Re: [RFC PATCH 2/3] hurd: Implement O_TMPFILE

[Sergey Bugaev]

On Mon, Jan 30, 2023 at 2:25 AM Samuel Thibault <samuel.thibault@gnu.org> wrote:
>
> Hello,
>
> Sergey Bugaev, le lun. 12 déc. 2022 14:46:35 +0300, a ecrit:
> > diff --git a/sysdeps/mach/hurd/bits/fcntl.h b/sysdeps/mach/hurd/bits/fcntl.h
> > index 17dcb384..b898a0c5 100644
> > --- a/sysdeps/mach/hurd/bits/fcntl.h
> > +++ b/sysdeps/mach/hurd/bits/fcntl.h
> > @@ -123,6 +123,11 @@
> >  # define O_CLOEXEC   0x00400000 /* Set FD_CLOEXEC.  */
> >  #endif
> >
> > +#ifdef __USE_GNU
> > +# define __O_TMPFILE 0x00800000 /* Make a new unnamed file.  */
> > +# define O_TMPFILE   (__O_TMPFILE | O_DIRECTORY)
> > +#endif
>
> I don't think we need the __O_TMPFILE variant, only the O_TMPFILE one?
>
> Linux uses __ variants just because it has per-arch definitions.

This was an attempt to mimic the Linux port's behavior, where it
automatically or's in O_DIRECTORY when you specify O_TMPFILE. Not that
it has any real meaning (at least for us) since my
__file_name_lookup_at () passes O_DIRECTORY automatically (instead of
any other flags) when O_TMPFILE is used. So should I remove this and
just have

# define O_TMPFILE 0x00800000 /* Make a new unnamed file.  */

?

Thanks for taking a look!

Sergey