Walter suggested adding something like this to the end of [class.compare.default]/1 to indicate that in the case that the implementation provides a hidden friend declaration of == that it's actually hidden (i.e. the implementation is not allowed to additionally provide a non-member declaration). Should be editorial:

In the latter case, there shall be no declaration or redeclaration of the == operator function outside C's member-specification [Note: The intent is that a two-argument defaulted operator function is to be found via argument-dependent lookup only. - end note]

@brevzin: Added a note; see the commit for details.

@CaseyCarter I assigned this to you thinking that you had done a complete review of the changes - was that a mistake? Just trying to make sure a "member" does a complete review here... Please let me know if I should reassign this to someone else. Thanks!

I assigned this to you thinking that you had done a complete review of the changes - was that a mistake?

I have not made a proper review, but I did click "Start review" instead of "Add single comment", presumably prompting you to assign the PR to me. I plan to do a proper review this evening as an object lesson to remind me to choose my buttons more carefully.

Otherwise, we're saying that such a friend function can only be found [and not, for example, be eaten, be painted, or be shot ☺︎] via ADL.

As a writer of generic libraries, I would be unsurprised if ADL were to eat, paint, or shoot my functions.

@burblebee, I think maybe anyone making a review comment gets assigned as "reviewer". In any case, I'd expect a "full" reviewer to approve after all the fixes have been applied.

@burblebee, I think maybe anyone making a review comment gets assigned as "reviewer". In any case, I'd expect a "full" reviewer to approve after all the fixes have been applied.

Yes, I'm often thrown off by this, which is why I've resorted to assigning reviewers, so that I know if someone is going to review the PR, and to make sure that someone knows to do a complete review (and approve, suggest changes, etc.). There are too many PRs for all of us (or Richard) to review each one; this spreads the load.

@CaseyCarter many thanks for the full review :)