[Bug libstdc++/33628] unary_function and pointer_to_unary_function issues with void template arguments

[jwakely dot gcc at gmail dot com]

------- Comment #2 from jwakely dot gcc at gmail dot com  2008-03-08 15:25 -------
(In reply to comment #0)
> // 1
> #include <functional>
> 
> typedef int     value_type;
> 
> // void argument type
> template class std::unary_function<void, value_type>;

eurgh, even more of an abomination than int f(void)  :-)

> template class std::pointer_to_unary_function<void, value_type>;

This fails, correctly I think.  It's analogous to this illegal declaration:
  typedef void T;
  int f(T);

> // void return type
> template class std::unary_function<value_type, void>;
> template class std::pointer_to_unary_function<value_type, void>;

These are OK, aren't they?

> And more specific things:
> 
> // 2
> #include <functional>
> void foo() { };
> 
> // pointer_to_unary_function
> void test01()
> {
>   typedef std::pointer_to_unary_function<void, void> pfunc;
>   pfunc p(&foo);

this is invalid, foo is not a unary function so the number of parameters
doesn't match.  The declaration void f(void) is valid for C compatibility, but
that doesn't make it a unary function.
14.8.2para2 says type deduction can fail due to "Attempting to create a
function type in which a parameter has a type of void."  That's discussing a
different situation, but the principle is the same. I couldn't find anything
more specific.

> // 3
> #include <functional>
> // unary_function
> struct void_unary : public std::unary_function<void, void>
> { 
>   typedef std::unary_function<void, void> base_type;
>   typedef base_type::Arg argument_type;
>   typedef base_type::Result result_type;
> };

Hmm, I suppose this is legal, although ill-advised.

> Is this usage standards-conforming?

I think the declarations and typedefs are legal, but using them isn't
necessarily possible. So you can't assign &foo to a unary_function, and you
couldn't define this member:
  Result void_unary::f(Arg);

Thankfully the void f(void) syntax is not valid in templates, because there's
no need for C compatibility in template code.

I don't think this is a bug.


-- 

jwakely dot gcc at gmail dot com changed:

           What    |Removed                     |Added
----------------------------------------------------------------------------
                 CC|                            |jwakely dot gcc at gmail dot
                   |                            |com


http://gcc.gnu.org/bugzilla/show_bug.cgi?id=33628