> If an implementation chooses to ignore the attribute,
> does it mean the implication doesn't hold?
Yes, on the surface I think that's the most promising direction. If you
"ignore" the attribute you "ignore" the implied annotation (sort of by
necessity).

To elaborate, I think this would primarily be a transitional tool (e.g.,
for old code where you can't or don't want to add/change to the annotation).

I suspect this would /eventually/ result in attributes being
"discouraged" for most use cases in favor of annotations that have more
well defined semantics. I could picture a world where "annotations" are
never /ignored/ but /sometimes/ have /no effect/. Given reflection and
the class based design, I suspect we could do a lot better than the
(several times discussed) points of "what does it mean for X to be
ignored or not ignored?"

e.g., we could provide facilities more catered to the specific
annotation and its effects (borrowing some example code from CPP ref
<https://en.cppreference.com/w/cpp/language/attributes/no_unique_address>):

    struct Empty{}; // empty class
      
    struct X
    {
         int i;
         Empty e;
    };
      
    struct Y
    {
         int i;
         @std::no_unique_address Empty e;
    };
      

    constexpr auto x = ^Y::e;
    constexprstd::optional<std::no_unique_address> opt_nua_annot = single_annotation_of<std::no_unique_address>(x);

    constexpr if (opt_nua_annot.has_value()) {
       constexprstd::no_unique_address nua_annot = *opt_nua_annot;

       constexpr if (nua_annot.layout_known_by_front_end()) {
         constexpr if (nua_annot.changes_layout()) {
           std::cout << "The layout of " << name_of(^T) << " has been changed in this build!\n";
         } else {
           std::cout << "The layout of " << name_of(^T) << " has not been changed in this build.\n";
         }
       } else {
         std::cout << "The layout of " << name_of(^T) << " may or may not be changed in this build. It's up to the back end.\n";
       }
    } else {
       std::cout << "The layout of " << name_of(^T) << " has not been changed in this build.\n";
    }

- Wyatt

On 10/16/23 13:48, Jens Maurer wrote:
>
> On 16/10/2023 19.26, Wyatt Childers via SG7 wrote:
>> [[deprecated]] struct foo {};
>> /* implies */
>> @std::deprecated() struct foo {};
> If an implementation chooses to ignore the attribute,
> does it mean the implication doesn't hold?
>
> Jens