LWG3639: Handling of fill character width is underspecified in std::format

[tahonermann]

SG16 needs to clarify the intended design for https://cplusplus.github.io/LWG/issue3639.

 20.20.2.2 [format.string.std] doesn't specify if implementations should consider the estimated width of the fill character when substituting it into the formatted results.

For example:

    auto s = std::format("{:🤡>10}", 42);

"🤡" (U+1F921) is a single code point but its estimated display width is two.

There are at least three possible resolutions:

    1. s == "🤡🤡🤡🤡42": use the estimated display width, correctly displayed on compatible terminals.

    2. s == "🤡🤡🤡🤡🤡🤡🤡🤡42": assume the display width of 1, incorrectly displayed.

    3. Require the fill character to have the estimated width of 1.

[tahonermann]

SG16 discussed (a draft of) P2572R0 at its 2022-06-08 telecon. The following poll was taken:

• Poll 1: Revise D2572R0 "std::format() fill character allowances" as discussed,
  and forward the paper so revised to LEWG as the recommended resolution of LWG3576
  and LWG3639.

  • Attendance: 8 (2 abstention)

    SF	F	N	A	SA
    3	3	0	0	0

  • Consensus in favor.

The changes in the published P2572R0 revision reflect the SG16 requested changes.

Removing the SG16 label; this is ready for LWG pending approval of issue #1260 and (a revision of) P2572 by LEWG.

[frederick-vs-ja]

This should have been fixed by cplusplus/draft#6144.