FTBFS w/ MSYS2 MinGW

[kmilos]

Both GCC and Clang fail w/ _mm_pow_ps error on Windows:

https://github.com/msys2/MINGW-packages/actions/runs/12086725004/job/33706625136?pr=22697#step:11:512
https://github.com/msys2/MINGW-packages/actions/runs/12086725004/job/33706625435?pr=22697#step:11:510

2.3.x was building just fine w/o patching.

Note that MSYS2 targets -march=nocona -msahf (so no SSSE3, SSE4.2, AVX...)

[lazka]

Similar code is guarded differently, compare:

• OpenColorIO/src/OpenColorIO/ops/fixedfunction/FixedFunctionOpCPU.cpp

  Lines 1792 to 1801 in 0f06f04

  	#if (_MSC_VER >= 1920) && (OCIO_USE_AVX)
  	// MSVC 2019+ has built-in _mm_pow_ps() SVML intrinsic implementation
  	// accessible through immintrin.h. Therefore precise SIMD version is available
  	// only when compiled with MSVC and AVX support.
  	template<>
  	__m128 Renderer_PQ_TO_LIN_SSE<false>::myPower(__m128 x, __m128 exp)
  	{
  	return _mm_pow_ps(x, exp);
  	}
  	#endif

• OpenColorIO/src/OpenColorIO/ops/fixedfunction/FixedFunctionOpCPU.cpp

  Lines 1847 to 1856 in 0f06f04

  	#ifdef _WIN32
  	// Only Windows compilers have built-in _mm_pow_ps() SVML intrinsic
  	// implementation, so non-fast SIMD version is available only on Windows for
  	// now.
  	template<>
  	__m128 Renderer_LIN_TO_PQ_SSE<false>::myPower(__m128 x, __m128 exp)
  	{
  	return _mm_pow_ps(x, exp);
  	}
  	#endif // _WIN32

[lazka]

I noticed that the guard is fixed as part of #2089

[lazka]

@kmilos with #2089 merged this can be closed, right?

[kmilos]

Indeed, thanks.