TestAssertionRewrite test failures on 3.7.0b5

[hroncok]

Running on Fedora 28.

$ tox -e py37 -- testing/test_assertrewrite.py

The tests pass with 3.7.0b4, bt two of them fail with 3.7.0b5:

============================= test session starts ==============================
platform linux -- Python 3.7.0b5, pytest-3.6.1.dev44+gd609b635, py-1.5.3, pluggy-0.6.0
rootdir: /home/churchyard/Dokumenty/pytest, inifile: tox.ini
plugins: hypothesis-3.57.0
collected 57 items                                                             

testing/test_assertrewrite.py FF.....................................s.. [ 73%]
ssss...........                                                          [100%]

=================================== FAILURES ===================================
_______________ TestAssertionRewrite.test_place_initial_imports ________________

self = <test_assertrewrite.TestAssertionRewrite object at 0x7fd164296a58>

    def test_place_initial_imports(self):
        s = """'Doc string'\nother = stuff"""
        m = rewrite(s)
        # Module docstrings in 3.7 are part of Module node, it's not in the body
        # so we remove it so the following body items have the same indexes on
        # all Python versions
        if sys.version_info < (3, 7):
            assert isinstance(m.body[0], ast.Expr)
            assert isinstance(m.body[0].value, ast.Str)
            del m.body[0]
        for imp in m.body[0:2]:
>           assert isinstance(imp, ast.Import)
E           AssertionError: assert False
E            +  where False = isinstance(<_ast.Expr object at 0x7fd16422ee48>, <class '_ast.Import'>)
E            +    where <class '_ast.Import'> = ast.Import

testing/test_assertrewrite.py:81: AssertionError
____________________ TestAssertionRewrite.test_dont_rewrite ____________________

self = <test_assertrewrite.TestAssertionRewrite object at 0x7fd1642cdbe0>

    def test_dont_rewrite(self):
        s = """'PYTEST_DONT_REWRITE'\nassert 14"""
        m = rewrite(s)
        if sys.version_info < (3, 7):
            assert len(m.body) == 2
            assert isinstance(m.body[0], ast.Expr)
            assert isinstance(m.body[0].value, ast.Str)
            del m.body[0]
        else:
>           assert len(m.body) == 1
E           assert 2 == 1
E            +  where 2 = len([<_ast.Expr object at 0x7fd164274710>, <_ast.Assert object at 0x7fd16428cba8>])
E            +    where [<_ast.Expr object at 0x7fd164274710>, <_ast.Assert object at 0x7fd16428cba8>] = <_ast.Module object at 0x7fd164274898>.body

testing/test_assertrewrite.py:133: AssertionError
=========================== short test summary info ============================
FAIL testing/test_assertrewrite.py::TestAssertionRewrite::()::test_place_initial_imports
FAIL testing/test_assertrewrite.py::TestAssertionRewrite::()::test_dont_rewrite
SKIP [1] testing/test_assertrewrite.py:845: python 2 only
SKIP [1] testing/test_assertrewrite.py:895: condition: sys.version_info[0] >= 3
SKIP [1] testing/test_assertrewrite.py:904: condition: sys.version_info[0] >= 3
SKIP [1] testing/test_assertrewrite.py:915: condition: sys.version_info[0] >= 3
SKIP [1] testing/test_assertrewrite.py:926: condition: sys.version_info[0] >= 3
================ 2 failed, 50 passed, 5 skipped in 6.19 seconds ================

Happens on both 3.6.0 and master.

The docstring chanages in AST change form 3.7 got reverted. This might be relevant.

See also PyCQA/mccabe#61 (comment)

[hroncok]

Blame gives a merge commit f3a119c

I think that if I just change the version in there to 3.8, it'll do.

[pytestbot]

GitMate.io thinks possibly related issues are #386 (tests fail with pexpect-3.0), #2975 (Pytest 3.3: tests being skipped without failing), #3011 (Python 3.7 Nightly failures due to Python bug or travis), #1083 (Regression with parallel tests on 2.7.3 and 2.8), and #206 (Test failure with disabled byte-compilation).

[RonnyPfannschmidt]

@hroncok this relates cpython undoing the change to ast, changing it to 3.8 would be absolute horror

[hroncok]

@RonnyPfannschmidt I'm afraid I don't understand why?

See #3531

[hroncok]

Any plans for a release in couple of days, or should I backport the fix as a patch (in Fedora)?

[nicoddemus]

Was planning on making a release this week, might as well prepare one right away: #3534.