Re: [PATCH] hash: pacify gcc -Wsuggest-attribute=const

[Pádraig Brady]

On 19/05/2023 12:03, Bruno Haible wrote:
> Hi Pádraig,
>
> > diff --git a/lib/hash.c b/lib/hash.c
> > index 918aa0d1c3..332cca6df9 100644
> > --- a/lib/hash.c
> > +++ b/lib/hash.c
> > @@ -492,7 +492,7 @@ check_tuning (Hash_table *table)
> >      TUNING, or return 0 if there is no possible way to allocate that
> >      many entries.  */
> >   
> > -static size_t _GL_ATTRIBUTE_PURE
> > +static size_t _GL_ATTRIBUTE_CONST
> >   compute_bucket_size (size_t candidate, const Hash_tuning *tuning)
> >   {
> >     if (!tuning->is_n_buckets)
>
> This looks wrong. Per the definition of the 'const' attribute [1]
>    "The const attribute prohibits a function from reading objects that
>     affect its return value between successive invocations. However,
>     functions declared with the attribute can safely read objects that
>     do not change their return value, such as non-volatile constants."
> In this function's body, testing tuning->is_n_buckets affects the
> return value.
>
> Bruno
>
> [1] 
> https://gcc.gnu.org/onlinedocs/gcc-13.1.0/gcc/Common-Function-Attributes.html

I was unsure about that TBH.

In coreutils usage tuning==NULL so the static const default_tuning is passed 
always.
So I guess _GL_ATTRIBUTE_CONST is fine there.

In general for any tuning though, since the tuning param struct is declared 
const,
can the compiler not assume successive invocations against the _same object_
returns the same value?

cheers,
Pádraig