On Fri, Sep 6, 2024 at 11:16 PM Martin Uecker <ma.uec...@gmail.com> wrote:
>
> Am Samstag, dem 07.09.2024 um 00:12 +0000 schrieb Qing Zhao:
> > Now, if
> >
> > 1. __builtin_get_counted_by should return a LVALUE instead of a pointer
> > (required by CLANG’s design)
> > And
> > 2. It’s better not to change the behavior of __builtin_choose_expr.
> >
> > Then the solution left is:
> >
> > __builtin_get_counted_by (p->FAM) returns a LVALUE as p->COUNT if p->FAM
> > has a counted_by attribute, if p->FAM does not have a counted_by attribute,
> > silently do nothing. (Or just issue warning if Linux is OKEY with such
> > waning).
>
> What does silently do nothing mean?
>
> /* do nothing */ = counter;
>
> will still fail to compile. So I guess you have something
> else in mind?
>
>
> The new _Generic selection also works if you return a
> lvalue of type void:
>
> struct foo x;
> _Generic(typeof(__counted_by(&x)), void: (int){ 0 },
> default: __counted_by(&x)) = 10;
>
> https://godbolt.org/z/41E3oj84o
>
> So why not do this then?
>
I'm for any solution that a.) works with the fewest changes needed to
other builtins, and 2.) isn't super convoluted. The '_Generic' looks
like a potential compromise.
-bw
