uniform_real_distribution fails to compile with result_type having explicit conversion constructor from float

[b7.10110111@…]

Consider the following test program:

#include <random>
#include <boost/random/uniform_real_distribution.hpp>
#include <half.hpp>

template<typename Float>
void test()
{
    std::random_device rd;
    std::mt19937 mt(rd());
    boost::random::uniform_real_distribution<Float> rnd(Float(1),Float(10));
    rnd(mt);
}

int main()
{
    test<float>(); // this works
    test<half_float::half>();  // and this doesn't compile
}

It uses half.hpp from half.sourceforge.net. This type has a peculiar behavior with arithmetic involving types other than it: the variable converts implicitly to float, and then the explicit constructor half_float::half(float) prevents some code from compiling. Namely, the code above doesn't compile due to the return 2 * generate_uniform_real(...) in generate_uniform_real(). I've seen there already exist explicit casts like for the arguments T(min_value / 2) and the like. I've been able to fix it locally with the attached patch.

[b7.10110111@…]

Proposed patch