esp-idf/conftest.py

333 wiersze
11 KiB
Python

# SPDX-FileCopyrightText: 2021-2022 Espressif Systems (Shanghai) CO LTD
# SPDX-License-Identifier: Apache-2.0
# pylint: disable=W0621 # redefined-outer-name
# This file is a pytest root configuration file and provide the following functionalities:
# 1. Defines a few fixtures that could be used under the whole project.
# 2. Defines a few hook functions.
#
# IDF is using [pytest](https://github.com/pytest-dev/pytest) and
# [pytest-embedded plugin](https://github.com/espressif/pytest-embedded) as its example test framework.
#
# This is an experimental feature, and if you found any bug or have any question, please report to
# https://github.com/espressif/pytest-embedded/issues
import logging
import os
import sys
import xml.etree.ElementTree as ET
from datetime import datetime
from fnmatch import fnmatch
from typing import Callable, List, Optional, Tuple
import pytest
from _pytest.config import Config
from _pytest.fixtures import FixtureRequest
from _pytest.main import Session
from _pytest.nodes import Item
from _pytest.python import Function
from _pytest.reports import TestReport
from _pytest.runner import CallInfo
from _pytest.terminal import TerminalReporter
from pytest_embedded.plugin import apply_count, parse_configuration
from pytest_embedded.utils import find_by_suffix
SUPPORTED_TARGETS = ['esp32', 'esp32s2', 'esp32c3', 'esp32s3']
PREVIEW_TARGETS = ['linux', 'esp32h2', 'esp32c2']
DEFAULT_SDKCONFIG = 'default'
##################
# Help Functions #
##################
def is_target_marker(marker: str) -> bool:
if marker.startswith('esp32'):
return True
if marker.startswith('esp8'):
return True
return False
def format_case_id(target: Optional[str], config: Optional[str], case: str) -> str:
return f'{target}.{config}.{case}'
def item_marker_names(item: Item) -> List[str]:
return [marker.name for marker in item.iter_markers()]
############
# Fixtures #
############
_TEST_SESSION_TMPDIR = os.path.join(
os.path.dirname(__file__),
'pytest_embedded_log',
datetime.now().strftime('%Y-%m-%d_%H-%M-%S'),
)
os.makedirs(_TEST_SESSION_TMPDIR, exist_ok=True)
@pytest.fixture(scope='session', autouse=True)
def session_tempdir() -> str:
return _TEST_SESSION_TMPDIR
@pytest.fixture
@parse_configuration
def config(request: FixtureRequest) -> str:
return getattr(request, 'param', None) or DEFAULT_SDKCONFIG
@pytest.fixture
def test_func_name(request: FixtureRequest) -> str:
return request.node.function.__name__ # type: ignore
@pytest.fixture
def test_case_name(request: FixtureRequest, target: str, config: str) -> str:
return format_case_id(target, config, request.node.originalname)
@pytest.fixture
@apply_count
def build_dir(app_path: str, target: Optional[str], config: Optional[str]) -> str:
"""
Check local build dir with the following priority:
1. build_<target>_<config>
2. build_<target>
3. build_<config>
4. build
Args:
app_path: app path
target: target
config: config
Returns:
valid build directory
"""
check_dirs = []
if target is not None and config is not None:
check_dirs.append(f'build_{target}_{config}')
if target is not None:
check_dirs.append(f'build_{target}')
if config is not None:
check_dirs.append(f'build_{config}')
check_dirs.append('build')
for check_dir in check_dirs:
binary_path = os.path.join(app_path, check_dir)
if os.path.isdir(binary_path):
logging.info(f'find valid binary path: {binary_path}')
return check_dir
logging.warning(
'checking binary path: %s... missing... try another place', binary_path
)
recommend_place = check_dirs[0]
logging.error(
f'no build dir valid. Please build the binary via "idf.py -B {recommend_place} build" and run pytest again'
)
sys.exit(1)
@pytest.fixture(autouse=True)
@apply_count
def junit_properties(
test_case_name: str, record_xml_attribute: Callable[[str, object], None]
) -> None:
"""
This fixture is autoused and will modify the junit report test case name to <target>.<config>.<case_name>
"""
record_xml_attribute('name', test_case_name)
##################
# Hook functions #
##################
def pytest_addoption(parser: pytest.Parser) -> None:
base_group = parser.getgroup('idf')
base_group.addoption(
'--sdkconfig',
help='sdkconfig postfix, like sdkconfig.ci.<config>. (Default: None, which would build all found apps)',
)
base_group.addoption(
'--known-failure-cases-file', help='known failure cases file path'
)
_idf_pytest_embedded_key = pytest.StashKey['IdfPytestEmbedded']
def pytest_configure(config: Config) -> None:
config.stash[_idf_pytest_embedded_key] = IdfPytestEmbedded(
target=config.getoption('target'),
sdkconfig=config.getoption('sdkconfig'),
known_failure_cases_file=config.getoption('known_failure_cases_file'),
