https://gcc.gnu.org/bugzilla/show_bug.cgi?id=128657
--- Comment #1 from Jonathan Wakely <redi at gcc dot gnu.org> --- It's definitely a bug that the declaration in <type_traits> does not match the definition in <bits/move.h>. But if the declaration is fixed to match the definition then the testcase here still fails. The standard just says that std::swap is constrained, it doesn't say how. That means we're allowed to use a requires-clause, or std::enable_if, or other SFINAE tricks. So the ns::swap overload fails to be more constrained, and the testcase is still ambiguous.
