On Mon, Sep 16, 2019 at 5:22 AM Paolo Carlini <paolo.carl...@oracle.com> wrote: > > On 15/09/19 16:22, Jason Merrill wrote: > > On 9/12/19 9:41 AM, Paolo Carlini wrote: > >> + if (!valid_array_size_p (dname > >> + ? declarator->id_loc : input_location, > > > > Use the id_loc local variable? > > This diagnostic is inside the loop over declarator->declarator. > Eventually, outside the loop, the id_loc local is updated to the final > declarator->id_loc or input_location. Norrmally in the loop we use the > current declarator->id_loc: what I tested seems more correct to me (we > have to account for input_location too because valid_array_size_p, > shared with the C front-end, wants a sound location)
OK. Jason