Re: [Boost-users] [move] non-inline move constructor - won't compile

Subject: Re: [Boost-users] [move] non-inline move constructor - won't compile
From: Vicente J. Botet Escriba (vicente.botet_at_[hidden])
Date: 2013-11-03 11:32:50

Le 03/11/13 11:19, Krzysztof Czainski a écrit :
> 2013/11/2 Vicente J. Botet Escriba <vicente.botet_at_[hidden] 
> <mailto:vicente.botet_at_[hidden]>>
>
>     Le 02/11/13 12:02, Krzysztof Czainski a écrit :
>
>         Hi,
>
>         I've been trying to writhe a non-inline move constructor, to
>         place its' implementation in a .cpp file. I'm having problems
>         compiling my code, which I managed to simplify (after
>         examining output of the preprocessor) down to this short code
>         snippet:
>
>         #include <boost/move/core.hpp>
>
>         struct X
>         {
>             X( ::boost::rv< X >& x );
>             operator ::boost::rv<X>&() { return *static_cast<
>         ::boost::rv<X>* >(this); }
>         };
>
>         X::X( ::boost::rv< X >& x ) {}
>
>         Compiling a .cpp file with the above contents gives me:
>
>         a.cpp:41:1: error: prototype for 'X::X(boost::rv<X>&)' does
>         not match any in class 'X'
>          X::X( ::boost::rv< X >& x ) {}
>          ^
>         a.cpp:35:8: error: candidates are: X::X(const X&)
>          struct X
>                 ^
>         a.cpp:37:5: error: X::X(boost::rv<X>&)
>              X( ::boost::rv< X >& x );
>              ^
>
>         I'm using MinGW gcc 4.8.1 in default -std mode (not-C++11
>         mode), boost-1.54.
>
>     It works for me with clang-3.1 and clang-3.2, but not with gcc :(
>     Really curious.
>
>
>         Am I doing something wrong here, or is it a bug in the compiler?
>
>     I would say that it is a bug on the compiler.
>
>     Best,
>     Vicente
>
>
> Thanks for checking this out, Vincente.
>
> I simplified it a bit more, found out, that it has to do with 
> attribute may_alias, and reported it to MinGW:
> https://sourceforge.net/p/mingw/bugs/2126/
>
> I also came up with a workaround, an example of which can be found in 
> the attached files.
>
Glad to see that you know where is the problem.

Ion, may I suggest you to add this on the documentation?

Best,
Vicente