Re: [Boost-users] [Boost.Move] Conflict between std::move and boost::move (Boost 1.49.0)

Subject: Re: [Boost-users] [Boost.Move] Conflict between std::move and boost::move (Boost 1.49.0)
From: Jeffrey Lee Hellrung, Jr. (jeffrey.hellrung_at_[hidden])
Date: 2012-07-10 17:55:04

On Tue, Jul 10, 2012 at 2:18 PM, Jonathan Jones <
Jonathan.Jones_at_[hidden]> wrote:

>  Hi folks,
>
>  Consider the following code (which distilled from a real world example):
>
>
>   #include <boost/move/move.hpp>
>  #include <boost/optional/optional.hpp>
>  #include <utility>
>  #include <vector>
>
>  typedef boost::optional<int> data_type;
>  typedef std::vector<data_type> vector_type;
>
>  vector_type func()
>  {
>      vector_type vv;
>      return move(vv);
>  }
>
>
>  If I compile this code using Boost 1.49 on Visual Studio 2010, or on a
> recent version of g++ (e.g. 4.7.1) which supports some C++11 features, the
> code fails to compile, complaining about ambiguity between std::move and
> boost::move.  Strangely, the code compiles fine with Apple Clang version
> 3.1.
>
>  Is there a way to work around the compilation problem which doesn't
> involve explicitly qualifying the calls to move?  I'd prefer to use
> argument-dependent-lookup.
>

ADL doesn't really seem appropriate here, does it? Sounds like you intend
to call std::move regardless of the arguments.

>  Removing the includes of boost/move.hpp is not possible.  In the real
> world example, that header is included indirectly.
>
>  I did notice that if I define the macro
> BOOST_MOVE_USE_STANDARD_LIBRARY_MOVE, the compilation errors disappear.  Is
> this the correct approach for compilers which support std::move?  I find it
> odd that this macro isn't already defined for such compilers.
>

Well, (a) based on my quick reading of move.hpp, that looks like it should
do the thing you want; and (b) I don't the know the rationale for not
defining it for C++11 compilers, perhaps Ion can comment.

HTH,

- Jeff