From: pavenis AT lanet DOT lv
To: "Eli Zaretskii" <eliz AT is DOT elta DOT co DOT il>, sandmann AT clio DOT rice DOT edu,
        pavenis AT lanet DOT lv, djgpp-workers AT delorie DOT com,
        ST001906 AT HRZ1 DOT HRZ DOT TU-Darmstadt DOT De
Date: Mon, 10 Sep 2001 11:35:39 +0300
MIME-Version: 1.0
Subject: Re: gcc-301 difficulty
Message-ID: <3B9CA58B.21353.20AB69@localhost>
In-reply-to: <7704-Sat08Sep2001211808+0300-eliz@is.elta.co.il>
References: <69117341E6B AT HRZ1 DOT hrz DOT tu-darmstadt DOT de>	(ST001906 AT HRZ1 DOT HRZ DOT TU-Darmstadt DOT De)
X-mailer: Pegasus Mail for Win32 (v4.0, beta 40)
Content-type: text/plain; charset=US-ASCII
Content-transfer-encoding: 7BIT
Content-description: Mail message body
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
Precedence: bulk

On 8 Sep 2001 at 21:18, Eli Zaretskii wrote:

> > From: "Juan Manuel Guerrero" <ST001906 AT HRZ1 DOT HRZ DOT TU-Darmstadt DOT De>
> > Organization: Darmstadt University of Technology
> > Date: Sat, 8 Sep 2001 19:29:49 +0200
> > 
> Andris, what significant changes were introduced into GCC between v3.0
> and 3.0.1?  (I mean not only DJGPP-specific changes.)

I'm afraid some details have been forgotten from initial posts (correct me Juan, if I'm wrong somewhere):

this crashes occur with one type of Cyrix CPU and motherboard when
CPU internal cache is enabled, so one can suspect problem with either
CPU or motherboard or their incompatibility. It is not the first time one is getting problems with GCC and 
some Cyrix CPU's (see
http://www.BitWizard.nl/sig11/ for details)

It didn't happen with gcc30b.zip from v2gnu/alphas

It didn't happen with one of my builds of gcc-3.0.1 where I errorously linked cc1 with current CVS libc 
(as I remeber because I have lost corresponding message from Juan) and reappeared when 
I linked cc1.exe with modified djdev203.

It didn't happen when Juan tried to build gcc-3.0.1 (C compiler only)
on plain DOS without LFN support (Juan, which libc version were 
used for that? with which GCC version was libc built?)

What it could be:

gcc301b.zip were linked with libc.a from a slightly modified djdev203.zip (only Win2k related fixes). I 
don't exactly know which exactly GCC (or EGCS) version was used to build most object files from libc 
(Perhaps Eli or DJ could remeber)

gcc30b.zip were linked with current CVS version of libc which were 
built with gcc-3.0

So I can guess (I'm not sure, of course) that some sequence of instructions in djdev203 triggers some 
problem with used Cyrix CPU.
It could be it does not happen when libc is built with gcc-3.0.X
(or if CVS version of runtime is being used)

What to do?

Perhaps it would be best to avoid this problem, by using a combination which doesn't trigger that bug. I'm 
afraid it could be too difficult to find exact reason of problem though.

Maybe I should build entire djdev203.zip with gcc-2.95.3 and see
whether it changes anything 

Andris