| # Copyright (C) 2014-2019 Apple Inc. All rights reserved. |
| # |
| # Redistribution and use in source and binary forms, with or without |
| # modification, are permitted provided that the following conditions |
| # are met: |
| # 1. Redistributions of source code must retain the above copyright |
| # notice, this list of conditions and the following disclaimer. |
| # 2. Redistributions in binary form must reproduce the above copyright |
| # notice, this list of conditions and the following disclaimer in the |
| # documentation and/or other materials provided with the distribution. |
| # |
| # THIS SOFTWARE IS PROVIDED BY APPLE INC. AND ITS CONTRIBUTORS ``AS IS'' AND |
| # ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE IMPLIED |
| # WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE ARE |
| # DISCLAIMED. IN NO EVENT SHALL APPLE INC. OR ITS CONTRIBUTORS BE LIABLE FOR |
| # ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL |
| # DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS OR |
| # SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION) HOWEVER |
| # CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT LIABILITY, |
| # OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY OUT OF THE USE |
| # OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE. |
| |
| import logging |
| import os |
| import time |
| |
| from webkitpy.common.memoized import memoized |
| from webkitpy.common.system.crashlogs import CrashLogs |
| from webkitpy.common.system.executive import ScriptError |
| from webkitpy.port.apple import ApplePort |
| from webkitpy.port.leakdetector import LeakDetector |
| |
| |
| _log = logging.getLogger(__name__) |
| |
| |
| class DarwinPort(ApplePort): |
| |
| CURRENT_VERSION = None |
| SDK = None |
| |
| def __init__(self, host, port_name, **kwargs): |
| ApplePort.__init__(self, host, port_name, **kwargs) |
| |
| self._leak_detector = LeakDetector(self) |
| if self.get_option("leaks"): |
| # DumpRenderTree slows down noticably if we run more than about 1000 tests in a batch |
| # with MallocStackLogging enabled. |
| self.set_option_default("batch_size", 1000) |
| |
| def default_timeout_ms(self): |
| if self.get_option('guard_malloc'): |
| return 350 * 1000 |
| return super(DarwinPort, self).default_timeout_ms() |
| |
| def _port_specific_expectations_files(self, device_type=None): |
| return list(reversed([self._filesystem.join(self._webkit_baseline_path(p), 'TestExpectations') for p in self.baseline_search_path(device_type=device_type)])) |
| |
| def check_for_leaks(self, process_name, process_id): |
| if not self.get_option('leaks'): |
| return |
| |
| # We could use http://code.google.com/p/psutil/ to get the process_name from the pid. |
| self._leak_detector.check_for_leaks(process_name, process_id) |
| |
| def print_leaks_summary(self): |
| if not self.get_option('leaks'): |
| return |
| # We're in the manager process, so the leak detector will not have a valid list of leak files. |
| # FIXME: This is a hack, but we don't have a better way to get this information from the workers yet. |
| # FIXME: This will include too many leaks in subsequent runs until the results directory is cleared! |
| leaks_files = self._leak_detector.leaks_files_in_directory(self.results_directory()) |
| if not leaks_files: |
| return |
| total_bytes_string, unique_leaks = self._leak_detector.count_total_bytes_and_unique_leaks(leaks_files) |
| total_leaks = self._leak_detector.count_total_leaks(leaks_files) |
| _log.info("%s total leaks found for a total of %s." % (total_leaks, total_bytes_string)) |
| _log.info("%s unique leaks found." % unique_leaks) |
| |
| def _path_to_webcore_library(self): |
| return self._build_path('WebCore.framework/Versions/A/WebCore') |
| |
| def show_results_html_file(self, results_filename): |
| # We don't use self._run_script() because we don't want to wait for the script |
| # to exit and we want the output to show up on stdout in case there are errors |
| # launching the browser. |
| with open(os.devnull) as devnull: |
| self._executive.popen([self.path_to_script('run-safari')] + self._arguments_for_configuration() + ['--no-saved-state', '-NSOpen', results_filename], |
| cwd=self.webkit_base(), stdout=devnull, stderr=devnull) |
