That all makes sense to me! On Thu, Oct 18, 2018 at 8:24 PM, Elana Hashman <[email protected]> wrote: > I'm trying to fix a regression introduced by a PR to auditwheel[1] that > attempted to add support for repairing wheels that are not Python extensions > but still contain binary dependencies (e.g. [2]). This seems like a > reasonable thing to support so I'd like to avoid a hard revert of the PR if > possible, but I'm not 100% sure how to proceed. I've fixed the first bug > locally but that only seems to have revealed more. > > Since such a wheel is not a Python extension, it is (perhaps incorrectly) > identified as a purelib when built. I imagine that we should avoid putting > binary things into purelib, so it sounds like I should repair such a wheel > to set Root-Is-Purelib to false, which should cause it to install into > platlib if I'm reading the spec[3] right? > > As for platform compatibility flags for such a wheel, I will set it to > py3-none-manylinux1 (or whatever Python it needs). > > Does this sound reasonable? I'm hoping if wheels like this start to > circulate it won't catch anyone by surprise. > > Cheers, > > - e > > > [1]: https://github.com/pypa/auditwheel/issues/107 > [2]: > https://github.com/pypa/auditwheel/pull/95/files#diff-01628eaed8ff1de796a126b6e752ea91 > [3]: https://www.python.org/dev/peps/pep-0427/#file-contents > _______________________________________________ > Wheel-builders mailing list > [email protected] > https://mail.python.org/mailman/listinfo/wheel-builders
-- Nathaniel J. Smith -- https://vorpus.org _______________________________________________ Wheel-builders mailing list [email protected] https://mail.python.org/mailman/listinfo/wheel-builders
