Why convertible_to instead of same_as?

My reasoning was, it used to be trailing-return-type which used to mean almost the same as convertible_to<T> prior to P1452r2. convertible_to preserves this intent.

Unless -> const tuple_element_t<N, T>& was a bug from the outset?

I think it probably should have always required exactly that type, it was just more awkward to specify that prior to P1452R2. But I'm not sure about the original intent.

I agree that as an editorial change, using convertible_to is closer to the earlier meaning. But I wonder if we actually have a non-editorial defect and should require exactly that type.

@CaseyCarter ?

The conversion was intended in the original design: std::get<X>(t) may return a reference to const, or reference to non-const, or even a prvalue for subrange. The exact type and value category are largely immaterial - this is just a "sanity check" requirement - since the only use to which the implementation puts get is to return std::get<X>(...) from operator*.

I should point out that LWG 3323 resolves this, which we will be voting to move into the WD this week in Prague.

Ok, so where do we go from here? Should I close this PR given that it is reported as a LWG issue?

Fixed by 74487d3.