X-Authentication-Warning: delorie.com: mailnull set sender to djgpp-workers-bounces using -f
Date: Sun, 3 Mar 2002 13:31:13 +0200 (IST)
From: Eli Zaretskii <eliz AT is DOT elta DOT co DOT il>
X-Sender: eliz AT is
To: djgpp-workers AT delorie DOT com
Subject: Re: libm signed/unsigned warnings
In-Reply-To: <200203022210.XAA05451@father.ludd.luth.se>
Message-ID: <Pine.SUN.3.91.1020303133052.705E-100000@is>
MIME-Version: 1.0
Content-Type: TEXT/PLAIN; charset=US-ASCII
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 Sat, 2 Mar 2002, Martin Str|mberg wrote:

> > > -                   if((j<<(52-k))==ly) yisint = 2-(j&1);
> > > +                   if((((__uint32_t)(j))<<(52-k))==ly) yisint = 2-(j&1);
> > 
> > This seems okay, but is mighty ugly.  Can you find a better way of
> > shutting up GCC in this case?
> 
> Perhaps. But I'd really like somebody who understands the code to make
> more extensive changes.

What additional changes would you like to suggest?  The code is not
hard to understand, I think.

> > > -               if(((t1&sign)==sign)&&(s1&sign)==0) s0 += 1;
> > > +               if((t1&sign)&&(s1&sign)==0) s0 += 1;
> > >                 ix0 -= t;
> > >                 if (ix1 < t1) ix0 -= 1;
> > >                 ix1 -= t1;
> > > 
> > > __int32_t sign = (int)0x80000000U;
> > > __uint32_t t1,s1;
> > > 
> > > (t1&sign)==sign checks if the sign bit is set or not. The comparision
> > > with sign should be uneccessary.
> > 
> > I don't understand the original problem here: both t1 and sign are
> > int's, so why does GCC complain?  What is the exact language of the
> > warning?
> 
> No. t1 is unsigned.

Oops, you are right.

> > > -               if(ix>0x80000000U) z = (invsqrtpi*cc)/__ieee754_sqrtf(x);
> > > +               if((__uint32_t)ix>0x80000000U) z = (invsqrtpi*cc)/__ieee754_sqrtf(x);
> > 
> > I tend to think that this is a bug in the original code, since ix is
> > computed a few lines above this like so:
> > 
> > 	ix = hx&0x7fffffff;
> > 
> > So its high bit can never be set, and the comparison should always
> > fail.  Am I missing something?
> 
> I'm missing which file we're talking about...

It's from ef_j1.c.