blob: 70ea2321fd91ca67997762bde9b8af75cc8aa80a [file] [log] [blame]
# Copyright 2014 The Chromium Authors
# Use of this source code is governed by a BSD-style license that can be
# found in the LICENSE file.
from __future__ import print_function
import argparse
import codecs
import contextlib
import json
import os
import logging
import platform
import subprocess
import sys
import tempfile
import time
import traceback
logging.basicConfig(level=logging.INFO)
# Add src/testing/ into sys.path for importing xvfb and test_env.
sys.path.append(
os.path.abspath(os.path.join(os.path.dirname(__file__), os.path.pardir)))
import test_env
if sys.platform.startswith('linux'):
import xvfb
SCRIPT_DIR = os.path.abspath(os.path.dirname(__file__))
SRC_DIR = os.path.abspath(
os.path.join(SCRIPT_DIR, os.path.pardir, os.path.pardir))
# Use result_sink.py in //build/util/lib/results/ for uploading the
# results of non-isolated script tests.
BUILD_UTIL_DIR = os.path.join(SRC_DIR, 'build', 'util')
sys.path.insert(0, BUILD_UTIL_DIR)
try:
from lib.results import result_sink
from lib.results import result_types
except ImportError:
# Some build-time scripts import this file and run into issues with
# result_sink's dependency on requests since we can't depend on vpython
# during build-time. So silently swallow the error in that case.
result_sink = None
# run_web_tests.py returns the number of failures as the return
# code, but caps the return code at 101 to avoid overflow or colliding
# with reserved values from the shell.
MAX_FAILURES_EXIT_STATUS = 101
# Exit code to indicate infrastructure issue.
INFRA_FAILURE_EXIT_CODE = 87
# ACL might be explicitly set or inherited.
CORRECT_ACL_VARIANTS = [
'APPLICATION PACKAGE AUTHORITY' \
'\\ALL RESTRICTED APPLICATION PACKAGES:(OI)(CI)(RX)', \
'APPLICATION PACKAGE AUTHORITY' \
'\\ALL RESTRICTED APPLICATION PACKAGES:(I)(OI)(CI)(RX)'
]
# pylint: disable=useless-object-inheritance
def set_lpac_acls(acl_dir, is_test_script=False):
"""Sets LPAC ACLs on a directory. Windows 10 only."""
if platform.release() != '10':
return
try:
existing_acls = subprocess.check_output(['icacls', acl_dir],
stderr=subprocess.STDOUT,
universal_newlines=True)
except subprocess.CalledProcessError as e:
logging.error('Failed to retrieve existing ACLs for directory %s', acl_dir)
logging.error('Command output: %s', e.output)
sys.exit(e.returncode)
acls_correct = False
for acl in CORRECT_ACL_VARIANTS:
if acl in existing_acls:
acls_correct = True
if not acls_correct:
try:
existing_acls = subprocess.check_output(
['icacls', acl_dir, '/grant', '*S-1-15-2-2:(OI)(CI)(RX)'],
stderr=subprocess.STDOUT)
except subprocess.CalledProcessError as e:
logging.error(
'Failed to retrieve existing ACLs for directory %s', acl_dir)
logging.error('Command output: %s', e.output)
sys.exit(e.returncode)
if not is_test_script:
return
# Bots running on luci use hardlinks that do not have correct ACLs so these
# must be manually overridden here.
with temporary_file() as tempfile_path:
subprocess.check_output(
['icacls', acl_dir, '/save', tempfile_path, '/t', '/q', '/c'],
stderr=subprocess.STDOUT)
# ACL files look like this, e.g. for c:\a\b\c\d\Release_x64
#
# Release_x64
# D:AI(A;OICI;0x1200a9;;;S-1-15-2-2)(A;OICIID;FA;;;BA)
# Release_x64\icudtl_extra.dat
# D:AI(A;ID;0x1200a9;;;S-1-15-2-2)(A;ID;FA;;;BA)(A;ID;0x1301bf;;;BU)
with codecs.open(tempfile_path, encoding='utf_16_le') as aclfile:
for filename in aclfile:
acl = next(aclfile).strip()
full_filename = os.path.abspath(
os.path.join(acl_dir, os.pardir, filename.strip()))
if 'S-1-15-2-2' in acl:
continue
if os.path.isdir(full_filename):
continue
subprocess.check_output(
['icacls', full_filename, '/grant', '*S-1-15-2-2:(RX)'],
stderr=subprocess.STDOUT)
def run_script(argv, funcs):
def parse_json(path):
with open(path) as f:
return json.load(f)
parser = argparse.ArgumentParser()
