[depr.c.macros] Missing description of `__alignof_is_defined` LWG4036

[frederick-vs-ja]

__alignas_is_defined and __alignof_is_defined are introduced in C11 (WG14 N1507, WG14 N1558) and removed in C23 (WG14 N2934).

However, C++ (11~23) has never mentioned __alignof_is_defined even though it's specified that "contents of the C++ header <stdalign.h> are the same as the C standard library header <stdalign.h>" (the wording was slightly different before C++20). There doesn't seem any C++ paper or issue for this.

Is it intended that __alignof_is_defined is not a C++ feature or is there a missing specification? In the latter case, we may still need to change Annex C for C++26.

[jwakely]

This was mentioned during the discussion of LWG 3827 but looks like it didn't make it into the issue.

It was mentioned in cplusplus/nbballot#502

It doesn't really matter now, because if it was supposed to be present in older standards, it's gone now anyway. I don't think we want a compatibility note saying that something which was never in the standard is gone now.

[jwakely]

Oh wait, they're not gone, we only deprecated them for now. Sigh.

I suppose we could add __alignof_is_defined to Annex D. I find it hard to care. Nobody should ever try to use that macro in C++.

[jensmaurer]

What's unclear about this specification in [stdalign.h.syn] ?

The contents of the C++ header <stdalign.h> are the same as the C standard library header <stdalign.h>, with the following changes: The header <stdalign.h> does not define a macro named alignas.
See also: ISO/IEC 9899:2018, 7.15

Our normative reference is to C17, so if this has a macro in <stdalign.h>, then C++23 has this macro as well.

What exactly is the problem? Do you want index entries for the macros?

[jwakely]

We explicitly deprecated __alignas_is_defined in C++23, it would be silly to leave __alignof_is_defined there, un-deprecated. It would conflict with the clear direction of an NB comment.

[jensmaurer]

Ah, ok, we're simply missing an entry for __alignof_is_defined in [depr.c.macros].

@jwakely , do you feel we should do that editorially, or is that an LWG matter? Strictly speaking, it's normatively deprecating a macro whose status quo is that it's not deprecated.

[jwakely]

Yeah I think we should probably do it as an LWG issue. I don't know why it was never listed in the contents of <cstdalign>, if it had been there, we'd have deprecated it as part of LWG 3827.

[frederick-vs-ja]

Hmm... I'm tried to submit an LWG issue but not sure whether it is wanted (it's possibly too late). Since C++26 will probably refer to C23 and thus should get __alignas_is_defined/__alignof_is_defined removed.

That's now LWG4036.

[frederick-vs-ja]

Should be fixed by #6915.