It seems that if you call skipTest *anywhere* in a test function, even in a subTest, the *entire* function gets marked as "skipped", even if only one sub-test is skipped.

Example:

    import unittest

    class SomeTest(unittest.TestCase):
        def test_something(self):
            for i in range(1, 3):
                with self.subTest(i):
                    if i > 1:
                        self.skipTest('Not supported')

                    self.assertEqual(i, 1)

If you run `python3.7 -m unittest -v` on this, you get:

    $ python -m unittest -v
    test_something (test_mod.SomeTest) ... skipped 'Not supported'

    ----------------------------------------------------------------------
    Ran 1 test in 0.000s

    OK (skipped=1)


Despite the fact that the test *was* run in the `i == 1` case. Similarly, pytest marks this as a single skipped test:

    ========= test session starts =======
    
    platform linux -- Python 3.7.1, pytest-4.0.1, py-1.7.0, pluggy-0.8.0
    rootdir: /tmp/test_mod, inifile:
    collected 1 item
    
    test_mod.py s                                           [100%]
    
    ===== 1 skipped in 0.00 seconds =====


The solution to this is not obvious, unfortunately. One way to solve it would be to have each subtest considered a "separate test". Another would be to detect whether subTests have been skipped and expose only the *skipped* tests as separate tests. You could also add a "partially skipped" marker, so to say, "Some parts of this were skipped"

I suspect the right answer is some combination of these three answers - possibly adding an extra-verbose mode that splits out all sub tests and having skipped subTests default to being displayed separately.

Somewhat related to issue #30997.

As "prior art" the way that pytest does this is to have parametrized tests show up as separate tests:

    import pytest

    @pytest.mark.parametrize("i", range(1, 3))
    def test_something(i):
        if i > 1:
            pytest.skip('Not supported')
        assert i == 1

    @pytest.mark.parametrize("i", range(1, 3))
    def test_something_else(i):
        assert 3 > i >= 1


Running `pytest -v` for this gives:


======================================= test session starts ========================================
platform linux -- Python 3.7.1, pytest-4.0.1, py-1.7.0, pluggy-0.8.0
cachedir: .pytest_cache
rootdir: /tmp/test_mod, inifile:
collected 4 items                                                                                  

test_mod.py::test_something[1] PASSED                                                        [ 25%]
test_mod.py::test_something[2] SKIPPED                                                       [ 50%]
test_mod.py::test_something_else[1] PASSED                                                   [ 75%]
test_mod.py::test_something_else[2] PASSED                                                   [100%]

=============================== 3 passed, 1 skipped in 0.01 seconds ================================

Did you notice that `skipped 'Not supported'` will be displayed once per skipped subtest so changing your `for i in range(1, 3):` by `for i in range(1, 5):` will show:

    python3 -m unittest -v
    test_something (test2.SomeTest) ... skipped 'Not supported'
    skipped 'Not supported'
    skipped 'Not supported'


Do you think it should show something like:

    python3 -m unittest -v
    test_something (test2.SomeTest) ... SubTest skipped 'Not supported'
                                        SubTest skipped 'Not supported'
                                        SubTest skipped 'Not supported'
?

@Rémi Interesting. Your suggested output does look clearer than the existing one, but it still doesn't indicate that anything *passed*. I think I like the way pytest does it the best, but if we can't expose the subtests as separate tests, I'd probably want it to be more like this:

    test_something (test2.SomeTest) ... ok (3 subtests skipped)
    test_something [2] skipped 'Not supported'
    test_something [3] skipped 'Not supported'
    test_something [4] skipped 'Not supported'

I think this is a nice output, taking a quick look at unittest source, all the information needed to display this is save in the TestResult object, showing skipped tests is done here: https://github.com/python/cpython/blob/master/Lib/unittest/runner.py#L85

It seems to me that the change would not be very hard to add.

Sounds very similar to Issue 25894, discussing how to deal with tests where different subtests errored, failed, skipped and passed.