On Mon, Feb 3, 2020 at 9:32 PM Paul Eggert <egg...@cs.ucla.edu> wrote: > ... > How about the attached (untested) patch instead? It looks at > __apple_build_version__.
Ack, tested OK on OS X 10.9 and a Linux smoke test. I was able to test _Noreturn.h through GnuTLS. I was not able to test gnulib-common.m4. (I'm still learning how to test Gnulib). I think CentOS 5 and NetBSD or FreeBSD would be good platforms to test also. I need some time to adjust my scripts. Recent package upgrades have broken some compiles. One warning.... I've never found __apple_build_version__ documented anywhere. I learned of __apple_build_version__ looking at diff's of 'clang -dM -E'. With that said, I've been using __apple_build_version__ for 5 or 10 years without a problem. Apple also uses __apple_build_version__ in places like https://opensource.apple.com/source/xnu/xnu-4570.61.1/san/kasan_internal.h.auto.html . > PS. I must say that Apple's hijacking and renumbering of __clang_major__ > seems pretty antisocial, and breaks some already-existing code in Gnulib > - what caused Apple to do that? Yeah, I have my feelings on it, too. Jeff