On Thu, 30 Aug 2018 at 19:30, Donald Stufft <don...@stufft.io> wrote: > I find it helpful to generally not think of compatibility tags as hard “this > wheel is supported on this platform”, but more along the lines of “if I built > a wheel in the specified environment, I would get the same results”. Those > results may or may not work. So for example, if you have a pure Python wheel > that *only* works on Python 3.5+, but it produces the same output when > building the wheel on Python 3.3 (even if it doesn’t ultimately work), then a > “py3” wheel is the right tag to use. Arguably you’d even use py2.py3 since > you’d produce the same output (but in practice most don’t since that’s extra > work).
I also quite like Daniel's description of the list of supported tags as "try these, in this order - the first one you match is the best chance you have of getting something that works". It may still not work, but there's nothing that the package builder declared as being any better. The thing I most dislike about the tag system is that if pip (or any installer) misses out a particular tag combination, it gets totally ignored. I sort of wish that there was some level of wildcard matching, so that weird combinations get picked up at least *somewhere* in the priority list. But I never managed to design a workable scheme for doing that, so it may not even be possible. Paul -- Distutils-SIG mailing list -- distutils-sig@python.org To unsubscribe send an email to distutils-sig-le...@python.org https://mail.python.org/mm3/mailman3/lists/distutils-sig.python.org/ Message archived at https://mail.python.org/mm3/archives/list/distutils-sig@python.org/message/HU2K3CEZPJF7SWRMPGZ5NCULMZIB65NO/
