Question

Could you please explain why following code does compile and works fine (checked on gcc-4.3.4). I thought selective inheritance cannot weaken or even strengthen access to members/methods. Doesn’t it break encapsulation rules?

#include <iostream>

class A {
protected:
    void foo() { std::cout << "foo" << std::endl;  }
};

class B : private A {
public:
    using A::foo;   //foo() becomes public?!
};

int main() {
    B b;
    b.foo();
    return 0;
}

Answer 1

From the language point of view, there’s nothing wrong with this (whether it’s good design is another matter).

Any class can choose to expose to a wider audience things that it has access to.

In principle, your example is no different to:

class B : private A {
public:
    void bar() { foo(); }
};

except that here foo() is exposed by proxy.

What you can’t do is the opposite: a publicly derived class can’t restrict access to things that are accessible via the base class.