rsmith added inline comments.
================
Comment at: include/tuple:185-186
+ // Allow "int&&" to bind to 'int const&'
+ || (is_rvalue_reference<_Tp>::value && is_const<_RawHp>::value &&
+ is_same<_RawHp, const _RawTp>::value)
>;
----------------
It would be reasonable to consider `is_base_of` here too.
================
Comment at: include/tuple:190
|| (is_lvalue_reference<_Hp>::value && _CheckLValueArg::value)
|| (is_rvalue_reference<_Hp>::value &&
!is_lvalue_reference<_Tp>::value);
+#else
----------------
This line looks wrong to me. This disallows an rvalue reference tuple member
from binding to an rvalue reference argument, no? Can you
```
return !is_reference<_Hp>::value || (is_reference<_Tp>::value &&
is_convertible<_RawTp*, _RawHp*>::value) || (reference wrapper special case);
```
instead? That should at least only reject valid code in cases where `_RawTp` is
a class type that converts to a reference type.
https://reviews.llvm.org/D41977
_______________________________________________
cfe-commits mailing list
cfe-commits@lists.llvm.org
http://lists.llvm.org/cgi-bin/mailman/listinfo/cfe-commits