| |
| @memoized |
| def path_to_crash_logs(self): |
| log_directory = self.host.filesystem.expanduser('~') |
| log_directory = self.host.filesystem.join(log_directory, 'Library', 'Logs') |
| diagnositc_reports_directory = self.host.filesystem.join(log_directory, 'DiagnosticReports') |
| if self.host.filesystem.exists(diagnositc_reports_directory): |
| return diagnositc_reports_directory |
| return self.host.filesystem.join(log_directory, 'CrashReporter') |
| |
| def _merge_crash_logs(self, logs, new_logs, crashed_processes): |
| for test, crash_log in new_logs.items(): |
| try: |
| if test.split('-')[0] == 'Sandbox': |
| process_name = test.split('-')[1] |
| pid = int(test.split('-')[2]) |
| else: |
| process_name = test.split('-')[0] |
| pid = int(test.split('-')[1]) |
| except (IndexError, ValueError): |
| continue |
| if not any(entry[1] == process_name and entry[2] == pid for entry in crashed_processes): |
| # if this is a new crash, then append the logs |
| logs[test] = crash_log |
| return logs |
| |
| def _look_for_all_crash_logs_in_log_dir(self, newer_than): |
| crash_log = CrashLogs(self.host, self.path_to_crash_logs(), crash_logs_to_skip=self._crash_logs_to_skip_for_host.get(self.host, [])) |
| return crash_log.find_all_logs(newer_than=newer_than) |
| |
| def _get_crash_log(self, name, pid, stdout, stderr, newer_than, time_fn=None, sleep_fn=None, wait_for_log=True, target_host=None): |
| # Note that we do slow-spin here and wait, since it appears the time |
| # ReportCrash takes to actually write and flush the file varies when there are |
| # lots of simultaneous crashes going on. |
| time_fn = time_fn or time.time |
| sleep_fn = sleep_fn or time.sleep |
| crash_log = '' |
| crash_logs = CrashLogs(target_host or self.host, self.path_to_crash_logs(), crash_logs_to_skip=self._crash_logs_to_skip_for_host.get(target_host or self.host, [])) |
| now = time_fn() |
| deadline = now + 5 * int(self.get_option('child_processes', 1)) |
| while not crash_log and now <= deadline: |
| crash_log = crash_logs.find_newest_log(name, pid, include_errors=True, newer_than=newer_than) |
| if not wait_for_log: |
| break |
| if not crash_log or not [line for line in crash_log.splitlines() if not line.startswith('ERROR')]: |
| sleep_fn(0.1) |
| now = time_fn() |
| |
| if not crash_log: |
| return (stderr, None) |
| return (stderr, crash_log) |
| |
| def look_for_new_crash_logs(self, crashed_processes, start_time): |
| """Since crash logs can take a long time to be written out if the system is |
| under stress do a second pass at the end of the test run. |
| |
| crashes: test_name -> pid, process_name tuple of crashed process |
| start_time: time the tests started at. We're looking for crash |
| logs after that time. |
| """ |
| crash_logs = {} |
| for (test_name, process_name, pid) in crashed_processes: |
| # Passing None for output. This is a second pass after the test finished so |
| # if the output had any logging we would have already collected it. |
| crash_log = self._get_crash_log(process_name, pid, None, None, start_time, wait_for_log=False)[1] |
| if not crash_log: |
| continue |
| crash_logs[test_name] = crash_log |
| all_crash_log = self._look_for_all_crash_logs_in_log_dir(start_time) |
| return self._merge_crash_logs(crash_logs, all_crash_log, crashed_processes) |
| |
| def sample_process(self, name, pid, target_host=None): |
| host = target_host or self.host |
| tempdir = host.filesystem.mkdtemp() |
| temp_tailspin_file_path = host.filesystem.join(str(tempdir), "{0}-{1}-tailspin-temp.txt".format(name, pid)) |
| command = [ |
| '/usr/bin/tailspin', |
| 'save', |
| '-n', |
| temp_tailspin_file_path, |
| ] |
| if host.platform.is_mac(): |
| command = ['/usr/bin/sudo', '-n'] + command |
| |
| exit_status = host.executive.run_command(command, return_exit_code=True) |
| if not exit_status: # Symbolicate tailspin log using spindump |
| spindump_command = [ |
| '/usr/sbin/spindump', |
| '-i', temp_tailspin_file_path, |
| '-file', DarwinPort.tailspin_file_path(host, name, pid, str(tempdir)), |
| ] |
| try: |
| exit_code = host.executive.run_command(spindump_command + ['-noBulkSymbolication'], return_exit_code=True) |
| |
| # FIXME: Remove the fallback when we no longer support Catalina. |
