Fix doctest_run_option_parser for Windows

[tkf]

It should fix (part of) #2477.

[takluyver]

We need a Windows user to test this. @jstenar , will you have time in the next few days?

[bfroehle]

I'd be in favor of a simple decorator with a known failure warning on non-posix systems

@dec.knownfailureif(os.name != 'posix', 'This test is known to fail on non-POSIX systems.')

[tkf]

You mean use that decorator for doctest_run_option_parser_for_windows? But passing doctest_run_option_parser_for_posix means failing doctest_run_option_parser_for_windows. I think known failure is for something like "if you fix this bug, this failure goes away". But in this case, if you fix the problem in POSIX system, you need to remove doctest_run_option_parser_for_posix as it will fail if %run behaves the way we want.

[ghost]

tested to have fixed the doctests failure on py2.7/win7.

@takluyver , shouldn't ShiningPanda have taken care of windows testing for you?

[takluyver]

ShiningPanda is set up to run tests on master. You can point it to other branches, but it has to be manually reconfigured. Travis-CI automatically runs the tests on each pull request, but that's only on Linux.

[ghost]

That's too bad, it looks like windows testing is a recurring problem. Perhaps permanently adding a few
free-for-all integration branches to throw patches into, test, then revert back to master would help
things move more quickly.

[takluyver]

There's a skip_if_not_win32 decorator we could use.

[tkf]

@takluyver fixed.

[takluyver]

Thanks @tkf . I'll merge this, so hopefully we'll have the Windows tests passing again.