C++26 freestanding feature-test macros

[frederick-vs-ja]

Fixes #3793. Fixes #3794. Fixes #3795.

(But we may need to reopen these issues and #2914 in future once freestanding mode is added. See #1289.)

It seems to me that hosted implementations are required to provide all of these macros (see "Rejected alternatives" in WG21-P2198R7). I think it makes more sense to define them in older modes when the corresponding feature/header is available.

Possibly related question: should we leave feature-test macros for non-core components not defined when _ENFORCE_ONLY_CORE_HEADERS is defined?

[frederick-vs-ja]

CC-ing @ben-craig.

[ben-craig]

(But we may need to reopen these issues and #2914 in future once freestanding mode is added. See #1289.)

You may want to drop a static assert somewhere checking __STDC_HOSTED__ so that these issues are noticed at that time.

It seems to me that hosted implementations are required to provide all of these macros (see "Rejected alternatives" in WG21-P2198R7). I think it makes more sense to define them in older modes when the corresponding feature/header is available.

The standard doesn't really talk about older modes, and I didn't do anything to audit which modes / standards would make sense. Sure, all the stuff in C++03 from the <utility> header is freestanding capable, but we've added stuff to <utility> since then. That could make backporting __cpp_lib_freestanding_utility weird.

[frederick-vs-ja]

Sure, all the stuff in C++03 from the <utility> header is freestanding capable, but we've added stuff to <utility> since then. That could make backporting __cpp_lib_freestanding_utility weird.

The value 202306L may be misleading for some macros, but pehaps fine for others.

Macro	Introduction of related portions	Latest change of related portions at the time of 202306L	Implemented in MSVC STL
__cpp_lib_freestanding_char_traits	C++98	C++20 (char8_t, constexpr)	Yes
__cpp_lib_freestanding_charconv	C++17	C++26 (operator bool)	No
__cpp_lib_freestanding_cstdlib	C++98	C++23 (constexpr)	No
__cpp_lib_freestanding_cstring	C++98	(none?)	Yes
__cpp_lib_freestanding_cwchar	C++98	(none?)	Yes
__cpp_lib_freestanding_errc	C++11	(none?)	Yes
__cpp_lib_freestanding_functional	C++98	C++26 (function_ref etc.)	No
__cpp_lib_freestanding_iterator	C++98	C++23 (basic_const_iterator etc.)	Yes
__cpp_lib_freestanding_memory	C++98	C++23 (start_lifetime_as)	No
__cpp_lib_freestanding_operator_new	C++98	C++17 (aligned overloads)	Yes (can be opted-out by /Zc:alignedNew-)
__cpp_lib_freestanding_ranges	C++20	C++23 (various range adaptors)	Yes
__cpp_lib_freestanding_ratio	C++11	C++26 (optional additions)	N/A (or Yes?)
__cpp_lib_freestanding_tuple	C++11	C++23 (tuple-like)	Yes
__cpp_lib_freestanding_utility	C++98	C++23 (tuple-like)	Yes

I don't actually know what __cpp_lib_freestanding_feature_test_macros intentionally covers.

[ben-craig]

From the paper:

In order for users of freestanding implementations to be able to detect extensions of freestanding, the users need a way of distinguishing the C++20 macro requirements from an accurate expression of extension. This paper will provide a new macro, __cpp_lib_freestanding_feature_test_macros, so that users can distinguish between these cases. When __cpp_lib_freestanding_feature_test_macros is present, users will know that the C++20 feature-test macros aren't lying.

Example...

#if (__STDC_HOSTED__ || \
     (defined(__cpp_lib_freestanding_feature_test_macros) && \
      __cpp_lib_freestanding_feature_test_macros >= 202112))
#define TRUTHFUL_MACROS 1
#else
#define TRUTHFUL_MACROS 0
#endif

// see if filesystem happens to be supported on this possibly freestanding environment
#if TRUTHFUL_MACROS && defined(__cpp_lib_filesystem) && __cpp_lib_filesystem >= 201703L
// hosted and freestanding extension success path
#else
// fallback path
#endif

So that macro should only be important for actual freestanding implementations.

[StephanTLavavej]

Because these feature-test macros are essentially defined unconditionally, we can accept this as a special exception to our "no C++26 PRs yet" policy.

[StephanTLavavej]

Thanks! After fixing a copy-paste typo, looks perfect! 😻

[StephanTLavavej]

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

[StephanTLavavej]

Thanks for implementing all of these feature-test macros! 🎉 ⚙️ 🚀