This is so *cosmetic*, but wouldn't it be more consistent with other macros like _LIBCPP_CONSTEXPR, etc..?
I didn't add _LIBCPP_INLINE_VARIABLE because we shouldn't make a habit of having variables that are sometimes inline and sometimes not. So it should be as ugly as possible when you have to do it. Hopefully the current usages in <utility> will be all we ever need.
Eric, I can see a potential for this macro. We have a lot of tag types instances in Standard Library -- seq, par, piecewise_construct, nullopt, ignore to name a few. All of them are considered to become inline to prevent odr-violations. There are 2 National Body comments addressing this issue (GB28, FI9). I absolutely agree about monstrosity of such macros, but from my perspective it's inevitable.