On Thu, 2023-05-04 at 21:57 +0200, Jens Maurer via Std-Proposals wrote:
>
>
> On 04/05/2023 14.27, Lauri Vasama via Std-Proposals wrote:
> > That's right. My intention is not at all to try to force
> > Microsoft's hand. Their hesitation to implement
> > [[no_unique_address]] serves to demonstrate the problem with the
> > ignorability of attributes, though of course there is no problem if
> > attributes never affect program behaviour, but that particular cat
> > is now out of the bag. I can't speak for Microsoft of course, but i
> > believe they would not have the same objection to implementing
> > no_unique_address if its use was ill formed in compiler versions or
> > language modes where it is not supported.
>
> Note carefully that the "ignorability" rule says that you must ignore
> all
> instances of an attribute in the entire program. Selectively
> ignoring
> a few instances (e.g. per translation unit) is not supported.
>
> Thus, anyone feeling that they want to make the ABI breakage of
> [[no_unique_address]] visible loudly might want to start mangling
> classes having a [[no_unique_address]] member differently. That
> way, you get linker errors if you pass one of those classes with
> inconsistent [[no_unique_address]] treatment across function call
> boundaries. That's not a complete approach because it doesn't
> detect "sizeof(T)" differences, but it's a start.

How do you do that when the class is incomplete?

Cheers,
Lénárd