Re: [PATCH] file-has-acl: avoid warning from bleeding-edge GCC

[Jim Meyering]

On Sun, May 28, 2023 at 10:44 AM Bruno Haible <bruno@clisp.org> wrote:
> How about adding -Wno-suggest-attribute to the list of warning options that
> are part of the GL_CFLAG_GNULIB_WARNINGS variable, no that such warnings
>   function might be candidate for attribute 'pure'
> or
>   function might be candidate for attribute 'const'
> no longer occur in gnulib source files, compiled as part of other packages?
>
> This would not only silence the warnings about static functions [1][2],
> but also about the global function definitions. Leading to slightly
> worse code generation by gcc in the long term.
>
> Paul, what do you think? Should we do that?
>
> Bruno
>
> [1] https://gcc.gnu.org/bugzilla/show_bug.cgi?id=109914
> [2] https://gcc.gnu.org/bugzilla/show_bug.cgi?id=109915

That would also solve the problem, but at a potential code-gen cost,
as you imply: code-gen for a few rare new and modified functions may
end up being degraded. I would endorse this approach if that GCC bug
is still unresolved in say October.