Fix Windows CUDA builds, maybe

[NicolasHug]

No description provided.

[pytorch-bot]

🔗 Helpful Links

🧪 See artifacts and rendered test results at hud.pytorch.org/pr/meta-pytorch/torchcodec/1468

• 📄 Preview Python docs built from this PR

Note: Links to docs will display an error until the docs builds have been completed.

This comment was automatically generated by Dr. CI and updates every 15 minutes.

[sh202603]

Hi @NicolasHug,

Thanks for the quick follow-up here.

Verified locally on the same Windows source-build setup that originally needed the /Zc:preprocessor env-var workaround (torch 2.11.0+cu130 / CUDA 13.2 / MSVC 14.44 / Win11).

Before this PR I had to set both CXXFLAGS=/Zc:preprocessor and CUDAFLAGS=-Xcompiler=/Zc:preprocessor before uv pip install -e ../torchcodec --no-build-isolation, otherwise the CCCL headers pulled in by color_conversion.cu failed with error C1189. With 54c4091e I built cleanly with both env vars unset — the COMPILE_LANGUAGE:CUDA-gated option is sufficient on its own, and the .cpp side does not appear to need the flag at this revision either.

So the "maybe" can be promoted to "yes" at least for this configuration. For other Windows source-build users this is a strict simplification — no need to set the two /Zc:preprocessor env vars anymore!