Re: [PATCH] tools/compiler: match glibc 2.42 definition of __attribute_const__

[Nathan Chancellor <nathan@kernel.org>]

On Tue, Jun 30, 2026 at 06:41:06PM +0100, David Laight wrote:
> On Tue, 30 Jun 2026 23:58:40 +0900
> "Joy H.J. Lee" <rkr0k0r@gmail.com> wrote:
> 
> > glibc 2.42 added __attribute_const__ to sys/cdefs.h:
> > 
> >     # define __attribute_const__ __attribute__ ((__const__))
> > 
> > GCC 15 warns when a macro is redefined to a different replacement list
> > (-Wbuiltin-macro-redefined). Since host tool Makefiles (resolve_btfids,
> > objtool) pass -Werror, this conflict becomes fatal when building with
> > glibc 2.42 and GCC 15.
> > 
> > Per C11 §6.10.3, identical replacement lists are accepted silently.
> > Match the glibc definition exactly, including the space before "((", so
> > the redefinition is accepted without warning.
> > 
> > Signed-off-by: Joy H.J. Lee <rkr0k0r@gmail.com>
> > ---
> >  tools/include/linux/compiler.h | 2 +-
> >  1 file changed, 1 insertion(+), 1 deletion(-)
> > 
> > diff --git a/tools/include/linux/compiler.h b/tools/include/linux/compiler.h
> > index f40bd2b04..f2f54b038 100644
> > --- a/tools/include/linux/compiler.h
> > +++ b/tools/include/linux/compiler.h
> > @@ -119,7 +119,7 @@
> >  #define __read_mostly
> >  
> >  #ifndef __attribute_const__
> > -# define __attribute_const__
> > +# define __attribute_const__ __attribute__ ((__const__))
> 
> That doesn't look right, you are completely changing the value
> when it isn't already defined.

That's the point, no? If compiler.h is included before sys/cdefs.h, you
would get an instance of -Wbuiltin-macro-redefined like described in the
commit message. I am sure this is fine, we unconditionally define it in
compiler_attributes.h.

For the record, I have not seen an error like the commit message
describes in any of my build tests or environment. It would be good to
know under what circumstances this actually happens (or is this just
some contrivied issue or AI hallucination?).

Lastly, I don't think this should go through Kbuild, we don't touch
tools/. Perhaps this should go through Andrew Morton?

> >  #endif
> >  
> >  #ifndef __maybe_unused
> 

-- 
Cheers,
Nathan