mirror of
https://github.com/u-boot/u-boot.git
synced 2025-04-22 04:44:46 +00:00

When printing a python tool's test results, the entire list of failed tests and their tracebacks are reprinted for every failed test. This makes the test output quite unreadable. Fix the loop to print failures and tracebacks one at a time. Signed-off-by: Alper Nebi Yasak <alpernebiyasak@gmail.com> Reviewed-by: Simon Glass <sjg@chromium.org>
198 lines
7.3 KiB
Python
198 lines
7.3 KiB
Python
# SPDX-License-Identifier: GPL-2.0+
|
|
#
|
|
# Copyright (c) 2016 Google, Inc
|
|
#
|
|
|
|
from contextlib import contextmanager
|
|
import doctest
|
|
import glob
|
|
import multiprocessing
|
|
import os
|
|
import sys
|
|
import unittest
|
|
|
|
from patman import command
|
|
|
|
from io import StringIO
|
|
|
|
use_concurrent = True
|
|
try:
|
|
from concurrencytest.concurrencytest import ConcurrentTestSuite
|
|
from concurrencytest.concurrencytest import fork_for_tests
|
|
except:
|
|
use_concurrent = False
|
|
|
|
|
|
def run_test_coverage(prog, filter_fname, exclude_list, build_dir, required=None,
|
|
extra_args=None):
|
|
"""Run tests and check that we get 100% coverage
|
|
|
|
Args:
|
|
prog: Program to run (with be passed a '-t' argument to run tests
|
|
filter_fname: Normally all *.py files in the program's directory will
|
|
be included. If this is not None, then it is used to filter the
|
|
list so that only filenames that don't contain filter_fname are
|
|
included.
|
|
exclude_list: List of file patterns to exclude from the coverage
|
|
calculation
|
|
build_dir: Build directory, used to locate libfdt.py
|
|
required: List of modules which must be in the coverage report
|
|
extra_args (str): Extra arguments to pass to the tool before the -t/test
|
|
arg
|
|
|
|
Raises:
|
|
ValueError if the code coverage is not 100%
|
|
"""
|
|
# This uses the build output from sandbox_spl to get _libfdt.so
|
|
path = os.path.dirname(prog)
|
|
if filter_fname:
|
|
glob_list = glob.glob(os.path.join(path, '*.py'))
|
|
glob_list = [fname for fname in glob_list if filter_fname in fname]
|
|
else:
|
|
glob_list = []
|
|
glob_list += exclude_list
|
|
glob_list += ['*libfdt.py', '*site-packages*', '*dist-packages*']
|
|
glob_list += ['*concurrencytest*']
|
|
test_cmd = 'test' if 'binman' in prog or 'patman' in prog else '-t'
|
|
prefix = ''
|
|
if build_dir:
|
|
prefix = 'PYTHONPATH=$PYTHONPATH:%s/sandbox_spl/tools ' % build_dir
|
|
cmd = ('%spython3-coverage run '
|
|
'--omit "%s" %s %s %s -P1' % (prefix, ','.join(glob_list),
|
|
prog, extra_args or '', test_cmd))
|
|
os.system(cmd)
|
|
stdout = command.output('python3-coverage', 'report')
|
|
lines = stdout.splitlines()
|
|
if required:
|
|
# Convert '/path/to/name.py' just the module name 'name'
|
|
test_set = set([os.path.splitext(os.path.basename(line.split()[0]))[0]
|
|
for line in lines if '/etype/' in line])
|
|
missing_list = required
|
|
missing_list.discard('__init__')
|
|
missing_list.difference_update(test_set)
|
|
if missing_list:
|
|
print('Missing tests for %s' % (', '.join(missing_list)))
|
|
print(stdout)
|
|
ok = False
|
|
|
|
coverage = lines[-1].split(' ')[-1]
|
|
ok = True
|
|
print(coverage)
|
|
if coverage != '100%':
|
|
print(stdout)
|
|
print("Type 'python3-coverage html' to get a report in "
|
|
'htmlcov/index.html')
|
|
print('Coverage error: %s, but should be 100%%' % coverage)
|
|
ok = False
|
|
if not ok:
|
|
raise ValueError('Test coverage failure')
