Limitations of the conditional operator ?:

Question

I am using GCC 4.5 and have observed very peculiar behavior. I am wondering if there is something with this operator that I do not completely understand. I thought I was proficient in C++. I have a thin C++ wrapper class Wnd for Windows HWND objects with an implemented cast operator operator HWND ....

If I use the conditional operator like this (given input Wnd *p and a sample function SetParent(HWND)):

SetParent((p!=NULL) ? (HWND)(*p) : NULL)

The parent is properly set to NULL or p depending. This is what I would expect. However if dare to be lazy and write:

SetParent(p ? *p : NULL)

things go haywire. After running GDB I find that destructor is called on variable p after the call to SetParent. Any ideas what is going on here?

Edit Here is my Wnd class:

class Wnd{
        HWND m_hwnd;        ///< the actual handle
        WndFake *fake;      ///< store state here if we do not have a handle
    public:
        virtual ~Wnd();
        //contructor s
        Wnd(HWND wnd=NULL):m_hwnd(wnd),fake(NULL){}
        Wnd(DWORD sty,const jchar *title,const RECT &sz);
        operator HWND(){return m_hwnd;}
        operator HWND() const {return m_hwnd;}
    }

Answer 1

I suspect that your Wnd has non-explicit conversion constructor too that takes HWND or even int? If so then make it explicit.

Your Wnd probably does not have copy constructor and operator= declared? declare these private and don't define them.

Also remove operator HWND and add member function HWND hwnd() const; to your Wnd. Then the code will look readable like:

Setparent( p ? p->hwnd() : NULL );

I trust that when these mods are done you will find out what is wrong with your Wnd.

The problem manifests itself because the operands at both sides of : in ?: have to be of same type so NULL (0) is somehow convertible to Wnd. So the temporary copy of *p is made as return value of ?: then operator HWND is called to it.