On Thu, Jul 21, 2005 at 06:29:03PM -0600, Warren Young wrote:
> Chris Frey wrote:
> >
> >The latest SVN version has this "fixed", but incorrectly in my
> >opinion. It currently just returns *this, whereas I don't think
> >operator=() should be defined at all, just declared.
>
> Does this actually suppress the compiler-generated version?
Yep.
tmp $ cat oe.cc
class A
{
private:
A(const A&);
A& operator=(const A&);
int i;
public:
A() : i(5) {}
};
int main()
{
A a, b;
a = b;
return 0;
}
tmp $ g++ -Wall -o oe oe.cc
oe.cc: In function `int main()':
oe.cc:5: error: `A& A::operator=(const A&)' is private
oe.cc:16: error: within this context
This also flags an error for derived classes too. The thing you want
to be careful about is if you're doing virtual operator=(), with a
base class as the const reference argument. For example:
class A {
public:
virtual A& operator=(const A&); // this overrides the default
};
class B : public A {
public:
virtual A& operator=(const A&); // this does not
// there is a hidden, automatic, operator=(const B&) here too!
// and the automatic operator=() will call the user-defined
// A::operator=() for us.
};
> Regardless, operator= does normally return *this, so the only problem
> with doing it this way is that it generates a tiny bit more code than if
> it were only declared and not defined.
I think all we care about is the compiler warning, so no copy is done
accidentally.
- Chris
