If I compare same string in different Unicode normal forms the error message doesn't show any differences between them when using Python 3. For example, if I have a test like

def test_nfc_nfd():
    nfc = 'hyv\xe4'
    nfd = 'hyva\u0308'
    assert nfc == nfd

the result I got is

>       assert nfc == nfd
E       AssertionError: assert 'hyvä' == 'hyvä'
E         - hyvä
E         + hyvä

which isn't very informative.

The problem is caused by repr('hyva\u0308') being 'hyva\u0308' which is rendered as 'hyvä'. I submitted an issue about repr() escaping combining characters (i.e. turning the result to 'hyva\\u0308') but it was closed as invalid.

In PyTest it would be possible to use ascii() instead of repr(), but that would make all non-ASCII strings unreadable and that would be worse. I guess the best solution would be using ascii() if strings look the same, but I have no idea how to detect that.