codecvt: no conversion for all char-like types of size 1.

[Igor Zhukov (fsb4000)]

Fixes #2109
Fixes #817

[Igor Zhukov (fsb4000)]

template <class>
_INLINE_VAR constexpr bool test = false;

template <>
_INLINE_VAR constexpr bool test<char> = true;

in two different cpp files gives fatal error LNK1169: one or more multiply defined symbols found with clang-cl and C++14.

Fortunately, there is no code like this in the STL.

Ordinary constexpr variables or constexpr templates variables without explicit specialization work fine.

[Igor Zhukov (fsb4000)]

Do we need remove_cv_t<_Ty> like there?:

STL/stl/inc/xtr1common

Lines 174 to 180 in 60decd0

	template
	_INLINE_VAR constexpr bool is_integral_v = _Is_any_of_v, bool, char, signed char, unsigned char,
	wchar_t,
	#ifdef __cpp_char8_t
	char8_t,
	#endif // __cpp_char8_t
	char16_t, char32_t, short, unsigned short, int, unsigned int, long, unsigned long, long long, unsigned long long>;

[Stephan T. Lavavej (StephanTLavavej)]

The type traits for the "primary type categories" like is_integral_v are required to ignore top-level cv-qualifiers (WG21-N4910 [meta.unary.cat]/2). Our internal type traits don't always remove_cv_t, but we need to keep this in mind whenever we're using them, as this is a potential source of bugs.

(_Is_any_of_v itself should not remove_cv_t as it is a generalized is_same which must care about cv-qualifiers; it's things like the "is it safe to activate the memmeow() optimization" helpers that sometimes expect usage to have already removed cv.)

[nicole mazzuca (strega-nil-ms)]

It might be nice to also check for _Is_two_byte_char_like_v, for wchar_t/char16_t?

[Igor Zhukov (fsb4000)]

Yes, there is an issue for that: #605 (comment)
but I have not yet explored what needs to be done there to speed up

[Stephan T. Lavavej (StephanTLavavej)]

I'm mirroring this to the MSVC-internal repo - please notify me if any further changes are pushed.

[Stephan T. Lavavej (StephanTLavavej)]

Thanks for this major performance improvement! 🚀 🐇 🏎️