WIP: test against Python 3.8

[grlee77]

This PR is an attempt to test against Python 3.8.

[grlee77]

Python 3.8 is passing!

The DISTUTILS_USE_SDK case is now failing because I removed Python 3.4 where it was being used and moved this case to 3.5. I think the util/appveyor/build.cmd would have to be updated to use a different version of MSVC/SDK than 3.4 did.

[grlee77]

I think we should just remove DISTUTILS_USE_SDK and all will be well

[grlee77]

Actually the tests for 3.8 were not run on Travis due to some pytest-related import error causing 0 tests to be found:

https://travis-ci.org/PyWavelets/pywt/jobs/599345508#L917-L954

[grlee77]

I don't understand the import error on Travis, particularly the following section

../../../../virtualenv/python3.8.0/lib/python3.8/distutils/__init__.py:4: in <module>

    import imp

In my local copy of Python 3.8 there is no import imp at this location.

The test suite ran normally on Python 3.8 when I tried it locally on a 64-bit linux machine.

[codecov-io]

Codecov Report

❗ No coverage uploaded for pull request base (master@2cced50). Click here to learn what that means.
The diff coverage is n/a.

@@           Coverage Diff            @@
##             master    #525   +/-   ##
========================================
  Coverage          ?   63.5%           
========================================
  Files             ?      23           
  Lines             ?    3891           
  Branches          ?     467           
========================================
  Hits              ?    2471           
  Misses            ?    1351           
  Partials          ?      69

---

Continue to review full report at Codecov.

Legend - Click here to learn more
Δ = absolute <relative> (impact), ø = not affected, ? = missing data
Powered by Codecov. Last update 2cced50...8139bff. Read the comment docs.

[grlee77]

Marking this as WIP for now. I don't think it is useful to merge as is since the tests on 3.8 are not currently being run.

They did get run over at MacPython/pywavelets-wheels#9 and are passing there, so I think we can work on this again after the 1.1 release.

[rgommers]

Ah I missed this one - should we merged as is?

[grlee77]

Let me take a quick look again first. I think #540 might have fixed one of the issues here, but I think we may not need 8bb8661 and I see that the Appveyor case is still commented out here for some reason.

[grlee77]

Okay, I think this looks good now. Probably want to squash the commits on merge to get rid of some of the noise (or I can do it manually if you prefer)