[PATCH 04/31] stddef: Avoid warning with clang with offsetof()

[Simon Glass]

Hi Tom, Rasmus,

On Thu, 13 Jan 2022 at 06:07, Tom Rini <trini at konsulko.com> wrote:
>
> On Thu, Jan 13, 2022 at 09:08:13AM +0100, Rasmus Villemoes wrote:
> > On 01/11/2021 02.17, Simon Glass wrote:
> > > Some bright sparks have decided that a cast on a constant cannot be a
> > > constant, so offsetof() produces this warning on clang-10:
> > >
> > > include/intel_gnvs.h:113:1: error: static_assert expression is not an
> > >     integral constant expression
> > > check_member(acpi_global_nvs, unused2, GNVS_CHROMEOS_ACPI_OFFSET);
> > > ^~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
> > > include/linux/kernel.h:284:2: note: expanded from macro 'check_member'
> > >         offsetof(struct structure, member) == (offset), \
> > >         ^~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
> > > include/linux/stddef.h:20:32: note: expanded from macro 'offsetof'
> > >                                 ^
> > > include/intel_gnvs.h:113:1: note: cast that performs the conversions of
> > >     a reinterpret_cast is ot allowed in a constant expression
> > > include/linux/stddef.h:20:33: note: expanded from macro 'offsetof'
> > >
> > > Fix it by using the compiler built-in version, if available.
> > >
> > > Signed-off-by: Simon Glass <sjg at chromium.org>
> > > ---
> > >
> > >  include/linux/stddef.h | 8 +++++++-
> > >  1 file changed, 7 insertions(+), 1 deletion(-)
> > >
> > > diff --git a/include/linux/stddef.h b/include/linux/stddef.h
> > > index c540f6100d4..a7f546fdfe5 100644
> > > --- a/include/linux/stddef.h
> > > +++ b/include/linux/stddef.h
> > > @@ -1,6 +1,8 @@
> > >  #ifndef _LINUX_STDDEF_H
> > >  #define _LINUX_STDDEF_H
> > >
> > > +#include <linux/compiler_types.h>
> > > +
> > >  #undef NULL
> > >  #if defined(__cplusplus)
> > >  #define NULL 0
> > > @@ -14,7 +16,11 @@
> > >
> > >  #ifndef __CHECKER__
> > >  #undef offsetof
> > > -#define offsetof(TYPE, MEMBER) ((size_t) &((TYPE *)0)->MEMBER)
> > > +#ifdef __compiler_offsetof
> > > +#define offsetof(TYPE, MEMBER)     __compiler_offsetof(TYPE, MEMBER)
> > > +#else
> > > +#define offsetof(TYPE, MEMBER)     ((size_t)&((TYPE *)0)->MEMBER)
> > > +#endif
> > >  #endif
> >
> >
> > Can we please just drop the useless indirections? Any compiler we care
> > about provides __builtin_offsetof(), so just make the whole thing
> >
> >   #undef offsetof
> >   #define offsetof(TYPE, MEMBER)  __builtin_offsetof(TYPE, MEMBER)
> >
> > And by "compilers we care about", that includes sparse - it has had
> > __builtin_offsetof ever since 2007.
> >
> > And we can nuke the __compiler_offsetof definition in compiler_types.h,
> > and if any users exist, just mechanically convert them to offsetof().
>
> We should re-sync this with upstream, but that will keep the indirection
> you're objecting to here, I think.

Yes, with this patch, this part of the file is the same as Linux (e.g. v5.16).

So I'd like to apply it as is and deal with improvements later.
Perhaps in lockstep with Linux?

Regards,
Simon