https://gcc.gnu.org/bugzilla/show_bug.cgi?id=103827
--- Comment #17 from Jason Merrill <jason at gcc dot gnu.org> --- (In reply to Jan Hubicka from comment #16) > We are not too good on using const on automatic variables and static > ones with constructors since they become effectively const only when > they are constructed and it is not clear when this is done. We could certainly represent that somehow. Or assume that when passing the address to a function, only passing it as the first argument to a [cd]tor could change the value?
