On FreeBSD 12.0/i386 I get a compilation error when building a testdir
for module 'rawmemchr'.

On the master branch, the first error is:

In file included from ../../gllib/rawmemchr.c:17:
In file included from ../config.h:1018:
In file included from /usr/include/stddef.h:41:
/usr/include/sys/_types.h:106:35: error: expected expression
        long long __max_align1 __aligned(_Alignof(long long));
                                         ^
../config.h:964:39: note: expanded from macro '_Alignof'
#     define _Alignof(type) offsetof (struct { char __a; type __b; }, __b)
                                      ^

On the stable-202301 branch, the first error is:

In file included from ../../gllib/rawmemchr.c:17:
In file included from ../config.h:925:
In file included from ./stdalign.h:131:
In file included from /usr/include/stddef.h:41:
/usr/include/sys/_types.h:106:35: error: expected expression
        long long __max_align1 __aligned(_Alignof(long long));
                                         ^
./stdalign.h:70:36: note: expanded from macro '_Alignof'
#  define _Alignof(type) offsetof (struct { char __a; type __b; }, __b)
                                   ^

What happens is that
  - config.h includes <stdalign.h>, per the patch that I submitted in
    <https://lists.gnu.org/archive/html/bug-gnulib/2022-09/msg00144.html>.
  - This <stdalign.h> is the gnulib-generated one.
    (In <https://lists.gnu.org/archive/html/bug-gnulib/2022-09/msg00144.html>
    the purpose of my patch was to avoid including the gnulib-generated
    stdalign.h, but it did not achieve that goal since on this platform
    HAVE_STDALIGN_H is 1, following these configure findings:
      checking for stdalign.h... yes
      checking for alignas and alignof... no
    . Anyway.)
  - At the end of <stdalign.h>, <stddef.h> gets included.
  - <stddef.h> includes <sys/_types.h>, which attempts to define max_align_t.
    For this it uses _Alignof(long long), which expands to
    offsetof (struct { char __a; long long __b; }, __b)
  - But at this point offsetof is not yet defined (since <stddef.h> is still
    being included).

The following patch fixes it. One in master, a different one for the
stable-202301 branch. (On the stable-202207 branch there is no problem.)