This concerns ML-KEM, ML-DSA and in the future, also the combined extractions.
- If possible (maybe once we have more of a handle on CI time), we should flag the need to re-extract early, or just have CI do it entirely. This would reduce some of the churn we have now, where PRs get rejected from the merge queue because the author forgot about C extraction.
- Make sure header-only extractions are header only
- Make sure C extractions are free of C++isms
This concerns ML-KEM, ML-DSA and in the future, also the combined extractions.