X-Authentication-Warning: delorie.com: mail set sender to djgpp-bounces using -f
X-Recipient: djgpp AT delorie DOT com
Date: Sat, 02 May 2009 19:16:10 +0300
From: Eli Zaretskii <eliz AT gnu DOT org>
Subject: Re: Please help resurrecting GDB for version 7.0 release
In-reply-to: <_ZKdnV74euw2-2HUnZ2dnUVZ_uednZ2d@earthlink.com>
X-012-Sender: halo1 AT inter DOT net DOT il
To: djgpp AT delorie DOT com
Message-id: <83ws8zeb05.fsf@gnu.org>
References: <83zldveryw DOT fsf AT gnu DOT org> <684bdd55-b464-4e61-b28e-28774dc6ced8 AT o27g2000vbd DOT googlegroups DOT com> <83y6tfem02 DOT fsf AT gnu DOT org> <_ZKdnV74euw2-2HUnZ2dnUVZ_uednZ2d AT earthlink DOT com>
Reply-To: djgpp AT delorie DOT com
Errors-To: nobody AT delorie DOT com
X-Mailing-List: djgpp AT delorie DOT com
X-Unsubscribes-To: listserv AT delorie DOT com
Precedence: bulk

> From: "Charles Sandmann" <cwsdpmi AT earthlink DOT net>
> Date: Sat, 2 May 2009 10:28:29 -0500
> 
> >"Eli Zaretskii" <eliz AT gnu DOT org> wrote in message 
> >news:83y6tfem02 DOT fsf AT gnu DOT org...
> >> From: Rugxulo <rugxulo AT gmail DOT com>
> >> It also doesn't like "info all-regs", IIRC.
> >Yes, for the same reason: it thinks DJGPP supports SSE registers.
> 
> What is required to make that work?

There are 2 possible fixes: the easy one and the more complicated one.

The easy one is a two-liner: tell GDB that we don't support SSE.  I
already did that, and the patch is in the GDB CVS repository since
about 2 days ago.  (The original problem happened because GDB switched
to supporting SSE by default in its x86 targets at some point, so any
target that doesn't support them needs to explicitly disable that.
Seems like no one from the DJGPP camp, yours truly included, payed
attention at the time.)

The more complex solution is to extend save_npx and load_npx to use
FXSAVE and FXRSTOR instructions, when they are available, instead of
FNSAVE and FRSTOR that we use now.  But that means hacking the library
or providing GDB-private versions of these functions.  Unless many
DJGPP users need access to SSE registers (which sounds improbable,
given that this feature is badly broken for at least 5 last years), I
don't see a justification for this hassle.