Updating testing and (minor) MSVC improvement

[reductor]

I have a bunch of fixes and improvements coming for future pushes, doing an overhaul of the way tests are run so then it can be verified what changed.

TODO: (Minor) - Let me know if any required for pull-request

• Cover more python versions (Not covered before)
• Fully verify non-stolen decompyle stolen tests are covered (If pull accepted, will bug)
• Documentation on how to update baseline (copy-based)

[ghost]

Regarding test cases from distributions 1.5.2 - 3.4.2: I have been using the Xeon Phi to execute many tests in parallel.

Some preliminary sorting:

These tests produced output, [not validated] https://github.com/mancoast/CPythonPyc_test/tree/master/cpython

These tests crashed [validated as incorrect]: https://github.com/mancoast/CPythonPyc_test/tree/master/crash

These tests produced bad output [validated as incorrect]: https://github.com/mancoast/CPythonPyc_test/tree/master/fail

[reductor]

Thanks for wider test suite. I welcome and wish to integrate this larger suite however should it be a current CL blocker or new commit?

[ghost]

those files were dumped from the full regression test foreach python distribution. what do you mean by current CL blocker or new commit?

[reductor]

What I mean by blocker or new commit is should your suggestion block committing/merge the change I currently have in the pull request (to make it more ideal with CPythonPyc_test regression tests), or should that be done as a later pull request/commit

[ghost]

I like your suggestions with the test suite. the additional test cases from cpythonpyc-test do not block this commit. The additional test cases require some restructuring.

[rocky]

I just came across this. It is a pity this hasn't been acted upon. If someone wants to rework them in the context of uncompyle6's test framework, I'd support that.

I see that a number of these come from decompyle's test suite, for example test_applyEquiv.py. This particular test exists under in python-uncompyle6/test/simple_source/bug22/02_apply_equiv.py.

However the better way used in testing now would be to adapt this program to add asserts, so that when this program runs, it checks itself. For example, instead of:

no_apply(*args, **args)

we would have:

assert no_apply(*args, **args) == ...