P1905 In-Source Mechanism to Identify Importable Headers

[wg21bot]

P1905R0 In-Source Mechanism to Identify Importable Headers (Corentin Jabot)

[jfbastien]

Let's start this in SG15 to get tooling expert's thoughts, and figure out the language stuff with that knowledge.

[brycelelbach]

Belfast 2019-11 Tooling Minutes

P1905R0 In-Source Mechanism to Identify Importable Headers

Chair: Bryce Adelstein Lelbach

Champion: Corentin Jabot

Minute Taker: Aaron Ballman

Start Review: 11-08 10:39

Start Polling: 11:07

POLL: We want an in-source mechanism for indicating but not enforcing that a header is importable.

Strongly For	Weakly For	Neutral	Weakly Against	Strongly Against
2	6	3	1	0

Attendance: 16

Authors Voting: 1

Author Position: SF

A: Not clear what diagnostics are expected, concerns about how this impacts implementations that don't support implicit modules.

That has consensus.

POLL: We want an in-source mechanism for indicating but not enforcing that a header is NOT importable.

Strongly For	Weakly For	Neutral	Weakly Against	Strongly Against
1	10	2	0	0

Attendance: 16

Authors Voting: 1

Author Position: SF

That has unanimous consent.

POLL: Prioritize scannability by tools in the design of a in-source mechanism that indicates whether headers are importable.

Strongly For	Weakly For	Neutral	Weakly Against	Strongly Against
8	5	1	0	0

Attendance: 16

Authors Voting: 1

Author Position: SF

That has unanimous consent.

POLL: Explore an in-source mechanism for enforcing that a header must not be imported.

Strongly For	Weakly For	Neutral	Weakly Against	Strongly Against
2	9	3	0	0

Attendance: 15

Authors Voting: 1

Author Position: F

That has unanimous consent.

POLL: Explore an in-source mechanism for enforcing that a header must only be imported.

Strongly For	Weakly For	Neutral	Weakly Against	Strongly Against
1	6	5	2	0

Attendance: 15

Authors Voting: 1

Author Position: A

That has weak consensus.

End: 11:30

CONSENSUS: Bring a revision of P1905R0 (In-Source Mechanism to Identify Importable Headers), with the guidance below, to the Tooling Study Group.

• Retain the existing proposed in-source mechanism for indicating but not enforcing that a header is importable.
• Add an in-source mechanism for indicating but not enforcing that a header is NOT importable.
• Explore an in-source mechanism for enforcing that a header must not be imported.
• Explore an in-source mechanism for enforcing that a header must only be imported.
• Prioritize scannability by tools in the design of a in-source mechanism that indicates whether headers are importable.
• Make your design intent regarding scannability clear in the next revision; make it clear what's allowed to come before the #pragma.
• Prior to this paper returning, please get review from EWG, EWGI, or SG2, and also specifically Michael Spencer, GDR, and Nathan Sidwell.

[brycelelbach]

This is ready for an EWGI and Modules Study Group pass. Note that if we don't want this in the International Standard, we still might want to suggest it in the Modules Ecosystem TR.

[erichkeane]

EWGI in Prague

The problem that P1905 is attempting to solve is worth solving.

SF	F	N	A	SA
4	6	0	0	0

The proper ship vehicle for this is the C++ Standard.

SF	F	N	A	SA
1	0	3	5	1

The proper ship vehicle for this is SG15 Ecosystem TR.

SF	F	N	A	SA
2	3	3	0	0

// The following polls taken in the context of a SG15 TR
We prefer the #pragma spelling listed as the choice in P1905.

SF	F	N	A	SA
4	6	1	0	0

We prefer an attribute spelling to solve the problem from P1905.

SF	F	N	A	SA
0	0	1	7	1

EWGI Believes that the Modules Ecosystem TR is the proper ship vehicle, and suggests a #pragma spelling.