Date: Thu, 16 Mar 2000 11:41:33 +0200 (IST)
From: Eli Zaretskii <eliz AT is DOT elta DOT co DOT il>
X-Sender: eliz AT is
To: Martin Stromberg <Martin DOT Stromberg AT lu DOT erisoft DOT se>
cc: djgpp-workers AT delorie DOT com
Subject: Re: Unnormals???
In-Reply-To: <200003151643.RAA20668@mars.lu.erisoft.se>
Message-ID: <Pine.SUN.3.91.1000316114116.3117B-100000@is>
MIME-Version: 1.0
Content-Type: TEXT/PLAIN; charset=US-ASCII
Reply-To: djgpp-workers AT delorie DOT com
Errors-To: dj-admin AT delorie DOT com
X-Mailing-List: djgpp-workers AT delorie DOT com
X-Unsubscribes-To: listserv AT delorie DOT com
Precedence: bulk


On Wed, 15 Mar 2000, Martin Stromberg wrote:

> > Perhaps you tried without the sign in the format specifier?  That case 
> > was left alone on purpose; see the discussions on djgpp-workers about 10 
> > months ago (IIRC).
> 
> But the sign of a negative nan and inf should be printed regardless of
> any sign format specifier.

Why ``should''?  I don't think the standard says that, since some
architectures don't support signed NaNs.

When this was discussed (I urge you to read those discussions), some
people said thay never want to see a sign, others said they want it
sometimes.  FWIW, the Cygnus test suite, now part of djtst203.zip,
clearly distinguishes between positive and negative NaNs and
infinities.  The compromise we've chosen IMHO makes sense: if the user
puts a sign into the format specifier, they are telling us they want
to see the sign.

> I don't think it is (from reading the
> source of <something...>special(), last in doprnt.c; not shown above).

I'm not sure I follow.  Are you telling that some cases where the sign
is in the format specifier are still printed without the sign?