X-Authentication-Warning:	acp3bf.physik.rwth-aachen.de: broeker owned process doing -bs
Date:	Thu, 27 Apr 2000 11:27:15 +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: gcc 3.0
In-Reply-To:	<200004262017.QAA24206@indy.delorie.com>
Message-ID:	<Pine.LNX.4.10.10004271121060.24606-100000@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 Wed, 26 Apr 2000, Eli Zaretskii wrote:

> > Date: Wed, 26 Apr 2000 11:20:40 -0400
> > From: DJ Delorie <dj AT delorie DOT com>
> > 
> > It would be nice if gcc 3.0 worked out-of-the-box with DJGPP.
> 
> Also, how about porting gcov?  Any takers?

When I worked on line-level profiling, I also tried out the
'-fprofile-arcs' method used by 'gcov'. The problem was that the resulting
programs almost immediately died, presumably because of stack corruption.
The modifications done to the function entry/exit code and basic block
entry broke DJGPP programs completely. If memory serves, there was a
problem with some register used by this profiling method that didn't
properly get conserved by gcc. As the problem routinely killed the
debugger along with the debuggee, I never managed to find out what the
actual core problem was.

In other words: porting 'gcov' itself is not the problem, I think. It's
the gcc-internal support code that needs to be adapted to allow this code
to work.

The last version I tried was 2.8.1, though, so there is a chance that it's
been improved, in the meantime. The new default '-mschedule-prologue'
method of outputting function prologues and epilogues may help, in this
area.

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

- Raw text -