[libcxx-commits] [libcxx] [libc++] Fix ambiguity due to non-uglified member typedefs (PR #121664)

[Peng Liu via libcxx-commits]

================
@@ -75,8 +75,10 @@ template <class, class _Cp, bool _IsConst, class _Tp, class _Proj, __enable_if_t
 _LIBCPP_HIDE_FROM_ABI _LIBCPP_CONSTEXPR_SINCE_CXX20 __iter_diff_t<__bit_iterator<_Cp, _IsConst> >
 __count(__bit_iterator<_Cp, _IsConst> __first, __bit_iterator<_Cp, _IsConst> __last, const _Tp& __value, _Proj&) {
   if (__value)
-    return std::__count_bool<true>(__first, static_cast<typename _Cp::size_type>(__last - __first));
-  return std::__count_bool<false>(__first, static_cast<typename _Cp::size_type>(__last - __first));
+    return std::__count_bool<true>(
+        __first, static_cast<typename __size_difference_type_traits<_Cp>::size_type>(__last - __first));
+  return std::__count_bool<false>(
----------------
winner245 wrote:

Yeah, that would be the ideal approach if these algorithms are only for `bitset::iterator`. However, these algorithms are also used by `vector<bool>::iterator`, where `_Cp` in `__bit_iterator<_Cp, _IsConst>` becomes `vector<bool>`, for which the required member type name is `difference_type`. Only when `_Cp` is `bitset`, the member type becomes `__difference_type`. To address this name difference, I introduced `__size_difference_type_traits`. 

https://github.com/llvm/llvm-project/pull/121664