https://gcc.gnu.org/bugzilla/show_bug.cgi?id=79420
--- Comment #6 from Jason Merrill <jason at gcc dot gnu.org> --- (In reply to Marek Polacek from comment #5) > Given PR79463 and Comment 4, it seems that we shouldn't change at least > VAR_DECLs and OVERLOADs back to dependent. We can treat them as dependent, but we shouldn't clobber the TREE_TYPE of a VAR_DECL or OVERLOAD. We should only do that for EXPR_P trees.
