Currently, Windows TestCases on Appveyor established on #519 are "artifically" ok.
Because in the code, the Test Cases known to fail on have been marked with the HIDE_WINDOWS_KNOWN_ERRORS variable so that we get an ALL GREEN response; that way we can detect when new commits break additional TCs on Windows.

Currently there are 10 failing TCs (search them with the above var-name), and you can view the errors in this Appveyor job:

https://ci.appveyor.com/project/ankostis/gitpython/build/1.0.196

The errors are mostly related to 2 issues:

1. Unicode: for those with reading time, I would suggest these 2 contradictory sources:
   • problem: Everything you did not want to know about Unicode in Python 3
   • solution: PEP 383 -- Non-decodable Bytes in System Character Interfaces - so using text-mode
2. Leaked resources: these are (possibly memmory-mapped) files that are not deallocated (and hence, not deleted afterwards). Unfortunately, dealing with them might mean that you have to jump projects, as some are inherited from the library projects smmp and (mostly) gitdb.

In order to reproduce the failed TCs on your PC, change locally git/test/lib/helper.py:#L38 from this:

    HIDE_WINDOWS_KNOWN_ERRORS = os.environ.get('HIDE_WINDOWS_KNOWN_ERRORS', True)

into this:

    HIDE_WINDOWS_KNOWN_ERRORS = os.environ.get('HIDE_WINDOWS_KNOWN_ERRORS', False)

This issue is about not forgetting these TCs - asking for help to fix them.