Date:	Sun, 7 Jan 2001 12:23:14 +0200 (IST)
From:	Eli Zaretskii <eliz AT is DOT elta DOT co DOT il>
X-Sender:	eliz AT is
To:	Tim Van Holder <tim DOT van DOT holder AT pandora DOT be>
cc:	djgpp-workers AT delorie DOT com
Subject:	RE: Backslashes in debug info
In-Reply-To:	<NEBBIOJNGMKPNOBKHCGHOEGJCAAA.tim.van.holder@pandora.be>
Message-ID:	<Pine.SUN.3.91.1010107121107.21289U-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 Sun, 7 Jan 2001, Tim Van Holder wrote:

> I'm pretty sure gdb currently handles this correctly

Don't be so sure: the problems are very subtle, and test cases that 
expose them are very hard to cook.

GDB will DTRT when presented with file names in DOS format,
including with backslashes; that was fixed a long time ago (last
problems were corrected when GDB 4.18 was ported).

However, GDB modules such as symtab.c and minsyms.c have quite a few 
literal '/'s in them, and some of these places are relevant to debug 
formats supported by DJGPP, including stabs and dwarf2.  That code
makes all kinds of decisions based on the file name (as opposed to just
passing the file name to some file-oriented library function).  That
code was never made to understand backslashes because I always thought
(and continue to think) that the debug info cannot cope with backslashes 
well, and that they are all replaced with forward slashes.  (And with 
COFF debugging this is a moot point, since the full path of a file is not 
recorded at all.)

I'm sure that if you succeed to fall on one of the fragments where GDB 
uses '/', you will see all kinds of breakage.

- Raw text -