On Mon, Feb 3, 2020 at 9:32 PM Paul Eggert <egg...@cs.ucla.edu> wrote: > > On 2/3/20 3:00 PM, Jeffrey Walton wrote: > > > Or, you can use __has_attribute(_Noreturn). __has_attribute(_Noreturn) > > behaves as expected. > > It isn't documented to work, and it doesn't work for me with clang > version 9.0.0 (Fedora 9.0.0-1.fc31). > > > I would not worry about other software using __has_attribute. It is a > > reserved symbol due to the leading undersocres. Userland software > > should not have been using it in the first place. > > Gnulib is userland software, so we cannot use your argument to justify > putting "#define __has_attribute(x) 0" into Gnulib.
Yeah, you're right. This is how GCC recommends using __has_attribute: https://gcc.gnu.org/onlinedocs/cpp/_005f_005fhas_005fattribute.html Jeff