On Thu, Apr 16, 2020 at 7:46 PM Daniel Shahaf <d...@daniel.shahaf.name> wrote: > > Stefan Sperling wrote on Thu, 16 Apr 2020 15:31 +00:00: > > I don't know the Python ecosystem well enough to have an informed > > opinion on this. I would support whatever the people actively > > working on Python bindings would want to do. > > This is not only about the bindings but also about the "core" uses of > Python (autogen.sh, Windows build, test suite).
I'm quickly going to point out (as I did a couple of days ago on IRC) that on Windows there are a lot of test failures if you run the test suite with Python 3 (69 failures IIRC). I'll try to post some more details (the actual fails.log) later tonight, so maybe some people can help with tracking down / squashing those problems. I'm guessing a lot of those failures have the same underlying cause (like encoding issues or eol issues). Perhaps that needs at least a mention in some "known issues" section. For normal users this doesn't matter much of course, and judging from the amount of Windows sigs for 1.14 I doubt even one single other person beside me is building and running tests on Windows :-). So I ended up running the test suite on Windows with Python 2.7. I did build the swig-python bindings with Python 3.8.2, but in the end I had some problems getting the swig-python tests to run (probably a local problem), so I had to punt and omitted the swig-python bindings from my tests / sig. -- Johan