# TODO(phajdan.jr): Make build-config-fs required after passing it in recipe.
parser.add_argument('--build-config-fs')
parser.add_argument('--paths', type=parse_json, default={})
# Properties describe the environment of the build, and are the same per
# script invocation.
parser.add_argument('--properties', type=parse_json, default={})
# Args contains per-invocation arguments that potentially change the
# behavior of the script.
parser.add_argument('--args', type=parse_json, default=[])
subparsers = parser.add_subparsers()
run_parser = subparsers.add_parser('run')
run_parser.add_argument(
'--output', type=argparse.FileType('w'), required=True)
run_parser.add_argument('--filter-file', type=argparse.FileType('r'))
run_parser.set_defaults(func=funcs['run'])
run_parser = subparsers.add_parser('compile_targets')
run_parser.add_argument(
'--output', type=argparse.FileType('w'), required=True)
run_parser.set_defaults(func=funcs['compile_targets'])
args = parser.parse_args(argv)
return args.func(args)
def run_command(argv, env=None, cwd=None):
print('Running %r in %r (env: %r)' % (argv, cwd, env), file=sys.stderr)
rc = test_env.run_command(argv, env=env, cwd=cwd)
print('Command %r returned exit code %d' % (argv, rc), file=sys.stderr)
return rc
@contextlib.contextmanager
def temporary_file():
fd, path = tempfile.mkstemp()
os.close(fd)
try:
yield path
finally:
os.remove(path)
def record_local_script_results(name, output_fd, failures, valid):
"""Records to a local json file and to RDB the results of the script test.
For legacy reasons, local script tests (ie: script tests that run
locally and that don't conform to the isolated-test API) are expected to
record their results using a specific format. This method encapsulates
that format and also uploads those results to Result DB.
Args:
name: Name of the script test.
output_fd: A .write()-supporting file descriptor to write results to.
failures: List of strings representing test failures.
valid: Whether the results are valid.
"""
local_script_results = {
'valid': valid,
'failures': failures
}
json.dump(local_script_results, output_fd)
if not result_sink:
return
result_sink_client = result_sink.TryInitClient()
if not result_sink_client:
return
status = result_types.PASS
if not valid:
status = result_types.UNKNOWN
elif failures:
status = result_types.FAIL
test_log = '\n'.join(failures)
result_sink_client.Post(name, status, None, test_log, None)
def parse_common_test_results(json_results, test_separator='/'):
def convert_trie_to_flat_paths(trie, prefix=None):
# Also see blinkpy.web_tests.layout_package.json_results_generator
result = {}
for name, data in trie.items():
if prefix:
name = prefix + test_separator + name
if len(data) and not 'actual' in data and not 'expected' in data:
result.update(convert_trie_to_flat_paths(data, name))
else:
result[name] = data
return result
results = {
'passes': {},
'unexpected_passes': {},
'failures': {},
'unexpected_failures': {},
'flakes': {},
'unexpected_flakes': {},
}
# TODO(dpranke): crbug.com/357866 - we should simplify the handling of
# both the return code and parsing the actual results, below.
passing_statuses = ('PASS', 'SLOW', 'NEEDSREBASELINE')
for test, result in convert_trie_to_flat_paths(
json_results['tests']).items():
key = 'unexpected_' if result.get('is_unexpected') else ''
data = result['actual']
actual_results = data.split()
last_result = actual_results[-1]
expected_results = result['expected'].split()
if (len(actual_results) > 1 and
(last_result in expected_results or last_result in passing_statuses)):
key += 'flakes'
elif last_result in passing_statuses:
key += 'passes'
# TODO(dpranke): crbug.com/357867 ... Why are we assigning result
# instead of actual_result here. Do we even need these things to be
# hashes, or just lists?
data = result
else:
key += 'failures'
results[key][test] = data
return results
def write_interrupted_test_results_to(filepath, test_start_time):
"""Writes a test results JSON file* to filepath.
