manila/manila/tests/test_hacking.py
Ha Van Tu 8407f72c2c [hacking] Ensure not to use LOG.warn
This patchs ensures not to use LOG.warn because LOG.warn deprecated
in Python 3 [1], use LOG.warning instead.

[1] https://docs.python.org/3/library/logging.html#logging.warning

TrivialFix

Change-Id: I50e44a89092e0ee15f5be294837a87c321427921
2016-11-30 06:36:51 +00:00

344 lines
14 KiB
Python

# Copyright 2014 Red Hat, Inc.
#
# Licensed under the Apache License, Version 2.0 (the "License"); you may
# not use this file except in compliance with the License. You may obtain
# a copy of the License at
#
# http://www.apache.org/licenses/LICENSE-2.0
#
# Unless required by applicable law or agreed to in writing, software
# distributed under the License is distributed on an "AS IS" BASIS, WITHOUT
# WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied. See the
# License for the specific language governing permissions and limitations
# under the License.
import sys
import textwrap
import ddt
import mock
import pep8
from manila.hacking import checks
from manila import test
@ddt.ddt
class HackingTestCase(test.TestCase):
"""Hacking test cases
This class tests the hacking checks in manila.hacking.checks by passing
strings to the check methods like the pep8/flake8 parser would. The parser
loops over each line in the file and then passes the parameters to the
check method. The parameter names in the check method dictate what type of
object is passed to the check method. The parameter types are::
logical_line: A processed line with the following modifications:
- Multi-line statements converted to a single line.
- Stripped left and right.
- Contents of strings replaced with "xxx" of same length.
- Comments removed.
physical_line: Raw line of text from the input file.
lines: a list of the raw lines from the input file
tokens: the tokens that contribute to this logical line
line_number: line number in the input file
total_lines: number of lines in the input file
blank_lines: blank lines before this one
indent_char: indentation character in this file (" " or "\t")
indent_level: indentation (with tabs expanded to multiples of 8)
previous_indent_level: indentation on previous line
previous_logical: previous logical line
filename: Path of the file being run through pep8
When running a test on a check method the return will be False/None if
there is no violation in the sample input. If there is an error a tuple is
returned with a position in the line, and a message. So to check the result
just assertTrue if the check is expected to fail and assertFalse if it
should pass.
"""
def test_no_translate_debug_logs(self):
self.assertEqual(1, len(list(checks.no_translate_debug_logs(
"LOG.debug(_('foo'))", "manila/scheduler/foo.py"))))
self.assertEqual(0, len(list(checks.no_translate_debug_logs(
"LOG.debug('foo')", "manila/scheduler/foo.py"))))
self.assertEqual(0, len(list(checks.no_translate_debug_logs(
"LOG.info(_('foo'))", "manila/scheduler/foo.py"))))
def test_check_explicit_underscore_import(self):
self.assertEqual(1, len(list(checks.check_explicit_underscore_import(
"LOG.info(_('My info message'))",
"manila/tests/other_files.py"))))
self.assertEqual(1, len(list(checks.check_explicit_underscore_import(
"msg = _('My message')",
"manila/tests/other_files.py"))))
self.assertEqual(0, len(list(checks.check_explicit_underscore_import(
"from manila.i18n import _",
"manila/tests/other_files.py"))))
self.assertEqual(0, len(list(checks.check_explicit_underscore_import(
"LOG.info(_('My info message'))",
"manila/tests/other_files.py"))))
self.assertEqual(0, len(list(checks.check_explicit_underscore_import(
"msg = _('My message')",
"manila/tests/other_files.py"))))
