Re: [Sab] BUG: arithmetic with NaN

[Kaiserovi]

Robin wrote:
>         #include <nan.h>
>         // ...
>         x = NAN;

Pretty easy. I'm not sure this is supported by MSVC; if not, the
division by 0 is better. 

> But why not go the whole way - drop the _NaN member,
> have setNaN() do sthg like the above, & isNaN use isnan()
> from math.h ?

I agree. There's also a problem with the +- infinity - can that be
generated in a portable way? 

Tom