The conversion to function pointer is relatively new: It was introduced with N3043 on February 15, 2010.
While e.g. GCC 4.5 implements it, Visual Studio 10 was released on April 12, 2010 and thus just didn't implement it in time. As James pointed out, this will be fixed in future releases.
For the moment you have to use one of the alternative solutions provided here.
Technically something like the following workaround would work, but without variadic templates its no fun to generalize it (Boost.PP to the rescue...) and there is no safety net against passing capturing lambdas in:
typedef uint64_t (*WeightFunc)(void* param);
template<class Func> WeightFunc make_function_pointer(Func& f) {
return lambda_wrapper<Func>::get_function_pointer(f);
}
template<class F> class lambda_wrapper {
static F* func_;
static uint64_t func(void* p) { return (*func_)(p); }
friend WeightFunc make_function_pointer<>(F& f);
static WeightFunc get_function_pointer(F& f) {
if (!func_) func_ = new F(f);
return func;
}
};
template<class F> F* lambda_wrapper<F>::func_ = 0;
// ...
WeightFunc fp = make_function_pointer([](void* param) -> uint64_t { return 0; });
与恶龙缠斗过久,自身亦成为恶龙;凝视深渊过久,深渊将回以凝视…