Re: C++17 skeleton with move semantics and std::variant

[Hans Åberg]

> On 9 Apr 2018, at 18:16, Frank Heckenbach <address@hidden> wrote:
> 
> Hans Åberg wrote:
> 
>>>> One can have a preprocessor macro '#if (__cplusplus == 201703L) ...', 
>>>> would it be of interest.
>>> 
>>> What would it do? (BTW, I think it should be ">=".)
>> 
>> Something like, but with suitable template arguments:
>> 
>> #if !(__cplusplus >= 201703L)
>> namespace std {
>>  // Workaround when C++17 <variant> not available.
>>  template<class T>
>>  using variant = some::variant<T>;
>> }
>> #endif
> 
> (Or simply "namespace std { using mpark::variant; }", I think.)

This is better. I use the workaround with Apple clang, which when finally got 
C++17 put those new types in std::experimental!  

>>> I don't want the skeleton to depend on an external library (and as I
>>> said, there might be at least two candidates), so I think it's
>>> better to leave it to the user. (For me, I've put mpark's variant
>>> header in /usr/local/include for g++-6 to use, while g++-7 will use
>>> its own variant implementation, so the parser will work with both
>>> compilers without change.)
>> 
>> Indeed, leave as it is. The idea above would leave it open for other users 
>> to introduce their own dependencies.
> 
> I think so. With the various "%code" blocks etc., there are enough
> possibilities for users to do the above or similar.

It should suffice in %code requires.