Even though there is less occurences of it, "standard layout" without the dash makes more sense to me. But maybe there is a reason for the other one to be widely spread in the draft.

AFAIK, ISO C++ only defines "standard-layout type", "standard-layout class", "standard-layout struct" and "standard-layout union". It doesn't define the meaning of the phrase "have standard layout", but this phrase is used in N4527 §29.5[atomics.types.generic]p5&6 and §29.7[atomics.flag]p3.

I think it's safe to replace "have standard layout" with "be standard-layout structs". According to N2427, this matches the original intent.

Incidentally, the sentense "They shall each support aggregate initialization syntax." in §29.5 is also problematic: std::atomic<T> doesn't seem to be a aggregate, as it has a user-provided constructor constexpr atomic(T) noexcept. This might need to be a non-editorial issue....

@"Incidentally, the sentense "They shall each support aggregate initialization syntax." in §29.5 is also problematic: std::atomic doesn't seem to be a aggregate, as it has a user-provided constructor constexpr atomic(T) noexcept."

With list-initialization, we already support "aggregate initialization syntax" here, even when the class has a constructor. That said, the requirement seems to be totally vacuous given the synopsis.

This is LWG issue 2715: http://cplusplus.github.io/LWG/lwg-active.html#2715

See also LWG 2715 for discussion on the (meaningless) phrase "aggregate initialization syntax".

@mclow: If the LWG issue was resolved as editorial, can you please either update and reopen this issue or create a new issue on GitHub? I will merge PR #1021 for now.