8/17    In C++ (and Java), virtual functions are "turned off" in the
        constructors and destructors, right?  Since the child class
        isn't available.
        \_ you can definitely have virtual destructors (in C++).
           It is strongly recommended.
           \_ I meant calling a VF from a ctor or dtor.  --OP
              \_ even if it were legal, that is really bad style.  A parent
                 class having knowledge of children classes? Bad bad bad.
                 \_ Um, the point of a VF is that you don't have to know the
                    exact behavior.  Fortunately, this isn't allowed anyway.
        \_ They're not "turned off"--that is, the compiler doesn't prevent you
           from calling them, but you shouldn't expect them to work.  When
           you're constructing, you start at the base constructor and work your
           way down the inheritance chain.  So if you try to call a virtual
           function in that process, the full type hasn't been constructed and
           you can't expect anything predictable.  Similarly in the destructor.
           This is a reason many people on the C++ community are asking for
           "post-construction" and "pre-destruction" functions.
           \_ This is incorrect.  It is perfectly predictable; in a constructor
              you are guaranteed to have the virtual functions defined in that
              class or any base class.  --pld

lazy emacs user was here