self.assertEqual(0, len(list(checks.check_explicit_underscore_import(
"from manila.i18n import _LE, _, _LW",
"manila/tests/other_files2.py"))))
self.assertEqual(0, len(list(checks.check_explicit_underscore_import(
"msg = _('My message')",
"manila/tests/other_files2.py"))))
self.assertEqual(0, len(list(checks.check_explicit_underscore_import(
"_ = translations.ugettext",
"manila/tests/other_files3.py"))))
self.assertEqual(0, len(list(checks.check_explicit_underscore_import(
"msg = _('My message')",
"manila/tests/other_files3.py"))))
# Complete code coverage by falling through all checks
self.assertEqual(0, len(list(checks.check_explicit_underscore_import(
"LOG.info('My info message')",
"manila.tests.unit/other_files4.py"))))
self.assertEqual(0, len(list(checks.check_explicit_underscore_import(
"from manila.i18n import _LW",
"manila.tests.unit/other_files5.py"))))
self.assertEqual(1, len(list(checks.check_explicit_underscore_import(
"msg = _('My message')",
"manila.tests.unit/other_files5.py"))))
# We are patching pep8 so that only the check under test is actually
# installed.
@mock.patch('pep8._checks',
{'physical_line': {}, 'logical_line': {}, 'tree': {}})
def _run_check(self, code, checker, filename=None):
pep8.register_check(checker)
lines = textwrap.dedent(code).strip().splitlines(True)
checker = pep8.Checker(filename=filename, lines=lines)
checker.check_all()
checker.report._deferred_print.sort()
return checker.report._deferred_print
def _assert_has_errors(self, code, checker, expected_errors=None,
filename=None):
actual_errors = [e[:3] for e in
self._run_check(code, checker, filename)]
self.assertEqual(expected_errors or [], actual_errors)
def _assert_has_no_errors(self, code, checker, filename=None):
self._assert_has_errors(code, checker, filename=filename)
def test_logging_format_no_tuple_arguments(self):
checker = checks.CheckLoggingFormatArgs
code = """
import logging
LOG = logging.getLogger()
LOG.info("Message without a second argument.")
LOG.critical("Message with %s arguments.", 'two')
LOG.debug("Volume %s caught fire and is at %d degrees C and"
" climbing.", 'volume1', 500)
"""
self._assert_has_no_errors(code, checker)
@ddt.data(*checks.CheckLoggingFormatArgs.LOG_METHODS)
def test_logging_with_tuple_argument(self, log_method):
checker = checks.CheckLoggingFormatArgs
code = """
import logging
LOG = logging.getLogger()
LOG.{0}("Volume %s caught fire and is at %d degrees C and "
"climbing.", ('volume1', 500))
"""
self._assert_has_errors(code.format(log_method), checker,
expected_errors=[(4, 21, 'M310')])
def test_str_on_exception(self):
checker = checks.CheckForStrUnicodeExc
code = """
def f(a, b):
try:
p = str(a) + str(b)
except ValueError as e:
p = str(e)
return p
"""
errors = [(5, 16, 'M325')]
self._assert_has_errors(code, checker, expected_errors=errors)
def test_no_str_unicode_on_exception(self):
checker = checks.CheckForStrUnicodeExc
code = """
def f(a, b):
try:
p = unicode(a) + str(b)
except ValueError as e:
p = e
return p
"""
self._assert_has_no_errors(code, checker)
def test_unicode_on_exception(self):
checker = checks.CheckForStrUnicodeExc
code = """
def f(a, b):
try:
p = str(a) + str(b)
except ValueError as e:
p = unicode(e)
return p
"""
errors = [(5, 20, 'M325')]
self._assert_has_errors(code, checker, expected_errors=errors)
def test_str_on_multiple_exceptions(self):
checker = checks.CheckForStrUnicodeExc
code = """
def f(a, b):
try:
p = str(a) + str(b)
except ValueError as e:
try:
p = unicode(a) + unicode(b)
except ValueError as ve:
p = str(e) + str(ve)
p = e
return p
"""
errors = [(8, 20, 'M325'), (8, 29, 'M325')]
self._assert_has_errors(code, checker, expected_errors=errors)
def test_str_unicode_on_multiple_exceptions(self):
checker = checks.CheckForStrUnicodeExc
