From: pavenis AT lanet DOT lv
Message-ID: <B0000101585@stargate.astr.lu.lv>
To: Eli Zaretskii <eliz AT is DOT elta DOT co DOT il>, djgpp-workers AT delorie DOT com
Date: Tue, 14 Sep 1999 20:04:56 +0300
MIME-Version: 1.0
Content-type: text/plain; charset=US-ASCII
Content-transfer-encoding: 7BIT
Subject: RE: gcc-2.95
References: <B0000101501 AT stargate DOT astr DOT lu DOT lv>
In-reply-to: <Pine.SUN.3.91.990914131203.18753A-100000@is>
X-mailer: Pegasus Mail for Win32 (v3.12a)
Reply-To: djgpp-workers AT delorie DOT com

On 14 Sep 99, at 13:16, Eli Zaretskii wrote:

> 
> On Tue, 14 Sep 1999 pavenis AT lanet DOT lv wrote:
> 
> > These things are rather broken. Some days ago related fixes were done
> > in gcc source tree (current only, not in gcc-2.95 branch). So perhaps 
> > now we have too possibilities:
> > 	1) simply build for i386-pc-msdosdjgpp and leave this topic for
> >             next version
> >        2) backport related changes to gcc-2.95.1 (currently patches fails) 
> > 
> > In current version options -m386, -m486, ... also are rather broken (try 
> > generating assembler source with -m386, -m486, ... and with -
> > mcpu=i386, -mcpu=i486, ... and compare results)
> 
> Based on your description, it seems that the best practical alternative 
> is indeed to build with i386 as the target, and leave the rest for the 
> future.  This would at least solve a known problem.  If -mx86 options are 
> broken, any other solution risks uncovering additional bugs which might 
> break user programs.
> 

I took related changes from current CVS versions of gcc and apllied 
them manually as sources are too different. Also fixed one more 
ancient bug that is present at least from gcc-2.8.0  time(it seems to 
effectively kill all -m386, -m486, etc options, but -mcpu=... works). 
These changes were for gcc/config/i386/i386.h

Also modified i386.c to always default to -march=i386. Similar change 
was some days ago done in mainline sources. Earlier value was always 
to default to -march=pentium (even when I'm building compiler  for 
i386-pc-msdosdjgpp)

So I'll leave building gcc-2.95.1 later in evening (when I'll be about to 
leave). Perhaps some additional testing will be needed after such 
changes, However it's rather unlikely I have broken something badly.

If all will work we'll have defaults options -march=i386 -mcpu=pentium
as suggested earlier in this mailing list.

Andris