std::shared_ptr upcasting to base class - best method?

Question

Which conversion is better, and what is the difference?

class Base
{};

class Derived : public Base, public std::enable_shared_from_this<Derived>
{};

int main(int argc, const char * argv[])
{
    std::shared_ptr<Base> ptr1 = std::dynamic_pointer_cast<Base>(std::shared_ptr<Derived>(new Derived())); // version 1
    std::shared_ptr<Base> ptr2 = std::shared_ptr<Derived>(new Derived()); // version 2
    return 0;
}

Answer 1

As in other use cases of shared_ptr, you should prefer using make_shared instead of constructing the shared_ptr manually:

std::shared_ptr<Base> ptr2 = std::make_shared<Derived>();

This is essentially your version 2, plus the various benefits of make_shared.

Version 1 does a bunch of unnecessary stuff: First you construct a temporary shared_ptr<Derived>, then you dynamic_cast its contents to a base class pointer (while a static_cast would be sufficient here) and then you store that pointer in a different shared_ptr<Base>. So you have a lot of unnecessary runtime operations, but no benefits regarding type safety over Version 2.