This JSON file is formatted to explain that something went wrong.
*src/docs/testing/json_test_results_format.md
Args:
filepath: A path to a file to write the output to.
test_start_time: The start time of the test run expressed as a
floating-point offset in seconds from the UNIX epoch.
"""
with open(filepath, 'w') as fh:
output = {
'interrupted': True,
'num_failures_by_type': {},
'seconds_since_epoch': test_start_time,
'tests': {},
'version': 3,
}
json.dump(output, fh)
def get_gtest_summary_passes(output):
"""Returns a mapping of test to boolean indicating if the test passed.
Only partially parses the format. This code is based on code in tools/build,
specifically
https://chromium.googlesource.com/chromium/tools/build/+/17fef98756c5f250b20bf716829a0004857235ff/scripts/slave/recipe_modules/test_utils/util.py#189
"""
if not output:
return {}
mapping = {}
for cur_iteration_data in output.get('per_iteration_data', []):
for test_fullname, results in cur_iteration_data.items():
# Results is a list with one entry per test try. Last one is the final
# result.
last_result = results[-1]
if last_result['status'] == 'SUCCESS':
mapping[test_fullname] = True
elif last_result['status'] != 'SKIPPED':
mapping[test_fullname] = False
return mapping
def extract_filter_list(filter_list):
"""Helper for isolated script test wrappers. Parses the
--isolated-script-test-filter command line argument. Currently, double-colon
('::') is used as the separator between test names, because a single colon may
be used in the names of perf benchmarks, which contain URLs.
"""
return filter_list.split('::')
def add_emulator_args(parser):
parser.add_argument(
'--avd-config',
type=os.path.realpath,
help=('Path to the avd config. Required for Android products. '
'(See //tools/android/avd/proto for message definition '
'and existing *.textpb files.)'))
parser.add_argument(
'--emulator-window',
action='store_true',
default=False,
help='Enable graphical window display on the emulator.')
class BaseIsolatedScriptArgsAdapter:
"""The base class for all script adapters that need to translate flags
set by isolated script test contract into the specific test script's flags.
"""
def __init__(self):
self._parser = argparse.ArgumentParser()
self._options = None
self._rest_args = None
self._script_writes_output_json = None
self._parser.add_argument(
'--isolated-outdir', type=str,
required=False,
help='value of $ISOLATED_OUTDIR from swarming task')
self._parser.add_argument(
'--isolated-script-test-output', type=os.path.abspath,
required=False,
help='path to write test results JSON object to')
self._parser.add_argument(
'--isolated-script-test-filter', type=str,
required=False)
self._parser.add_argument(
'--isolated-script-test-repeat', type=int,
required=False)
self._parser.add_argument(
'--isolated-script-test-launcher-retry-limit', type=int,
required=False)
self._parser.add_argument(
'--isolated-script-test-also-run-disabled-tests',
default=False, action='store_true', required=False)
self._parser.add_argument(
'--xvfb',
help='start xvfb. Ignored on unsupported platforms',
action='store_true')
# Used to create the correct subclass.
self._parser.add_argument(
'--script-type', choices=['isolated', 'typ', 'bare'],
help='Which script adapter to use')
# Arguments that are ignored, but added here because it's easier to ignore
# them to to update bot configs to not pass them.
self._parser.add_argument('--isolated-script-test-chartjson-output')
self._parser.add_argument('--isolated-script-test-perf-output')
def parse_args(self, args=None):
self._options, self._rest_args = self._parser.parse_known_args(args)
@property
def parser(self):
return self._parser
@property
def options(self):
return self._options
@property
def rest_args(self):
return self._rest_args
def generate_test_output_args(self, output):
del output # unused
return []
def generate_test_filter_args(self, test_filter_str):
del test_filter_str # unused
raise RuntimeError('Flag not supported.')
