Applying multiple skipifs to parametrized test using param(marks=[]) is order dependent

[KevKeating]

If I execute

import pytest


@pytest.mark.parametrize(
    "param",
    ["foo",
     pytest.param("bar", marks=[pytest.mark.skipif(False, reason=""),
                                pytest.mark.skipif(True, reason="")])])
def test_that_passes(param):
    assert param == "foo"


@pytest.mark.parametrize(
    "param",
    ["foo",
     pytest.param("bar", marks=[pytest.mark.skipif(True, reason=""),
                                pytest.mark.skipif(False, reason="")])])
def test_that_fails(param):
    assert param == "foo"

with Python 3.6.3 and pytest 3.5 on Windows 10 64-bit, then I get

============================= test session starts =============================
platform win32 -- Python 3.6.3, pytest-3.5.0, py-1.5.3, pluggy-0.6.0
rootdir: D:\Schrodinger-Kevin\pytest_bug, inifile:
collected 4 items

pytest_bug.py .s.F                                                       [100%]

================================== FAILURES ===================================
____________________________ test_that_fails[bar] _____________________________

param = 'bar'

    @pytest.mark.parametrize(
        "param",
        ["foo",
         pytest.param("bar", marks=[pytest.mark.skipif(True, reason=""),
                                    pytest.mark.skipif(False, reason="")])])
    def test_that_fails(param):
>       assert param == "foo"
E       AssertionError: assert 'bar' == 'foo'
E         - bar
E         + foo

pytest_bug.py:19: AssertionError
================ 1 failed, 2 passed, 1 skipped in 0.07 seconds ================

The "bar" version of the test is skipped in test_that_passes, but it's executed in test_that_fails. The only difference between these two tests is the order of the skipif marks in the pytest.param argument. I would expect the "bar" version of both tests to be skipped since they both have a skipif(True) mark applied. I think this is a slightly different issue than #1296, but presumably related.

[pytestbot]

GitMate.io thinks possibly related issues are #379 (Can't not apply multiple markers to individual test when using parametrize), #552 (Can't mark.xfail a callable test param), #2846 (Unexpected order of tests using parameterized fixtures), #473 (Cannot use @mark and multiple @patch class decorators), and #2703 (Multiple use of fixture in a single test).

[RonnyPfannschmidt]

looks like yet another mark issue - please take a look if its fixed by #3317

[RonnyPfannschmidt]

actually its indeed a duplicate and fixed in #3317