Tests fail with Python < 3.5

[sbraz]

Hi Jason,
I just ran the tests with various Python versions and some fail, mostly those related to exceptions because the messages changed throughout versions
For instance, Python 3.4 will yield ValueError: need more than 0 values to unpack whereas 3.5 will have ValueError: too many values to unpack (expected 1)

I don't really know if you can fix this only through doctests but I thought it was worth opening an issue for.

[sbraz]

Sorry, I should have mentioned, when I wrote mostly, that there are also two errors with Python 2: one due to the prefix b' and another because range isn't an iterable.

================================================================================== FAILURES ===================================================================================
_________________________________________________________________ [doctest] jaraco.itertools.always_iterable __________________________________________________________________
819     already iterable, return a tuple containing only the item. If item is
820     None, an empty iterable is returned.
821 
822     >>> always_iterable([1,2,3])
823     [1, 2, 3]
824     >>> always_iterable('foo')
825     ('foo',)
826     >>> always_iterable(None)
827     ()
828     >>> always_iterable(range(10))
Expected:
    range(0, 10)
Got:
    [0, 1, 2, 3, 4, 5, 6, 7, 8, 9]

/root/git/jaraco.itertools/jaraco/itertools.py:828: DocTestFailure
_____________________________________________________________________ [doctest] jaraco.itertools.flatten ______________________________________________________________________
456     True
457 
458     Note this will normally ignore string types as iterables.
459 
460     >>> flatten(['ab', 'c'])
461     ['ab', 'c']
462 
463     Same for bytes
464 
465     >>> flatten([b'ab', b'c'])
Expected:
    [b'ab', b'c']
Got:
    ['ab', 'c']

/root/git/jaraco.itertools/jaraco/itertools.py:465: DocTestFailure
===================================================================== 2 failed, 29 passed in 0.08 seconds =====================================================================

[jaraco]

Indeed. I see those as well - they now show up in the travis-run tests.

I found that I could update the always_iterable test to pass on all platforms with this:

@@ -826,7 +828,9 @@
    >>> always_iterable(None)
    ()
-   >>> always_iterable(range(10))
-   range(0, 10)
+
+   >>> print(repr(always_iterable(range(10))).lstrip('x'))
+   range(...10)
+
    >>> def _test_func(): yield "I'm iterable"
    >>> print(next(always_iterable(_test_func())))
    I'm iterable

But that violates the primary purpose of providing clear documentation. I note that I also discovered that it's not possible to include ellipses at the beginning of the expected output because it's interpreted as continuation syntax for the previous line's >>>. I don't think it will be possible to cleanly update that test to pass on 2.7, so I'd like to skip it, but I'm not sure how to do that. I'll see what I can do.

[jaraco]

Verified