Regression: duplicate 'fixt' when overriding parametrized fixture with parametrize

[The-Compiler]

This test:

import pytest

@pytest.fixture(params=[1, 2])
def fixt(request):
    return request.param

@pytest.mark.parametrize('fixt, val', [(3, 'x'), (4, 'x')])
def test_foo(fixt, val):
    pass

works with 2.7.2 but fails since #926 / 41cef6f:

.venv/lib/python3.4/site-packages/pluggy.py:724: in __call__
    return self._hookexec(self, self._nonwrappers + self._wrappers, kwargs)
.venv/lib/python3.4/site-packages/pluggy.py:338: in _hookexec
    return self._inner_hookexec(hook, methods, kwargs)
.venv/lib/python3.4/site-packages/pluggy.py:333: in <lambda>
    _MultiCall(methods, kwargs, hook.spec_opts).execute()
.venv/lib/python3.4/site-packages/pluggy.py:595: in execute
    return _wrapped_call(hook_impl.function(*args), self.execute)
.venv/lib/python3.4/site-packages/pluggy.py:249: in _wrapped_call
    wrap_controller.send(call_outcome)
_pytest/python.py:285: in pytest_pycollect_makeitem
    res = list(collector._genfunctions(name, obj))
_pytest/python.py:442: in _genfunctions
    self.ihook.pytest_generate_tests(metafunc=metafunc)
.venv/lib/python3.4/site-packages/pluggy.py:724: in __call__
    return self._hookexec(self, self._nonwrappers + self._wrappers, kwargs)
.venv/lib/python3.4/site-packages/pluggy.py:338: in _hookexec
    return self._inner_hookexec(hook, methods, kwargs)
.venv/lib/python3.4/site-packages/pluggy.py:333: in <lambda>
    _MultiCall(methods, kwargs, hook.spec_opts).execute()
.venv/lib/python3.4/site-packages/pluggy.py:596: in execute
    res = hook_impl.function(*args)
_pytest/python.py:195: in pytest_generate_tests
    metafunc.parametrize(*marker.args, **marker.kwargs)
_pytest/python.py:967: in parametrize
    param_index)
_pytest/python.py:819: in setmulti
    self._checkargnotcontained(arg)
_pytest/python.py:802: in _checkargnotcontained
    raise ValueError("duplicate %r" %(arg,))
E   ValueError: duplicate 'fixt'

It only seems to fail if:

• The fixture is parametrized
• It's overridden by a parametrized argument
• There is at least one other parametrized argument.

I agree overriding a fixture this way is a bit weird, but I think it should still work.

/cc @untitaker @bubenkoff

by the way - thanks to git bisect this was easy to find out:

git bisect start
git bisect bad master
git bisect good 2.7.2
git bisect run ./.venv/bin/py.test test_foo.py

[bubenkoff]

can you please check if it matters that the fixture is defined in same module?

[The-Compiler]

If I have a test_foo.py with:

import pytest

@pytest.mark.parametrize('fixt, val', [(3, 'x'), (4, 'x')])
def test_foo(fixt, val):
    pass

and a test_bar.py with:

import pytest

@pytest.fixture(params=[1, 2])
def fixt(request):
    return request.param

def test_bar(fixt):
    pass

that seems to work:

test_bar.py ..
test_foo.py ..

==== 4 passed in 0.01 seconds ====

When I have the fixture in conftest.py, it fails. So it seems to fail whenever the fixt fixture is available in test_foo.py.

[bubenkoff]

does the test folder have init.py?

[The-Compiler]

no:

$ mkdir foo
$ cd foo
$ ls
$ virtualenv .venv
Running virtualenv with interpreter /usr/bin/python2
New python executable in .venv/bin/python2
Also creating executable in .venv/bin/python
Installing setuptools, pip...done.
$ ./.venv/bin/pip install git+https://github.com/pytest-dev/pytest.git
Downloading/unpacking git+https://github.com/pytest-dev/pytest.git
  Cloning https://github.com/pytest-dev/pytest.git to /tmp/pip-zpWML1-build
  Running setup.py (path:/tmp/pip-zpWML1-build/setup.py) egg_info for package from git+https://github.com/pytest-dev/pytest.git

    Installed /tmp/pip-zpWML1-build/setuptools_scm-1.7.0-py2.7.egg
    your setuptools is too old (<12)
    setuptools_scm functionality is degraded

Downloading/unpacking py>=1.4.29 (from pytest==2.7.3.dev380-ng553aef5)
  Downloading py-1.4.30-py2.py3-none-any.whl (81kB): 81kB downloaded
Downloading/unpacking pluggy>=0.3.0,<0.4.0 (from pytest==2.7.3.dev380-ng553aef5)
  Downloading pluggy-0.3.0-py2.py3-none-any.whl
Installing collected packages: py, pluggy, pytest
  Running setup.py install for pytest
    your setuptools is too old (<12)
    setuptools_scm functionality is degraded

