🚀 go-pugleaf

wil@ittpub.nl (Wil Evers) wrote:

>The ARM, section section 12.1, says:
>
>"A copy constructor for a class X is a constructor that can be called to  
>copy an object of class X; that is, one that can be called with a single  
>argument of type X."
>
...
>In contrast, the DWP says (12.8.2, [class.copy])
>
>"A constructor for class X is a copy constructor if its first parameter is  
>of type X& or const X&, and either there are no other parameters or all  
>other parameters have default arguments."
>
...
>Is this a deliberate change from the definition in the ARM? If so, does  
>anyone know what the rationale behind this change is? (Personally, I think  
>the definition in the DWP is preferable because I occasionally need to  
>write such a constructor to `upgrade' a base class object to a derived  
>class object and I got bitten once or twice because the compiler wouldn't  
>generate a copy constructor for the derived class if I did so. Still, I  
>can also imagine code that would break because of this change).

Actually, I prefer the ARM version specifically because of the issue we
are discussing in the thread about auto_ptr<>. If you want the argument
to your copy constructor going through some sort of intermediate proxy
conversion, then you don't want that work undone by the compiler:

template <class T>
class auto_ptr<T>
{
  class proxy { ... };
  ...
  operator proxy();	// conversion to proxy<T>
  auto_ptr<T>(const proxy&);  // line 3
  ...
};

According to the ARM, line 3 declares a copy constructor because you can
call it with an auto_ptr<T> argument. The compiler will not generate a
copy constructor for you. This is good, in this case. But according to
the DWP, line 3 does NOT declare a copy constructor and all of your
lovely work is lost as the compiler generates its own (incorrect) copy
constructor. Declaring a private copy constructor does not help because
the compiler still sees the copy constructor and just makes it illegal
to try to use it.

I can always manually make a copy constructor that does the same thing
that the compiler-generated one would do, but I can't always suppress
the compiler's urge to do something for me.
-------------------------------------------------------------
Pablo Halpern                   phalpern@truffle.ultranet.com


[ comp.std.c++ is moderated.  Submission address: std-c++@ncar.ucar.edu.
  Contact address: std-c++-request@ncar.ucar.edu.  The moderation policy
  is summarized in http://dogbert.lbl.gov/~matt/std-c++/policy.html. ]