[boost] Re: numeric_cast problem

From: Alan Stokes (alan_at_[hidden])
Date: 2004-01-07 10:13:37

Michael D. Borghardt wrote:
> ====> new lines
> 
> template <>
> struct greater_than_type_max<false, true>
> {
>     template <class X, class Y>
>     static inline bool check(X x, Y y_max)
>     {
>         return x >=0
>         // && static_cast<X>(static_cast<Y>(x)) != x; } // orignal line
> ===>               && ((static_cast<X>(static_cast<Y>(x)) != x)
> ===>                      || (sizeof(Y) >= sizeof(X) ? static_cast<Y>(x) >
> y_max : x > static_cast<X>(y_max)));
>     }

That still retains the static_cast<X>(static_cast<T>(x)) != x from the original.

I've just come across a problem relating to this with the original numeric_cast.

numeric_cast<unsigned>(1.5) throws, when it should succeed (and return 1U). This 
is because static_cast<double>(static_cast<unsigned>(1.5)) is 1.0, not 1.5.

I think the same problem must exist in your reformulation. Do you have a test 
case for this? Can you do away with the original test altogether?

- Alan