This is the mail archive of the
libc-alpha@sourceware.org
mailing list for the glibc project.
Re: [hurd,commited] hurd: do not check Mach and Hurd headers
On Sun, 4 Mar 2018, Samuel Thibault wrote:
> Joseph Myers, on sam. 03 mars 2018 22:53:23 +0000, wrote:
> > I think it's dubious to have a #error requiring _GNU_SOURCE to be used,
> > because it should always be possible to write a header in a way that
> > doesn't have such a requirement. But in any case where, after analysis,
> > such a #error is found to make sense
>
> Notably, most Hurd interfaces use the error_t type, which is GNU-only.
> I assume this is enough to make it a #error case?
No. Create bits/types/error_t.h (which would have a generic version and a
Hurd version), which would define error_t (subject to __error_t_defined as
a multiple-include guard). Then make errno.h include
<bits/types/error_t.h> if __USE_GNU, while all the Hurd headers using that
type would include <bits/types/error_t.h> unconditionally.
(sysdeps/mach/hurd/bits/errno.h would no longer need to define error_t and
wouldn't need to include <bits/types/error_t.h> either because of errno.h
doing so - I'm assuming the current reason for defining error_t there is
simply to get a Hurd-specific definition instead of the default from
errno.h, which would be dealt with by having a Hurd-specific
bits/types/error_t.h.)
Having such a bits/types/*.h header is the normal way in glibc of dealing
with types that different headers need under different conditions.
--
Joseph S. Myers
joseph@codesourcery.com