Mailing list review 22-02-2021 - 29-03-2021

• default_delete need to be made constexpr too
• Why are non member swap/ comparison are not constexpr?
• The wording should instruct editors to make the declaration constexpr too
• Consider having a single feature test macro

Low participation: Update the paper for a second round of Mailing review

P2273R1 Making std::unique_ptr constexpr (Andreas Fertig)

2021-05-25 Library Evolution Telecon

P2273R1: constexpr unique_ptr

2021-05-25 Library Evolution Telecon Minutes

Chair: Nevin Liber

Champion: Andreas Fertig

Minute Taker: Ben Craig

Summary

Discussion on whether or not the ordered comparison operators between two unique_ptrs could, and if so, should be made constexpr. Poll results are meh.

Discussion took a tangent on whether or not the standard should mark functions constexpr knowing that they cannot be used as such. It was pointed out that if there are no values which can be used in a constexpr context, the standard already says IFNDR.

Similar discussion on if make_shared_for_overwrite could be made constexpr (as default initialized values are uninitialized for some types). As the paper already proposes this, the author will check to see if it is currently allowed.

Finally, a brief discussion on whether or not we want shared_ptr, make_shared, etc. to be constexpr. There is weak consensus for this, with the dissenters feeling that it isn’t worth our limited time and resources to do so.

Outcome

LEWG is in favor of making unique_ptr constexpr, very neutral on making the ordered comparisons between two unique_ptrs constexpr, and weak consensus for making the shared_ptr family constexpr.

On 2021-05-25, we reviewed P2273R1, which proposed making std::unique_ptr constexpr.

Library Evolution had the following opinions after the review:

• We were in favor of making unique_ptr constexpr, which P2273 does.
• We were neutral on making ordered comparisons between two unique_ptr constexpr. P2273 does NOT make ordered comparisons between unique_ptr constexpr, it only makes comparisons between a unique_ptr and a nullptr constexpr.
• We were weakly in favor of exploring making shared_ptr constexpr; that's outside of the scope of what P2273 is proposing.

During the review, we asked the author to verify that making make_unique_for_overwrite constexpr is implementable, as P2273
proposes making it constexpr. The author informed me this week that he has looked into this. He believes constexpr
make_unique_for_overwrite is implementable, as we adopted P1331R2 (Permitting trivial default initialization in constexpr contexts). Here is a minimal implementation of it.

There is implementation experience for the entire P2273 proposal in a fork of libc++:

The proposal has wording, which is quite straightforward. Basically, it's just adding constexpr in synopses and adding a feature test macro. I've asked two wordsmiths to review it, and they've provided some minor feedback and have not identified any design issues.

Library Evolution did not ask for any changes to P2273 in our last review, and we affirmed that we wanted to pursue it. We asked the author to confirm that something contained in the paper was implementable, which he has done.

As such, I determined that another telecon review of P2273 is probably not the best use of Library Evolution time.

Instead, I presented the above information to Library Evolution via mailing list and asked who would support taking an electronic poll advancing P2273 to LWG for C++23 as a B2 (improvement) item in the next Library Evolution electronic polling period. 8 people explicitly spoke in favor of this on the mailing list; none spoke against it.

Therefore, I have determined that we have consensus to take an electronic poll advancing P2273 to LWG for C++23 as a B2 (improvement) item in the next Library Evolution electronic polling period

P2273R2 Making std::unique_ptr constexpr (Andreas Fertig)

The paper is being placed on the Summer 2021 electronic ballot.

2021 Summer Library Evolution Polls

P2435: 2021 Summer Library Evolution Poll Outcomes

POLL 5: Send P2273R2 (Making unique_ptr constexpr) to Library Working Group for C++23, classified as an improvement of an existing feature (P0592R4 bucket 2 item).

Outcome: Unanimous consensus in favor.

This was seen and effectively approved by LWG with a small addition that needs LEWG approval

https://wiki.edg.com/bin/view/Wg21telecons2021/P2273-20211105

poll: add P2273r3 to the C++23 working paper with the changes (reviewed by CC and JW)

JW: action item for me to send the version to LEWG.

P2273R3 Making std::unique_ptr constexpr (Andreas Fertig)

LEWG action resolved - November 8. No objections. Short thread starting with https://lists.isocpp.org/lib-ext/2021/11/21308.php

Ready for plenary.

Approved by WG21 plenary 2022-02-07