)
config.pluginmanager.register(config.stash[_idf_pytest_embedded_key])
def pytest_unconfigure(config: Config) -> None:
_pytest_embedded = config.stash.get(_idf_pytest_embedded_key, None)
if _pytest_embedded:
del config.stash[_idf_pytest_embedded_key]
config.pluginmanager.unregister(_pytest_embedded)
class IdfPytestEmbedded:
def __init__(
self,
target: Optional[str] = None,
sdkconfig: Optional[str] = None,
known_failure_cases_file: Optional[str] = None,
):
# CLI options to filter the test cases
self.target = target
self.sdkconfig = sdkconfig
self.known_failure_patterns = self._parse_known_failure_cases_file(
known_failure_cases_file
)
self._failed_cases: List[
Tuple[str, bool]
] = [] # (test_case_name, is_known_failure_cases)
@property
def failed_cases(self) -> List[str]:
return [case for case, is_known in self._failed_cases if not is_known]
@property
def known_failure_cases(self) -> List[str]:
return [case for case, is_known in self._failed_cases if is_known]
@staticmethod
def _parse_known_failure_cases_file(
known_failure_cases_file: Optional[str] = None,
) -> List[str]:
if not known_failure_cases_file or not os.path.isfile(known_failure_cases_file):
return []
patterns = []
with open(known_failure_cases_file) as fr:
for line in fr.readlines():
if not line:
continue
if not line.strip():
continue
without_comments = line.split('#')[0].strip()
if without_comments:
patterns.append(without_comments)
return patterns
@pytest.hookimpl(tryfirst=True)
def pytest_sessionstart(self, session: Session) -> None:
if self.target:
self.target = self.target.lower()
session.config.option.target = self.target
@pytest.hookimpl(tryfirst=True)
def pytest_collection_modifyitems(self, items: List[Function]) -> None:
# sort by file path and callspec.config
# implement like this since this is a limitation of pytest, couldn't get fixture values while collecting
# https://github.com/pytest-dev/pytest/discussions/9689
def _get_param_config(_item: Function) -> str:
if hasattr(_item, 'callspec'):
return _item.callspec.params.get('config', DEFAULT_SDKCONFIG) # type: ignore
return DEFAULT_SDKCONFIG
items.sort(key=lambda x: (os.path.dirname(x.path), _get_param_config(x)))
# add markers for special markers
for item in items:
if 'supported_targets' in item_marker_names(item):
for _target in SUPPORTED_TARGETS:
item.add_marker(_target)
if 'preview_targets' in item_marker_names(item):
for _target in PREVIEW_TARGETS:
item.add_marker(_target)
if 'all_targets' in item_marker_names(item):
for _target in [*SUPPORTED_TARGETS, *PREVIEW_TARGETS]:
item.add_marker(_target)
# filter all the test cases with "--target"
if self.target:
items[:] = [
item for item in items if self.target in item_marker_names(item)
]
# filter all the test cases with cli option "config"
if self.sdkconfig:
items[:] = [
item for item in items if _get_param_config(item) == self.sdkconfig
]
def pytest_runtest_makereport(
self, item: Function, call: CallInfo[None]
) -> Optional[TestReport]:
if call.when == 'setup':
return None
report = TestReport.from_item_and_call(item, call)
if report.outcome == 'failed':
test_case_name = item.funcargs.get('test_case_name', '')
is_known_failure = self._is_known_failure(test_case_name)
self._failed_cases.append((test_case_name, is_known_failure))
return report
def _is_known_failure(self, case_id: str) -> bool:
for pattern in self.known_failure_patterns:
if case_id == pattern:
return True
if fnmatch(case_id, pattern):
return True
return False
@pytest.hookimpl(trylast=True)
def pytest_runtest_teardown(self, item: Function) -> None:
"""
Format the test case generated junit reports
"""
tempdir = item.funcargs.get('test_case_tempdir')
if not tempdir:
return
junits = find_by_suffix('.xml', tempdir)
if not junits:
return
target = item.funcargs['target']
config = item.funcargs['config']
for junit in junits:
xml = ET.parse(junit)
testcases = xml.findall('.//testcase')
for case in testcases:
case.attrib['name'] = format_case_id(
target, config, case.attrib['name']
)
if 'file' in case.attrib:
case.attrib['file'] = case.attrib['file'].replace(
'/IDF/', ''
) # our unity test framework
xml.write(junit)
def pytest_sessionfinish(self, session: Session, exitstatus: int) -> None:
if exitstatus != 0 and self.known_failure_cases and not self.failed_cases:
session.exitstatus = 0
def pytest_terminal_summary(self, terminalreporter: TerminalReporter) -> None:
if self.known_failure_cases:
terminalreporter.section('Known failure cases', bold=True, yellow=True)
terminalreporter.line('\n'.join(self.known_failure_cases))
if self.failed_cases:
terminalreporter.section('Failed cases', bold=True, red=True)
terminalreporter.line('\n'.join(self.failed_cases))