https://gcc.gnu.org/bugzilla/show_bug.cgi?id=118009
--- Comment #10 from Jonathan Wakely <redi at gcc dot gnu.org> ---
(In reply to Jonathan Wakely from comment #8)
> To put it another way: the obsolete fixincludes rule creates an unnecessary
> and extremely fragile dependency from GCC to some Glibc internal details. If
> we just remove that unnecessary dependency, users won't need to know how to
> use mkheaders because GCC won't care about changes to those Glibc internals.
Oh wow, it's even worse than I thought.
The line we're "fixing" in GCC isn't even used by GCC:
#if __GNUC_PREREQ (11, 0)
extern int __REDIRECT_NTHNL (__sigsetjmp_cancel,
(struct __cancel_jmp_buf_tag __env[1],
int __savemask),
__sigsetjmp) __attribute_returns_twice__;
#else
# define __sigsetjmp_cancel(env, savemask) \
__sigsetjmp ((struct __jmp_buf_tag *) (void *) (env), (savemask))
extern int __sigsetjmp (struct __jmp_buf_tag __env[1],
int __savemask) __THROWNL;
#endif
We "fix" the declaration that is only used for GCC 10 and older. So we create a
fragile copy of pthread.h that couples GCC to Glibc internals, and then don't
even compile the line that we "fixed".
