Variant.type bug ?

[Voitech via Digitalmars-d-learn]

On Wednesday, 23 March 2016 at 12:52:24 UTC, Adam D. Ruppe wrote:
> On Wednesday, 23 March 2016 at 08:01:36 UTC, Voitech wrote:
>> Hi Variant stores variant.type as not the "highest" in 
>> hierarchy.
>
> Yeah, it stores the static type. You can use it to get that 
> then do a normal dynamic cast to test for a more derived type.

Ok but how to handle sittuation like this ?

class TypeHolder{
	import std.variant;
	Variant[TypeInfo] data;

	void add(T)(T value){
		data[typeid(value)]=value;
	}

	T getByType(T)(){
		Variant retVar=data.get(typeid(T),Variant(null));
		T val=retVar.get!T; //fails
		return val;
	}

}
unittest{
	import std.variant;
	A a= new A;
	B b= new B;
	C c = new C;

	A ab= new B;
	A ac = new C;
	TypeHolder holder = new TypeHolder;
	holder.add(a);
	holder.add(ab);
	holder.add(ac);
	assert(holder.data.length==3);
	A result=holder.getByType!A;
	assert(result==a);
	result=holder.getByType!B; //fails
	assert(result==ab);
	result=holder.getByType!C; //fails
	assert(result==ac);
}

I can hold objects in other AA but Object[TypeInfo] rather  than 
Variant. Or is there a way to get super type of provided T ?