[intro.object] Unconditional semantic effect of `[[no_unique_address]]`

[frederick-vs-ja]

Full name of submitter (unless configured in github; will be published with the issue): Jiang An

Reference (section label): [intro.object]

Link to reflector thread (if any):

Issue description:

It's currently specified in [intro.object] p7.2 that the no_unique_address attribute unconditionally makes a subobject potentially-overlapping, even if __has_cpp_attribute(no_unique_address) is 0.

There're already some open CWG issues with [[no_unique_address]] (CWG2545 and CWG2866). If some implementation decides to ignore [[no_unique_address]], it seems better to make such an implementation (e.g. MSVC) free from these issues. For such an implementation, it's perhaps better to make [[no_unique_address]] not to bring the potential-overlapping property in. See also CWG2759.

By the way, I guess that [intro.object] p7.2 should say "a member subobject corresponding to a non-static data member" instead. But perhaps this should belong to another issue.

Suggested resolution:

Modify [intro.object] p7.2 as indicated.

• (7.2) if the value of __has_cpp_attribute(no_unique_address) ([cpp.cond]) is nonzero, a non-static data member declared with the no_unique_address attribute.

[jensmaurer]

The only currently known program-detectable semantic effect of [[no_unique_attribute]], after CWG2759, is in constant expressions, per CWG2866.

While the above change might be a way to approach CWG2866 (causing different constant evaluation behavior on different implementations), I think that's in the realm to EWG to decide on. I'll add the suggested change to CWG2866.