Prepare for adding compile-time checks in <format>

[vitaut]

This PR prepares the formatting API for adding compile-time checks. Specifically, it adds a format string type, _Basic_format_string and updates formatting functions to use it instead of basic_string_view. The constructor of _Basic_format_string is constexpr instead of consteval because of a MSVC issue: https://godbolt.org/z/aMrbe6xzn.

It implements most of the first part of P2216 and is relative to #1874.

[SuperWig]

Am I blind or are the accidental removals of

template <class... _Types>

from a couple functions not in the diff?

[vitaut]

Am I blind or are the accidental removals of ... from a couple functions not in the diff?

I'm not sure what you are referring to, please comment on specific line. All the _Types parametrization is still there AFAICS.

[SuperWig]

Oh nvm you changed concrete types, that's why.
But this is why the tests are failing

STL/stl/inc/format

Lines 2955 to 2961 in f0226c2

	_NODISCARD inline string vformat(const _Fmt_string<_Types...> _Fmt, const format_args _Args) {
	string _Str;
	_Str.reserve(_Fmt._Str.size() + _Args._Estimate_required_capacity());
	_STD vformat_to(_STD back_inserter(_Str), _Fmt._Str, _Args);
	return _Str;
	}

_Types... is?

[vitaut]

the tests are failing

Yeah, I accidentally changed wrong overloads. Fixed now.

[barcharcraz]

This needs a rebase after the debloating changes were merged.

[vitaut]

Rebased.

[barcharcraz]

Let's mark this PR WIP until the feature is done. We're happy to review it incrementally, but don't want to merge a partial implementation of the paper (this is because things in our main branch will flow into future Visual Studio releases, and we'd prefer not to release a partial feature even if it doesn't break anything. That way we have more ability to change implementation strategies partway through).

[barcharcraz]

I know we merged the "other" part of P2216 separately, but that was really it's own feature (and it would have been nice if it had been it's own paper).

[fsb4000]

@barcharcraz I saw in https://github.com/microsoft/STL/projects/6 that need to cite DevCom bug. I found it's easy for me.
https://developercommunity.visualstudio.com/t/consteval-constructor-and-C7595/1453392
We could do it in April :(

[barcharcraz]

@barcharcraz I saw in https://github.com/microsoft/STL/projects/6 that need to cite DevCom bug. I found it's easy for me.
https://developercommunity.visualstudio.com/t/consteval-constructor-and-C7595/1453392
We could do it in April :(

thanks, that looks like it is indeed the correct bug. I'll make sure the internal version of that bug (VSO-1344623) is categorized correctly so folks know stl features depend on it.

[vitaut]

Closing as I don't plan to work on this PR.

[barcharcraz]

Closing as I don't plan to work on this PR.

Thanks, I'll pick it up

[fsb4000]

@barcharcraz https://developercommunity.visualstudio.com/t/consteval-constructor-and-C7595/ is fixed.
"A fix for this issue has been internally implemented and is being prepared for release. We'll update you once it becomes available for download."