def generate_test_repeat_args(self, repeat_count):
del repeat_count # unused
raise RuntimeError('Flag not supported.')
def generate_test_launcher_retry_limit_args(self, retry_limit):
del retry_limit # unused
raise RuntimeError('Flag not supported.')
def generate_sharding_args(self, total_shards, shard_index):
del total_shards, shard_index # unused
raise RuntimeError('Flag not supported.')
def generate_test_also_run_disabled_tests_args(self):
raise RuntimeError('Flag not supported.')
def select_python_executable(self):
return sys.executable
def generate_isolated_script_cmd(self):
isolated_script_cmd = [ self.select_python_executable() ] + self.rest_args
if self.options.isolated_script_test_output:
output_args = self.generate_test_output_args(
self.options.isolated_script_test_output)
self._script_writes_output_json = bool(output_args)
isolated_script_cmd += output_args
# Augment test filter args if needed
if self.options.isolated_script_test_filter:
isolated_script_cmd += self.generate_test_filter_args(
self.options.isolated_script_test_filter)
# Augment test repeat if needed
if self.options.isolated_script_test_repeat is not None:
isolated_script_cmd += self.generate_test_repeat_args(
self.options.isolated_script_test_repeat)
# Augment test launcher retry limit args if needed
if self.options.isolated_script_test_launcher_retry_limit is not None:
isolated_script_cmd += self.generate_test_launcher_retry_limit_args(
self.options.isolated_script_test_launcher_retry_limit)
# Augment test also run disable tests args if needed
if self.options.isolated_script_test_also_run_disabled_tests:
isolated_script_cmd += self.generate_test_also_run_disabled_tests_args()
# Augment shard args if needed
env = os.environ.copy()
total_shards = None
shard_index = None
if 'GTEST_TOTAL_SHARDS' in env:
total_shards = int(env['GTEST_TOTAL_SHARDS'])
if 'GTEST_SHARD_INDEX' in env:
shard_index = int(env['GTEST_SHARD_INDEX'])
if total_shards is not None and shard_index is not None:
isolated_script_cmd += self.generate_sharding_args(
total_shards, shard_index)
return isolated_script_cmd
def clean_up_after_test_run(self):
pass
def do_pre_test_run_tasks(self):
pass
def do_post_test_run_tasks(self):
pass
def _write_simple_test_results(self, start_time, exit_code):
if exit_code is None:
failure_type = 'CRASH'
elif exit_code == 0:
failure_type = 'PASS'
else:
failure_type = 'FAIL'
test_name = os.path.basename(self._rest_args[0])
# See //docs/testing/json_test_results_format.md
results_json = {
'version': 3,
'interrupted': False,
'num_failures_by_type': { failure_type: 1 },
'path_delimiter': '/',
'seconds_since_epoch': start_time,
'tests': {
test_name: {
'expected': 'PASS',
'actual': failure_type,
'time': time.time() - start_time,
},
},
}
with open(self.options.isolated_script_test_output, 'w') as fp:
json.dump(results_json, fp)
def run_test(self, cwd=None):
self.parse_args()
cmd = self.generate_isolated_script_cmd()
self.do_pre_test_run_tasks()
env = os.environ.copy()
env['CHROME_HEADLESS'] = '1'
print('Running command: %s\nwith env: %r' % (
' '.join(cmd), env))
sys.stdout.flush()
start_time = time.time()
try:
if self.options.xvfb and sys.platform.startswith('linux'):
exit_code = xvfb.run_executable(cmd, env, cwd=cwd)
else:
exit_code = test_env.run_command(cmd, env=env, cwd=cwd, log=False)
print('Command returned exit code %d' % exit_code)
sys.stdout.flush()
self.do_post_test_run_tasks()
except Exception:
traceback.print_exc()
exit_code = None
finally:
self.clean_up_after_test_run()
if (self.options.isolated_script_test_output
and not self._script_writes_output_json):
self._write_simple_test_results(start_time, exit_code)
return exit_code if exit_code is not None else 2