I suspect the relevant failure here (which is not listed in Pablo's post) is this one:

======================================================================
ERROR: test_realpath_curdir (test.test_ntpath.TestNtpath)
----------------------------------------------------------------------
Traceback (most recent call last):
  File "C:\buildbot.python.org\3.x.kloth-win64\build\lib\test\test_ntpath.py", line 210, in test_realpath_curdir
    tester("ntpath.realpath('/'.join(['.'] * 100))", expected)
  File "C:\buildbot.python.org\3.x.kloth-win64\build\lib\test\test_ntpath.py", line 30, in tester
    raise TestFailed("%s should return: %s but returned: %s" \
test.support.TestFailed: ntpath.realpath('/'.join(['.'] * 100)) should return: C:\buildbot.python.org\3.x.kloth-win64\build\build\test_python_5340\test_python_worker_1408 but returned: \\?\C:\buildbot.python.org\3.x.kloth-win64\build\build\test_python_5340\test_python_worker_1408\.\.\.\.\.\.\.\.\.\.\.\.\.\.\.\.

The path being normalized is longer than MAX_PATH, which is clearly causing failures with Windows APIs as implemented on Windows 7.

I can add the normpath() call back in, which should remove all the dots before ever actually testing realpath() and get it back down under MAX_PATH. That won't help anyone using real long paths on Win7 though, but then again the only reason this test was passing before was because there was a normpath() call, so we're no worse off (apart from performance-wise).

The venv issues seem unrelated. I'll get the realpath one fixed first and then take a look at those once I've finished today's tasks.