    Installing py.test-2.7 script to /home/florian/foo/.venv/bin
    Installing py.test script to /home/florian/foo/.venv/bin
Successfully installed py pluggy pytest
Cleaning up...
$ cat > test_foo.py
import pytest

@pytest.fixture(params=[1, 2])
def fixt(request):
    return request.param

@pytest.mark.parametrize('fixt, val', [(3, 'x'), (4, 'x')])
def test_foo(fixt, val):
    pass
$ ./.venv/bin/py.test test_foo.py 
========================================== test session starts ==========================================
platform linux2 -- Python 2.7.9, pytest-2.7.3.dev380+ng553aef5, py-1.4.30, pluggy-0.3.0
rootdir: /home/florian/foo, inifile: 
collected 0 items / 1 errors 

================================================ ERRORS =================================================
_____________________________________ ERROR collecting test_foo.py ______________________________________
.venv/local/lib/python2.7/site-packages/pluggy.py:724: in __call__
    return self._hookexec(self, self._nonwrappers + self._wrappers, kwargs)
.venv/local/lib/python2.7/site-packages/pluggy.py:338: in _hookexec
    return self._inner_hookexec(hook, methods, kwargs)
.venv/local/lib/python2.7/site-packages/pluggy.py:333: in <lambda>
    _MultiCall(methods, kwargs, hook.spec_opts).execute()
.venv/local/lib/python2.7/site-packages/pluggy.py:595: in execute
    return _wrapped_call(hook_impl.function(*args), self.execute)
.venv/local/lib/python2.7/site-packages/pluggy.py:249: in _wrapped_call
    wrap_controller.send(call_outcome)
.venv/local/lib/python2.7/site-packages/_pytest/python.py:285: in pytest_pycollect_makeitem
    res = list(collector._genfunctions(name, obj))
.venv/local/lib/python2.7/site-packages/_pytest/python.py:442: in _genfunctions
    self.ihook.pytest_generate_tests(metafunc=metafunc)
.venv/local/lib/python2.7/site-packages/pluggy.py:724: in __call__
    return self._hookexec(self, self._nonwrappers + self._wrappers, kwargs)
.venv/local/lib/python2.7/site-packages/pluggy.py:338: in _hookexec
    return self._inner_hookexec(hook, methods, kwargs)
.venv/local/lib/python2.7/site-packages/pluggy.py:333: in <lambda>
    _MultiCall(methods, kwargs, hook.spec_opts).execute()
.venv/local/lib/python2.7/site-packages/pluggy.py:596: in execute
    res = hook_impl.function(*args)
.venv/local/lib/python2.7/site-packages/_pytest/python.py:195: in pytest_generate_tests
    metafunc.parametrize(*marker.args, **marker.kwargs)
.venv/local/lib/python2.7/site-packages/_pytest/python.py:967: in parametrize
    param_index)
.venv/local/lib/python2.7/site-packages/_pytest/python.py:819: in setmulti
    self._checkargnotcontained(arg)
.venv/local/lib/python2.7/site-packages/_pytest/python.py:802: in _checkargnotcontained
    raise ValueError("duplicate %r" %(arg,))
E   ValueError: duplicate 'fixt'
======================================== 1 error in 0.22 seconds ========================================

[bubenkoff]

try to add init.py, sorry being boring :)

[The-Compiler]

Adding an __init__.py doesn't change anything.

http://www.the-compiler.org | me@the-compiler.org (Mail/XMPP)
GPG: 916E B0C8 FD55 A072 | http://the-compiler.org/pubkey.asc
I love long mails! | http://email.is-not-s.ms/

[bubenkoff]

you're right - checked the

pytest/testing/python/fixture.py

Line 227 in 73fdda0

def test_override_parametrized_fixture_conftest_module(self, testdir):

and lower, there's no such test
would you mind to add such?
and probably work on a fix?

[The-Compiler]

I can add a test, but that probably won't help much without a fix seeing that I already have a minimal testcase above.

As for the fix, I don't really understand the fixture code and I'm quite busy with my own project - so I'd really appreciate if @untitaker or someone else could take a look!

[bubenkoff]

well, making test fail is a big first step anyway!

[nicoddemus]

@The-Compiler I think @bubenkoff is asking for you to write a test for it and mark it as xfail, preferably adding this issue number as reason.

But of course, if you don't have the time that's understandable! 😄

[bubenkoff]

thanks @nicoddemus :)

[untitaker]

I currently don't have the time for this, but I suspect that in my commit, the condition should've been refined instead of removed (as discussed with @RonnyPfannschmidt in the PR).

On 28 August 2015 15:01:04 CEST, Bruno Oliveira notifications@github.com wrote:

@The-Compiler I think @bubenkoff is asking for you to write a test for
it and mark it as xfail, preferably adding this issue number as
reason.

But of course, if you don't have the time that's understandable!
😄

---

Reply to this email directly or view it on GitHub:
#979 (comment)

Sent from my phone. Please excuse my brevity.