Reply-To:	<200107280854.EAA31809@delorie.com>
Message-ID:	<Pine.LNX.4.10.10107311059390.13133-100000@acp3bf>
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 Sat, 28 Jul 2001, Eli Zaretskii wrote:

>   (gdb) break main
>   Breakpoint 1 at 0x15a6: file fpreg.c, line 984
> 
> Huh?  What line 984?  This file has only 12 lines.

And main() is not in fpreg.c, but rather in ftest.c, too...

> On this machine, I have GCC 3.0 and Binutils 2.11.  I vaguely recall
> something about problems with COFF debug info either in GCC or in
> Binutils, but I don't recall the details, and cannot test other
> combinations of the compiler and Binutils on this machine.

I guess the problems you recall would be those that were introduced when
GCC switched to emitting of function prologues (stack frame setup,
treatment of profiling and such) as RTL instead of raw assembly. Since
about then, debugging information for the very first active source code
line of a function has been flaky at best. If this problem strikes, the
debug info for this line points to the last line of the previous function
in the object file, instead.

To see whether that's indeed the problem, one would compile with the
undocumented(!) option -mno-schedule-prologue.

-- 
Hans-Bernhard Broeker (broeker AT physik DOT rwth-aachen DOT de)
Even if all the snow were burnt, ashes would remain.

- Raw text -