ns32k fixes

[Andrew Pinski]

Your patch does not follow C alias rules, you should use an union.

Thanks,
Andrew Pinski


> 
> 
> --pG7rHtfCel
> Content-Type: text/plain; charset=us-ascii
> Content-Description: message body text
> Content-Transfer-Encoding: 7bit
> 
> Richard Henderson writes:
>  > On Mon, Jul 29, 2002 at 11:43:20PM +0930, Ian Dall wrote:
>  > > + # define ISNAN(x) (				\
>  > > +   {						\
>  > > +     unsigned int t = (x);			\
>  > 
>  > This isn't right.  You want the bit pattern for X, not
>  > the float->int conversion of X.
> 
> Well spotted. It turns out that there isn't a test case for this
> single precision case. I have created one and will send it in a
> following message.
> 
> Here is an additional patch (rather than a replacement one, is that OK?)
> 
> 2002-07-30  Ian Dall  <ian@sibyl.beware.dropbear.id.au>
> 
> 	* config/ns32k/__unordsf2.c (ISNAN): Avoid converting float to int.
> 
> 
> 
> --pG7rHtfCel
> Content-Type: text/x-patch
> Content-Disposition: inline;
> 	filename="gcc.diffs"
> Content-Transfer-Encoding: 7bit
> 
> Index: config/ns32k/__unordsf2.c
> ===================================================================
> RCS file: /usr/local/cvs-rep/gcc/gcc/config/ns32k/__unordsf2.c,v
> retrieving revision 1.1
> diff -c -3 -p -r1.1 __unordsf2.c
> *** config/ns32k/__unordsf2.c	29 Jul 2002 15:12:23 -0000	1.1
> --- config/ns32k/__unordsf2.c	30 Jul 2002 01:38:11 -0000
> *************** Boston, MA 02111-1307, USA.  */
> *** 22,28 ****
>   
>   # define ISNAN(x) (				\
>     {						\
> !     unsigned int t = (x);			\
>       ((t & 0x7f800000) == 0x7f800000) &&		\
>       ((t & 0x7fffff) != 0);			\
>     })
> --- 22,28 ----
>   
>   # define ISNAN(x) (				\
>     {						\
> !     unsigned int t = * (unsigned int *) &(x);		\
>       ((t & 0x7f800000) == 0x7f800000) &&		\
>       ((t & 0x7fffff) != 0);			\
>     })
> 
> --pG7rHtfCel--
> 
>