pypy failure on master

Not sure what’s happening but the failure persists after I rerun the job. ping @rth maybe? See https://circleci.com/gh/scikit-learn/scikit-learn/43934

Installing collected packages: scikit-learn
  Running setup.py develop for scikit-learn
    Running command /root/project/pypy-env/bin/pypy3 -c "import setuptools, tokenize;__file__='/root/project/setup.py';f=getattr(tokenize, 'open', open)(__file__);code=f.read().replace('\r\n', '\n');f.close();exec(compile(code, __file__, 'exec'))" develop --no-deps
    Partial import of sklearn during the build process.
    Traceback (most recent call last):
      File "<string>", line 1, in <module>
      File "/root/project/setup.py", line 243, in <module>
        setup_package()
      File "/root/project/setup.py", line 216, in setup_package
        numpy_status = get_numpy_status()
      File "/root/project/setup.py", line 149, in get_numpy_status
        import numpy
      File "/root/project/pypy-env/site-packages/numpy/__init__.py", line 142, in <module>
        from . import core
      File "/root/project/pypy-env/site-packages/numpy/core/__init__.py", line 16, in <module>
        from . import multiarray
      File "/root/project/pypy-env/site-packages/numpy/core/multiarray.py", line 44, in <module>
        arange.__module__ = 'numpy'
    AttributeError: readonly attribute '__module__'

[qinhanmin2014]

rerun the job but still fails, reopen

[rth]

Do we know why this happens now?

Yes, it turns out I was trying to build numpy 1.5 (from 2010) with PyPy… Instead of 1.15 – makes sense it didn’t work and was attempting to run 2to3 (presumably to get Python 3.1 compatible code) 😃

Should be fixed in https://github.com/scikit-learn/scikit-learn/pull/13018