https://gcc.gnu.org/bugzilla/show_bug.cgi?id=80183
--- Comment #7 from François Dumont <fdumont at gcc dot gnu.org> --- (In reply to Jonathan Wakely from comment #4) > I think this is my fault, I asked François to remove a store to _M_color > that I thought was redundant. I was surprised that _M_color wasn't move along with other members, with your patch it is now. However to strictly restore previous behavior you could have only set _M_color to red in the move constructor whether _M_move_data or _M_reset was called. But I prefer current code of course.