Closed henryiii closed 3 months ago
I'd like to make a follow-up dropping 2.7. Better if an SDist build is made, and better for dropping manylinux1. We could keep manylinux1 if we forced only binaries for cmake, I think, as it would then get the last manylinux1 cmake. But better just do update, I think.
We could keep manylinux1 if we forced only binaries for cmake, I think, as it would then get the last manylinux1 cmake. But better just do update, I think.
Should we build on manylinux2010 (as we do with CMake) ? Maybe it's not worth it given the low number of downloads for manylinux 2010 (glibc 2.12) vs 2014 (glibc 2.17) for ninja the last 30 days (I removed some rows in there):
libc_version | cum_percent | download_count |
---|---|---|
2.35 | 32.88 % | 2,359,423 |
2.34 | 33.79 % | 74,313 |
2.31 | 72.24 % | 3,154,519 |
2.28 | 76.20 % | 324,935 |
2.27 | 80.20 % | 327,986 |
2.26 | 96.37 % | 1,326,990 |
2.24 | 96.39 % | 1,972 |
2.23 | 96.61 % | 17,466 |
2.17 | 99.95 % | 271,337 |
2.12 | 100.00 % | 4,214 |
2.9 | 100.00 % | 52 |
2.5 | 100.00 % | 41 |
Total | 8,206,430 |
Not "worth it", but fine to be consistent.
(Also when we drop 3.6 and below, that distribution may change)
Moving to scikit-build-core.