The concerned part : http://hg.python.org/cpython/rev/b836a0cd68f7#l4.15

"NameError: name 'keySet' is not defined."

New changeset af1351800c7a by Terry Jan Reedy in branch '2.7':
Issue #21982: Add minimal unittest for configDialog with 46% coverage.
http://hg.python.org/cpython/rev/af1351800c7a

New changeset 681979c6e6b2 by Terry Jan Reedy in branch '3.4':
Issue #21982: Add minimal unittest for configDialog with 46% coverage.
http://hg.python.org/cpython/rev/681979c6e6b2

Thanks for catching this. Either I misread pyflakes output or it has a bug. In either case, I should have searched this one like I did some others.  The htest caught this. I added a minimal unittest that initially failed, covers half the module, and now passes.

2.7 did not have the error, but the rest of the changes apply.

---
More info on a recurring 'nuisance': Start Idle on Windows from debug build interpreter. After ConfigDialog(), test termination causes the following to be printed in the interpreter window.

can't invoke "event" command: application has been destroyed
    while executing
"event generate $w <<ThemeChanged>>"
    (procedure "ttk::ThemeChanged" line 6)
    invoked from within
"ttk::ThemeChanged"

This happens *after* tearDownClass runs. I determined this by adding
def tearDownModule(): input('mod') to pause. Changing to "d = ConfigDialog(...); d.destroy()" has no effect. Commenting out parts of dialog construction might narrow down the responsible component.