Sometimes, test_pathlib fails because shutil.rmtree() cannot remove a test directory. Example:

http://buildbot.python.org/all/builders/AMD64%20Windows7%20SP1%203.x/builds/4833/steps/test/logs/stdio

======================================================================
ERROR: test_touch_common (test.test_pathlib.WindowsPathTest)
----------------------------------------------------------------------
Traceback (most recent call last):
  File "C:\buildbot.python.org\3.x.kloth-win64\build\lib\shutil.py", line 477, in rmtree
    return _rmtree_unsafe(path, onerror)
  File "C:\buildbot.python.org\3.x.kloth-win64\build\lib\shutil.py", line 367, in _rmtree_unsafe
    _rmtree_unsafe(fullname, onerror)
  File "C:\buildbot.python.org\3.x.kloth-win64\build\lib\shutil.py", line 376, in _rmtree_unsafe
    onerror(os.rmdir, path, sys.exc_info())
  File "C:\buildbot.python.org\3.x.kloth-win64\build\lib\shutil.py", line 374, in _rmtree_unsafe
    os.rmdir(path)
OSError: [WinError 145] The directory is not empty: 'C:\\buildbot.python.org\\3.x.kloth-win64\\build\\build\\test_python_3444\\@test_3444_tmp\\dirB'

The error comes from this cleanup function:

        self.addCleanup(shutil.rmtree, BASE)

I don't understand how rmtree() can fail with "The directory is not empty".

Note: this buildbot runs 4 tests in parallel using the "-j4" command line option of regrtest.

FWIW, every use of rmtree I have on Windows occasionally fails this way, parallelism seems not to be a factor.

This looks like a duplicate of #19811, which was closed as a duplicate of #19629.

New changeset 1db5cde4958f by Victor Stinner in branch '3.4':
Issue #19811, #22022: test_pathlib uses support.rmtree() instead of
http://hg.python.org/cpython/rev/1db5cde4958f

New changeset e405bcbf761c by Victor Stinner in branch 'default':
Merge Python 3.4
http://hg.python.org/cpython/rev/e405bcbf761c