Hi,

On 20 December 2022 13:25:48 GMT, Timur Doumler via Std-Proposals <std-proposals_at_[hidden]> wrote:
>
>
>
>> On 19 Dec 2022, at 12:11, Bo Persson via Std-Proposals <std-proposals_at_[hidden]> wrote:
>> At least originally, there was a basic idea that ignoring an attribute should not affect the validity of a program. Like [[nodiscard]] and [[maybe_unused]] would affect warnings, but not the generated code.
>>
>> Adding ABI-modifying attributes seems to be totally contrary to this.
>
>That basic idea has been thrown out of the window a while ago. We now have multiple potentially-ABI-modifying attributes in the standard.
>
>[[no_unique_address]] modifies ABI: its purpose is to literally change the class layout.
>
>[[assume]] can also modify ABI in one particular edge case (https://www.open-std.org/jtc1/sc22/wg21/docs/papers/2022/p2552r1.pdf, see last code example on page 5).
>
>Cheers,
>Timur

Are the precedences confirm that it is a good idea though? [[no_unique_address]] had to be "backported" to earlier -std options in clang and gcc. The ABI mismatch exists between old and new versions of the compilers on older standards.

MSVC just continues to ignore the attribute, AFAIK at least partly to avoid ABI issues.

I think the ABI issues could have been side-stepped, if it was introduced by a non-ignorable syntax. Unfortunately there is no similarly convenient way to introduce those.

I did not know about [[assume]].

Cheers,
Lénárd