Are all types extending std::is_floating_point required to implement FENV in C++11 standard?

I'm implementing a function as follows:

template <typename FP>
void do_something(FP f){
    static_assert(std::is_floating_point<FP>::value, "not a floating point");
    ...
}

In this context, any floating point can be used (native or custom, i.e. multi-precision float libraries).

I wonder if standard states anything about what to expect from types overloading the is_floating_point. Can I count for them being required to work with FENV?

Can I assume the following code should work properly if a custom floating point implementation is overloading the std::is_floating_point?

template <typename FP>
void do_something(FP f){
    static_assert(std::is_floating_point<FP>::value, "not a floating point");
    if (std::fetestexcept(FE_DIVBYZERO)){
        ...
    }
}