On Mon, 24 Mar 2025 at 15:56, Matheus Izvekov via Std-Proposals <
std-proposals_at_[hidden]> wrote:

> I don't disagree in principle with the idea of allowing more kinds of
> attributes, which currently ignorability rules out.
>
> But I think the status quo with vendor attributes is not very safe.
> Either they should be mandatory to implement, or the compiler should
> make it ill-formed
> to use an attribute it doesn't know about, which seems would be quite
> a breaking change.
>

"ill-formed" only requires a diagnostic, which can be a warning. Most
compilers already do that for unknown attributes, and have something like
-Werror=attributes to make it a hard error.