code = """
def f(a, b):
try:
p = str(a) + str(b)
except ValueError as e:
try:
p = unicode(a) + unicode(b)
except ValueError as ve:
p = str(e) + unicode(ve)
p = str(e)
return p
"""
errors = [(8, 20, 'M325'), (8, 33, 'M325'), (9, 16, 'M325')]
self._assert_has_errors(code, checker, expected_errors=errors)
def test_trans_add(self):
checker = checks.CheckForTransAdd
code = """
def fake_tran(msg):
return msg
_ = fake_tran
_LI = _
_LW = _
_LE = _
_LC = _
def f(a, b):
msg = _('test') + 'add me'
msg = _LI('test') + 'add me'
msg = _LW('test') + 'add me'
msg = _LE('test') + 'add me'
msg = _LC('test') + 'add me'
msg = 'add to me' + _('test')
return msg
"""
# Python 3.4.0 introduced a change to the column calculation during AST
# parsing. This was reversed in Python 3.4.3, hence the version-based
# expected value calculation. See #1499743 for more background.
if sys.version_info < (3, 4, 0) or sys.version_info >= (3, 4, 3):
errors = [(13, 10, 'M326'), (14, 10, 'M326'), (15, 10, 'M326'),
(16, 10, 'M326'), (17, 10, 'M326'), (18, 24, 'M326')]
else:
errors = [(13, 11, 'M326'), (14, 13, 'M326'), (15, 13, 'M326'),
(16, 13, 'M326'), (17, 13, 'M326'), (18, 25, 'M326')]
self._assert_has_errors(code, checker, expected_errors=errors)
code = """
def f(a, b):
msg = 'test' + 'add me'
return msg
"""
errors = []
self._assert_has_errors(code, checker, expected_errors=errors)
def test_dict_constructor_with_list_copy(self):
self.assertEqual(1, len(list(checks.dict_constructor_with_list_copy(
" dict([(i, connect_info[i])"))))
self.assertEqual(1, len(list(checks.dict_constructor_with_list_copy(
" attrs = dict([(k, _from_json(v))"))))
self.assertEqual(1, len(list(checks.dict_constructor_with_list_copy(
" type_names = dict((value, key) for key, value in"))))
self.assertEqual(1, len(list(checks.dict_constructor_with_list_copy(
" dict((value, key) for key, value in"))))
self.assertEqual(1, len(list(checks.dict_constructor_with_list_copy(
"foo(param=dict((k, v) for k, v in bar.items()))"))))
self.assertEqual(1, len(list(checks.dict_constructor_with_list_copy(
" dict([[i,i] for i in range(3)])"))))
self.assertEqual(1, len(list(checks.dict_constructor_with_list_copy(
" dd = dict([i,i] for i in range(3))"))))
self.assertEqual(0, len(list(checks.dict_constructor_with_list_copy(
" create_kwargs = dict(snapshot=snapshot,"))))
self.assertEqual(0, len(list(checks.dict_constructor_with_list_copy(
" self._render_dict(xml, data_el, data.__dict__)"))))
def test_no_xrange(self):
self.assertEqual(1, len(list(checks.no_xrange("xrange(45)"))))
self.assertEqual(0, len(list(checks.no_xrange("range(45)"))))
def test_validate_assertTrue(self):
test_value = True
self.assertEqual(0, len(list(checks.validate_assertTrue(
"assertTrue(True)"))))
self.assertEqual(1, len(list(checks.validate_assertTrue(
"assertEqual(True, %s)" % test_value))))
def test_validate_assertIsNone(self):
test_value = None
self.assertEqual(0, len(list(checks.validate_assertIsNone(
"assertIsNone(None)"))))
self.assertEqual(1, len(list(checks.validate_assertIsNone(
"assertEqual(None, %s)" % test_value))))
def test_check_uuid4(self):
code = """
fake_uuid = uuid.uuid4()
"""
errors = [(1, 0, 'M354')]
self._assert_has_errors(code, checks.check_uuid4,
expected_errors=errors)
code = """
hex_uuid = uuid.uuid4().hex
"""
self._assert_has_no_errors(code, checks.check_uuid4)
def test_no_log_warn_check(self):
self.assertEqual(0, len(list(checks.no_log_warn_check(
"LOG.warning('This should not trigger LOG.warn "
"hacking check.')"))))
self.assertEqual(1, len(list(checks.no_log_warn_check(
"LOG.warn('We should not use LOG.warn')"))))
foo = """
LOG.warn('Catch me too, please'
)
"""
self.assertEqual(1, len(list(checks.no_log_warn_check(foo))))