[libc++] Should formatting `year{-99}` with `%C` produce `"-1"` or `"-01"`?

[StephanTLavavej]

N4964 [tab:time.format.spec]:

Specifier	Replacement
%C	The year divided by 100 using floored division. If the result is a single decimal digit, it is prefixed with 0.

llvm-project/libcxx/test/std/time/time.syn/formatter.year.pass.cpp

Lines 271 to 272 in f1200ca

	check(SV("%C='-1'\t%y='99'\t%Y='-0099'\t\n"), fmt, std::chrono::year{-99});
	check(SV("%C='-1'\t%y='09'\t%Y='-0009'\t\n"), fmt, std::chrono::year{-9});

For year{-99} and year{-9}, this test expects %C='-1', but MSVC's STL emits %C='-01'. This Standardese isn't exactly a paragon of crystal clarity, but I think this might be a bug in libc++. (Otherwise, it's a bug in MSVC's STL that we'll need to fix.)

[frederick-vs-ja]

I've mailed LWG Chair to report this. In either case, it would be clearer to say "in range [-9, 9]" (or [0, 9]) instead of "a single decimal digit".

[jwakely]

It seems pretty obvious to me that -1 is a single digit. The minus sign is not a digit.

Libstdc++ adds the 0 (let's ignore that we print -00 instead of -01 though).

[mordante]

I'm quite sure this point was raised before by somebody working on MSVC STL.

To me it is less obvious, the result is a minus sign and a digit, not a single digit.
It would be clearer if the wording would be
The year divided by 100 using floored division. If the result contains exactly one decimal digit, it is prefixed with 0.

I've no objection to adjust libc++'s implementation to match MSVC STL's and libstdc++'s implementation.
I see Daniel already created LWG-4022 so I'll wait to see the reaction on the reflector on Monday.