|
|
|
|
|
|
# Use this to suppress stdout/stderr output:
|
|
# with capture_sys_output() as (stdout, stderr)
|
|
# ...do something...
|
|
@contextmanager
|
|
def capture_sys_output():
|
|
capture_out, capture_err = StringIO(), StringIO()
|
|
old_out, old_err = sys.stdout, sys.stderr
|
|
try:
|
|
sys.stdout, sys.stderr = capture_out, capture_err
|
|
yield capture_out, capture_err
|
|
finally:
|
|
sys.stdout, sys.stderr = old_out, old_err
|
|
|
|
|
|
def report_result(toolname:str, test_name: str, result: unittest.TestResult):
|
|
"""Report the results from a suite of tests
|
|
|
|
Args:
|
|
toolname: Name of the tool that ran the tests
|
|
test_name: Name of test that was run, or None for all
|
|
result: A unittest.TestResult object containing the results
|
|
"""
|
|
# Remove errors which just indicate a missing test. Since Python v3.5 If an
|
|
# ImportError or AttributeError occurs while traversing name then a
|
|
# synthetic test that raises that error when run will be returned. These
|
|
# errors are included in the errors accumulated by result.errors.
|
|
if test_name:
|
|
errors = []
|
|
|
|
for test, err in result.errors:
|
|
if ("has no attribute '%s'" % test_name) not in err:
|
|
errors.append((test, err))
|
|
result.testsRun -= 1
|
|
result.errors = errors
|
|
|
|
print(result)
|
|
for test, err in result.errors:
|
|
print(test.id(), err)
|
|
for test, err in result.failures:
|
|
print(test.id(), err)
|
|
if result.skipped:
|
|
print('%d %s test%s SKIPPED:' % (len(result.skipped), toolname,
|
|
's' if len(result.skipped) > 1 else ''))
|
|
for skip_info in result.skipped:
|
|
print('%s: %s' % (skip_info[0], skip_info[1]))
|
|
if result.errors or result.failures:
|
|
print('%s tests FAILED' % toolname)
|
|
return 1
|
|
return 0
|
|
|
|
|
|
def run_test_suites(result, debug, verbosity, test_preserve_dirs, processes,
|
|
test_name, toolpath, class_and_module_list):
|
|
"""Run a series of test suites and collect the results
|
|
|
|
Args:
|
|
result: A unittest.TestResult object to add the results to
|
|
debug: True to enable debugging, which shows a full stack trace on error
|
|
verbosity: Verbosity level to use (0-4)
|
|
test_preserve_dirs: True to preserve the input directory used by tests
|
|
so that it can be examined afterwards (only useful for debugging
|
|
tests). If a single test is selected (in args[0]) it also preserves
|
|
the output directory for this test. Both directories are displayed
|
|
on the command line.
|
|
processes: Number of processes to use to run tests (None=same as #CPUs)
|
|
test_name: Name of test to run, or None for all
|
|
toolpath: List of paths to use for tools
|
|
class_and_module_list: List of test classes (type class) and module
|
|
names (type str) to run
|
|
"""
|
|
for module in class_and_module_list:
|
|
if isinstance(module, str) and (not test_name or test_name == module):
|
|
suite = doctest.DocTestSuite(module)
|
|
suite.run(result)
|
|
|
|
sys.argv = [sys.argv[0]]
|
|
if debug:
|
|
sys.argv.append('-D')
|
|
if verbosity:
|
|
sys.argv.append('-v%d' % verbosity)
|
|
if toolpath:
|
|
for path in toolpath:
|
|
sys.argv += ['--toolpath', path]
|
|
|
|
suite = unittest.TestSuite()
|
|
loader = unittest.TestLoader()
|
|
for module in class_and_module_list:
|
|
if isinstance(module, str):
|
|
continue
|
|
# Test the test module about our arguments, if it is interested
|
|
if hasattr(module, 'setup_test_args'):
|
|
setup_test_args = getattr(module, 'setup_test_args')
|
|
setup_test_args(preserve_indir=test_preserve_dirs,
|
|
preserve_outdirs=test_preserve_dirs and test_name is not None,
|
|
toolpath=toolpath, verbosity=verbosity)
|
|
if test_name:
|
|
try:
|
|
suite.addTests(loader.loadTestsFromName(test_name, module))
|
|
except AttributeError:
|
|
continue
|
|
else:
|
|
suite.addTests(loader.loadTestsFromTestCase(module))
|
|
if use_concurrent and processes != 1:
|
|
concurrent_suite = ConcurrentTestSuite(suite,
|
|
fork_for_tests(processes or multiprocessing.cpu_count()))
|
|
concurrent_suite.run(result)
|
|
else:
|
|
suite.run(result)
|