| if exit_code: |
| host.executive.run_command(spindump_command) |
| host.filesystem.move_to_base_host(DarwinPort.tailspin_file_path(host, name, pid, str(tempdir)), |
| DarwinPort.tailspin_file_path(self.host, name, pid, self.results_directory())) |
| except (IOError, ScriptError) as e: |
| _log.warning('Unable to symbolicate tailspin log of process:' + str(e)) |
| else: # Tailspin failed, run sample instead |
| try: |
| host.executive.run_command([ |
| '/usr/bin/sample', |
| pid, |
| 10, |
| 10, |
| '-file', |
| DarwinPort.sample_file_path(host, name, pid, str(tempdir)), |
| ]) |
| host.filesystem.move_to_base_host(DarwinPort.sample_file_path(host, name, pid, str(tempdir)), |
| DarwinPort.sample_file_path(self.host, name, pid, self.results_directory())) |
| except ScriptError as e: |
| _log.warning('Unable to sample process:' + str(e)) |
| host.filesystem.rmtree(str(tempdir)) |
| |
| @staticmethod |
| def sample_file_path(host, name, pid, directory): |
| return host.filesystem.join(directory, "{0}-{1}-sample.txt".format(name, pid)) |
| |
| @staticmethod |
| def tailspin_file_path(host, name, pid, directory): |
| return host.filesystem.join(directory, "{0}-{1}-tailspin.txt".format(name, pid)) |
| |
| def look_for_new_samples(self, unresponsive_processes, start_time): |
| sample_files = {} |
| for (test_name, process_name, pid) in unresponsive_processes: |
| sample_file = DarwinPort.sample_file_path(self.host, process_name, pid, self.results_directory()) |
| if self._filesystem.isfile(sample_file): |
| sample_files[test_name] = sample_file |
| else: |
| tailspin_file = DarwinPort.tailspin_file_path(self.host, process_name, pid, self.results_directory()) |
| if self._filesystem.isfile(tailspin_file): |
| sample_files[test_name] = tailspin_file |
| return sample_files |
| |
| def _path_to_image_diff(self): |
| # ImageDiff for DarwinPorts is a little complicated. It will either be in |
| # a directory named ../mac relative to the port build directory, in a directory |
| # named ../<build-type> relative to the port build directory or in the port build directory |
| _image_diff_in_build_path = super(DarwinPort, self)._path_to_image_diff() |
| _port_build_dir = self.host.filesystem.dirname(_image_diff_in_build_path) |
| |
| # Test ../mac |
| _path_to_test = self.host.filesystem.join(_port_build_dir, '..', 'mac', 'ImageDiff') |
| if self.host.filesystem.exists(_path_to_test): |
| return _path_to_test |
| |
| # Test ../<build-type> |
| _build_type = self.host.filesystem.basename(_port_build_dir).split('-')[0] |
| _path_to_test = self.host.filesystem.join(_port_build_dir, '..', _build_type, 'ImageDiff') |
| if self.host.filesystem.exists(_path_to_test): |
| return _path_to_test |
| |
| return _image_diff_in_build_path |
| |
| def make_command(self): |
| return self.xcrun_find('make', '/usr/bin/make') |
| |
| def xcrun_find(self, command, fallback=None): |
| fallback = fallback or command |
| try: |
| return self._executive.run_command(['xcrun', '--sdk', self.SDK, '-find', command]).rstrip() |
| except ScriptError: |
| _log.warn("xcrun failed; falling back to '%s'." % fallback) |
| return fallback |
| |
| @memoized |
| def _plist_data_from_bundle(self, app_bundle, entry): |
| plist_path = self._filesystem.join(app_bundle, 'Info.plist') |
| if not self._filesystem.exists(plist_path): |
| plist_path = self._filesystem.join(app_bundle, 'Contents', 'Info.plist') |
| if not self._filesystem.exists(plist_path): |
| return None |
| return self._executive.run_command(['/usr/libexec/PlistBuddy', '-c', 'Print {}'.format(entry), plist_path]).rstrip() |
| |
| def app_identifier_from_bundle(self, app_bundle): |
| return self._plist_data_from_bundle(app_bundle, 'CFBundleIdentifier') |
| |
| def app_executable_from_bundle(self, app_bundle): |
| return self._plist_data_from_bundle(app_bundle, 'CFBundleExecutable') |
| |
| def environment_for_api_tests(self): |
| environment = super(DarwinPort, self).environment_for_api_tests() |
| build_root_path = str(self._build_path()) |
| for name in ['DYLD_LIBRARY_PATH', '__XPC_DYLD_LIBRARY_PATH', 'DYLD_FRAMEWORK_PATH', '__XPC_DYLD_FRAMEWORK_PATH']: |
| self._append_value_colon_separated(environment, name, build_root_path) |
| return environment |