On Sun, 29 Jan 2023 at 22:51, Giuseppe D'Angelo via Std-Proposals
<std-proposals_at_[hidden]> wrote:
>
> Il 29/01/23 21:49, Ville Voutilainen via Std-Proposals ha scritto:
> > When a proposal says "This is a pure library extension.", I take that
> > to mean that it requires no compiler magic,
> > not that it's an extension confined purely in the library sections of
> > the standard. The latter meaning is useless,
> > I can see that from the wording at a glance, but I think it would be
> > rather more helpful to say that the proposal
> > is about adding a trait that needs compiler help.
>
> Apologies for that, I'll clarify. What I meant was that I am not
> proposing to touch the language in any way, nor to change any other
> existing parts of the library.

"Impact on the standard:

This proposal is a library-only change proposing a compiler-backed trait."

and I'm a happy camper. :)

So, we're looking at an implementation-specific result here,
considering the mention of PMFs on the itanium ABI,
and other implementations may produce different results. That is, the
trait isn't detecting language semantics,
but rather implementation semantics.

That gives me naming suggestions.. I mean, we're saying "is this type
suitable for optimization with zero-fill on this implementation".
Anyway, naming aside, that's what this trait tells us, and different
implementations will have different answers, and that's intentional.
I suppose that could be emphasized a little in the proposal.