capture: ensure name of EncodedFile being a string

[blueyed]

Fixes #2555.

[nicoddemus]

Awesome, thanks @blueyed!

Could you please also add a new "bugfix" file to the changelog?

[nicoddemus]

Please add a small comment here explaining why this special case

[blueyed]

Will do, but it is not really clear to me if this is the way to go after all, since it seems to be only necessary for logging.StreamHandler ?!

@RonnyPfannschmidt
What API where you referring to that says this to be a string?

[nicoddemus]

I think the overall file.name being a string, IIUC.

[RonnyPfannschmidt]

hmm after revisiting, why not just make name a property?

[nicoddemus]

Ahh good idea. 👍

[coveralls]

Coverage increased (+0.002%) to 92.14% when pulling 1264971 on blueyed:EncodedFile-name into 7cd03d7 on pytest-dev:master.

[nicoddemus]

This failed in py26 because the field index is mandatory... but I suggest just using repr(self.buffer) as it is shorter and simpler.

[blueyed]

Makes sense, amended.

[coveralls]

Coverage increased (+0.002%) to 92.14% when pulling 97bdc6f on blueyed:EncodedFile-name into 7cd03d7 on pytest-dev:master.

[RonnyPfannschmidt]

while reading, why is get-attribute used there to begin with ?

[nicoddemus]

This code was already there, I think the idea is that EncodedFile should act as a proxy to the underlying buffer.

[RonnyPfannschmidt]

but then buffer is directly usable to begin with, from my pov that call can very likely be shortened to self.buffer

[nicoddemus]

OH that! Hmm yeah at first glance you are right, but I suspect this is there for a reason. We would have to hunt the Git history to see if we can find something about it.

[blueyed]

Let's see if it is covered by some test(s): #2614.

[blueyed]

test_pickling_and_unpickling_enocded_file fails:

    def test_pickling_and_unpickling_enocded_file():
        # See https://bitbucket.org/pytest-dev/pytest/pull-request/194
        # pickle.loads() raises infinite recursion if
        # EncodedFile.__getattr__ is not implemented properly
        ef = capture.EncodedFile(None, None)
        ef_as_str = pickle.dumps(ef)
>       pickle.loads(ef_as_str)

testing/test_capture.py:1151: 
_ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _
.tox/py36/lib/python3.6/site-packages/_pytest/capture.py:258: in __getattr__
    return getattr(self.buffer, name)
.tox/py36/lib/python3.6/site-packages/_pytest/capture.py:258: in __getattr__
    return getattr(self.buffer, name)
.tox/py36/lib/python3.6/site-packages/_pytest/capture.py:258: in __getattr__
    return getattr(self.buffer, name)
E   RecursionError: maximum recursion depth exceeded while calling a Python object
!!! Recursion detected (same locals & position)

[coveralls]

Coverage increased (+0.002%) to 92.058% when pulling 5c6a4c9 on blueyed:EncodedFile-name into 70d9f86 on pytest-dev:master.

[nicoddemus]

Hmm the check brakes in py26:

___________________________ test_dupfile_on_bytesio ___________________________
    def test_dupfile_on_bytesio():
        io = py.io.BytesIO()
        f = capture.safe_text_dupfile(io, "wb")
        f.write("hello")
        assert io.getvalue() == b"hello"
>       assert f.name.startswith('<_io.BytesIO object at')
E       AssertionError: assert False
E        +  where False = <built-in method startswith of str object at 0x03FB6A88>('<_io.BytesIO object at')
E        +    where <built-in method startswith of str object at 0x03FB6A88> = '<io.BytesIO object at 0x042FCDB0>'.startswith
E        +      where '<io.BytesIO object at 0x042FCDB0>' = <_pytest.capture.EncodedFile object at 0x03C57930>.name

I think changing the assertion to assert "BytesIO object" in f.name would be good enough.

[coveralls]

Coverage increased (+0.002%) to 92.058% when pulling 0603d1d on blueyed:EncodedFile-name into 79097e8 on pytest-dev:master.

[nicoddemus]

Thanks @blueyed!