P2758 R3 Emitting messages at compile time

[wg21bot]

P2758R0 Emitting messages at compile time (Barry Revzin)

[tahonermann]

Since this paper has character encoding considerations, I'm going to schedule it for SG16 review.

[michaelbprice]

Relevant to https://isocpp.org/files/papers/P2758R0.html#improving-static_assert, see the previously discussed paper (which I authored) https://open-std.org/JTC1/SC22/WG21/docs/papers/2015/n4433.html and the corresponding EWG notes on the wiki from the Lenexa meeting.

Of note, it already contains potential wording, but was never reviewed by a Core expert and likely needs updating since we've done lots of character encoding work since then.

[erichkeane]

This was discussed in EWG during the February 8, 2023 afternoon session in Issaquah. The following polls were taken:

EWG would like to encourage more work on P2741R0 (static_assert-with-constexpr-message), solicit feedback from SG16, and see it again with core expert reviewed wording.

SF	F	N	A	SA
8	8	0	1	0

Result: Consensus

EWG would like to encourage more work on P2758 in the direction of constexpr_print_str/constexpr_error_str.

SF	F	N	A	SA
11	6	1	0	0

Result: Consensus

EWG ALSO would like to encourage more work on this topic that can print multiple errors per constant evaluation, but still result in a failed TU.	SF	F	N	A	SA
5	10	3	1	0

Result: Consensus

EWG would like to encourage more work on P2758 in the direction of a constexpr std::format.	SF	F	N	A	SA
13	4	1	0	0

Result: Consensus

[cor3ntin]

@tahonermann we can remove the SG16 tag from thar, right? SG16 guidance was given in #1434

[tahonermann]

@cor3ntin, the guidance provided in #1434 is clearly applicable, but I would prefer to see an updated paper that removes the duplication with P2741 and that has wording that implements that guidance first so that SG16 can definitively approve forwarding a specific revision.

[cor3ntin]

@tahonermann you are right, but i think we lost a needs-revision on this paper then. @jfbastien you might want to ask barry whether he actually wants to present in Varna.

[jfbastien]

I checked with @brevzin, he says:

once SG16 is happy with arbitrary object, the rest of my paper just follows suit I think. Just need to drop the part that Corentin is covering.

So let's mark it as needing revision and track @cor3ntin 's paper first. Once @cor3ntin 's paper proceeds (or dies!) we can move this one accordingly.

[brevzin]

I checked with @brevzin, he says:

once SG16 is happy with arbitrary object, the rest of my paper just follows suit I think. Just need to drop the part that Corentin is covering.

So let's mark it as needing revision and track @cor3ntin 's paper first. Once @cor3ntin 's paper proceeds we can move this one accordingly.

Ahem. FTFY.

[tahonermann]

I'm going to retain the SG16 label for now. EWG can, and should, proceed with any review desired of course, but I want to run a new revision by SG16 since we haven't discussed the parts of the paper that are not covered by P2741 (user-generated static_assert messages).

[wg21bot]

P2758R1 Emitting messages at compile time (Barry Revzin, )

[wg21bot]

P2758R2 Emitting messages at compile time (Barry Revzin)

[hanickadot]

EWG discussed P2758R2 at Tokyo meeting on afternoon Thursday.

P2758R2 Emitting messages at compile time: EWG wants to have distinctive functions std::constexpr_error (placeholder) and std::constexpr_fatal_error (placeholder2). With distinct recommended behaviour.

SF	F	N	A	SA
1	6	7	4	0

No consensus.

[michaelbprice]

I'm confused by the update from the Tokyo meeting.

What is the question for the straw poll? Was the straw poll taken with the question of advancing the paper through to Core? Or was the straw poll taken on the question of having std::constexpr_error and std::constexpr_fatal_error? The addition of the needs-revision label makes me think the former, but the presence of the "wants" in the description of the poll makes me think the latter.

[tahonermann]

@michaelbprice, the poll was for design direction, not for forwarding. This paper has not yet been seen by LEWG (and honestly, that poll seems like more of a question for LEWG than for EWG). SG16 will be reviewing this paper tomorrow if you are interested. All that being said, these questions are more appropriate for the mailing lists.

[tahonermann]

SG16 reviewed P2758R2 during its 2024-04-10 meeting. No polls were taken. The following feedback was given to the author.

• The paper could use an overall update to reflect the adoption of P2741R3 (user-generated static_assert messages).
• Support for an annotation that could be used to suppress diagnostics produced from the proposed interfaces would be useful.
• Requests were made for each of the error, warning, and print interfaces to accept a tag that could be used to suppress diagnostics or elevate a warning to an error.
• The proposed wording needs an update to specify that the provided message will be interpreted as a sequence of characters in the ordinary literal encoding.
• The design and proposed wording need updates to specify which characters are allowed in tags. The set of characters should be limited to those from the basic literal character set and should exclude characters that might be challenging to pass in a command line argument including but not limited to quotes, semicolon, and non-printable characters.
• The proposed wording will need an update to [intro.compliance.general] to specify the effect that a call to print_error_str() has on translation success.
• These interfaces will require compiler support in the form of a builtin function. Consideration should be given to whether these facilities should include a core language feature that could then power a more std::format()-like interface.
• Consideration should be given to whether the tag argument should be required to be a string literal so that it would be possible to audit the set of tags used in a program.

Adding the needs-revision tag, SG16 looks forward to reviewing this paper again.

[wg21bot]

P2758R3 Emitting messages at compile time (Barry Revzin)

[jfbastien]

Seen in EWG in St Louis:

Poll: P2758R3 — Emitting messages at compile time, forward to SG15 for tooling consideration of tags, then forward to LEWG and CWG for inclusion in C++26, adding a prohibition for tags being empty.

| SF | F | N | A | SA | | 7 | 13 | 1 | 0 | 0 |

[inbal2l]

@tahonermann @Bigcheese - please add "ready-for-lewg-meeting-review" tag once you're done.

[Bigcheese]

POLL: Tooling is happy with the tag support in P2758r3 and has no blocking concerns with the paper.	SF	F	N	A	SA
4	2	0	0	0

[tahonermann]

I confirmed that all of the SG16 concerns enumerated in https://github.com/cplusplus/papers/issues/1445#issuecomment-2080315619 have been addressed in P2758R3. SG16 has not yet polled to forward the paper. If the paper does not appear in the plenary motions in Wrocław, then I'll schedule a poll in an SG16 meeting before the Austria meeting. I see no reason for any WG to await further review from SG16 before proceeding.

[jensmaurer]

CWG 2024-11-22 in Wroclaw: Reviewed; needs updates.

There are concerns about the interaction with trial evaluation for static const variable initialization.