[patch] C++14: N3671 Making non-modifying sequence operations more robust

[Jonathan Wakely]

On 10 June 2013 21:16, François Dumont wrote:
> @@ -810,6 +823,17 @@ _GLIBCXX_BEGIN_NAMESPACE_VERSION
>           return !__builtin_memcmp(__first1, __first2, sizeof(_Tp)
>                                    * (__last1 - __first1));
>         }
> +
> +#if __cplusplus > 201103L
> +      template<typename _Tp>
> +        static bool
> +        equal(const _Tp* __first1, const _Tp* __last1, const _Tp* __first2,
> +             const _Tp* __last2)
> +        {
> +         return !__builtin_memcmp(__first1, __first2, sizeof(_Tp)
> +                                  * (__last1 - __first1));
> +       }
> +#endif
>      };
>
>  I think there is a problem here. I had a closer look because __last2 is not
> consider in the version taking pointers.

Oops, yes, I copied the existing code then didn't update it.

> We could start reading after
> __last2, something Valgrin could catch perhaps. A consistent version would
> be:
>
> +#if __cplusplus > 201103L
> +      template<typename _Tp>
> +        static bool
> +        equal(const _Tp* __first1, const _Tp* __last1, const _Tp* __first2,
> +             const _Tp* __last2)
> +        {
> +         if (__builtin_memcmp(__first1, __first2, sizeof(_Tp)
> +                             * std::min(__last1 - __first1, __last2 -
> __first2))
> +           return false;
> +         return true;
> +       }
> +#endif
>
> But I am not sure it is the intended behavior because I can't see where
> those equal overloads are used. It looks like __equal2 has been implemented
> and used instead. Maybe some dead code to clean, no ?

It should be used, I'll check it.