create_autospec() doesn't respect configure_mock style kwargs

[marchantjm]

BPO	46690
Nosy	@cjw296, @voidspace, @lisroach, @mariocj89, @tirkarthi

^{Note: these values reflect the state of the issue at the time it was migrated and might not reflect the current state.}

Show more details

GitHub fields:

assignee = None
closed_at = None
created_at = <Date 2022-02-09.11:54:00.145>
labels = ['3.8', 'type-bug', 'library', '3.9', '3.10']
title = "create_autospec() doesn't respect configure_mock style kwargs"
updated_at = <Date 2022-02-10.13:44:04.417>
user = 'https://bugs.python.org/marchantjm'

bugs.python.org fields:

activity = <Date 2022-02-10.13:44:04.417>
actor = 'xtreak'
assignee = 'none'
closed = False
closed_date = None
closer = None
components = ['Library (Lib)']
creation = <Date 2022-02-09.11:54:00.145>
creator = 'marchant.jm'
dependencies = []
files = []
hgrepos = []
issue_num = 46690
keywords = []
message_count = 2.0
messages = ['412898', '413001']
nosy_count = 6.0
nosy_names = ['cjw296', 'michael.foord', 'lisroach', 'mariocj89', 'xtreak', 'marchant.jm']
pr_nums = []
priority = 'normal'
resolution = None
stage = None
status = 'open'
superseder = None
type = 'behavior'
url = 'https://bugs.python.org/issue46690'
versions = ['Python 3.8', 'Python 3.9', 'Python 3.10']

Linked PRs

• gh-90848: Fixed create_autospec ignoring configure_mock style kwargs #118163
• [3.12] gh-90848: Fixed create_autospec ignoring configure_mock style kwargs (GH-118163) #118517

[marchantjm]

When using create_autospec() to create a mock object, it doesn't respect values passed through in the style described for passing mock configurations in the Mock constructor (https://docs.python.org/3.8/library/unittest.mock.html#unittest.mock.Mock.configure_mock). Instead, they seem to get discarded somewhere here (

cpython/Lib/unittest/mock.py

Lines 2693 to 2741 in 77bab59

	for entry in dir(spec):
	if _is_magic(entry):
	# MagicMock already does the useful magic methods for us
	continue
	# XXXX do we need a better way of getting attributes without
	# triggering code execution (?) Probably not - we need the actual
	# object to mock it so we would rather trigger a property than mock
	# the property descriptor. Likewise we want to mock out dynamically
	# provided attributes.
	# XXXX what about attributes that raise exceptions other than
	# AttributeError on being fetched?
	# we could be resilient against it, or catch and propagate the
	# exception when the attribute is fetched from the mock
	try:
	original = getattr(spec, entry)
	except AttributeError:
	continue
	kwargs = {'spec': original}
	if spec_set:
	kwargs = {'spec_set': original}
	if not isinstance(original, FunctionTypes):
	new = _SpecState(original, spec_set, mock, entry, instance)
	mock._mock_children[entry] = new
	else:
	parent = mock
	if isinstance(spec, FunctionTypes):
	parent = mock.mock
	skipfirst = _must_skip(spec, entry, is_type)
	kwargs['_eat_self'] = skipfirst
	if iscoroutinefunction(original):
	child_klass = AsyncMock
	else:
	child_klass = MagicMock
	new = child_klass(parent=parent, name=entry, _new_name=entry,
	_new_parent=parent,
	**kwargs)
	mock._mock_children[entry] = new
	_check_signature(original, new, skipfirst=skipfirst)
	# so functions created with _set_signature become instance attributes,
	# *plus* their underlying mock exists in _mock_children of the parent
	# mock. Adding to _mock_children may be unnecessary where we are also
	# setting as an instance attribute?
	if isinstance(new, FunctionTypes):
	setattr(mock, entry, new)

).

Here's a simple test case:

from unittest.mock import create_autospec

class Test:
    def test_method(self):
        pass

autospec_mock = create_autospec(Test, instance=True, **{"test_method.side_effect": ValueError})

# Should throw a ValueError exception
autospec_mock.test_method()

# Assign manually
autospec_mock.test_method.side_effect = ValueError
# Throws as expected
autospec_mock.test_method()

[tirkarthi]

I guess the problem is that during the initial mock creation kwargs is passed so calling test_method immediately after mock creation raises ValueError. But as we loop through the attributes and create new child mock for the attributes the original configured mock for the method that raises ValueError is overridden by another object without the configuration info. Probably it makes sense to call configure_mock again after all children mock are constructed. Something like below works and I don't see any test failures in mock related test cases.

index 2719f74d6f..585e875c95 100644
--- a/Lib/unittest/mock.py
+++ b/Lib/unittest/mock.py
@@ -2637,6 +2637,7 @@ def create_autospec(spec, spec_set=False, instance=False, _parent=None,
                                f'[object={spec!r}]')
     is_async_func = _is_async_func(spec)
     _kwargs = {'spec': spec}
+    original_kwargs = kwargs
     if spec_set:
         _kwargs = {'spec_set': spec}
     elif spec is None:
@@ -2740,6 +2741,9 @@ def create_autospec(spec, spec_set=False, instance=False, _parent=None,
         if isinstance(new, FunctionTypes):
             setattr(mock, entry, new)
 
+    if _is_instance_mock(mock):
+        mock.configure_mock(**original_kwargs)
+
     return mock