stillwater-sc / universal / 24585301926

84%

master: 84%

push

github

fix(tools): guard <bit> header include for MSVC without /Zc:__cplusplus (#711)

* fix(native): guard <bit> include for MSVC without /Zc:__cplusplus

The <bit> include in utility/bit_cast.hpp was gated behind
__cplusplus >= 202002L, which fails on MSVC without /Zc:__cplusplus
(MSVC reports 199711L regardless of /std:c++20). Meanwhile,
<type_traits> transitively includes <version> on MSVC, defining
__cpp_lib_bit_cast -- so the macro fired, BIT_CAST expanded to
`using std::bit_cast`, but <bit> was never included and std::bit_cast
was undeclared. This broke downstream consumers (e.g., mp-dsp-python)
that include Universal headers without Universal's CMake flags.

Fix: drop the __cplusplus pre-check; __has_include(<bit>) already
returns false when the header is unavailable, making the extra guard
redundant and harmful.

Additionally, replace all 14 direct std::bit_cast<To, From>(v) calls
in native/extract_fields.hpp, native/set_fields.hpp, and
native/attributes.hpp with sw::bit_cast<To>(v). The sw::bit_cast
wrapper is always available after bit_cast.hpp is included and has
identical constexpr behavior, eliminating the fragile coupling to
<bit> being included upstream. Same treatment for 2 test files for
consistency.

Resolves #710

Co-Authored-By: Claude Opus 4.6 (1M context) <noreply@anthropic.com>

* fix(fixpnt): remove noisy #pragma message from mod_complex_mul

The #pragma message fired on every build warning that the complex
multiplication regression suite is disabled. Replace with a plain
comment explaining why MANUAL_TESTING is intentionally ON (known
rounding failures for small fixpnts where individual terms round
down but the sum should round up).

---------

Co-authored-by: Claude Opus 4.6 (1M context) <noreply@anthropic.com>

8 of 8 new or added lines in 2 files covered. (100.0%)

44839 of 53143 relevant lines covered (84.37%)

6525296.87 hits per line