1# Copyright 2016 The Chromium OS Authors. All rights reserved. 2# Use of this source code is governed by a BSD-style license that can be 3# found in the LICENSE file. 4 5# repohooks/pre-upload.py currently does not run pylint. But for developers who 6# want to check their code manually we disable several harmless pylint warnings 7# which just distract from more serious remaining issues. 8# 9# The instance variables _host and _install_paths are not defined in __init__(). 10# pylint: disable=attribute-defined-outside-init 11# 12# Many short variable names don't follow the naming convention. 13# pylint: disable=invalid-name 14# 15# _parse_result() and _dir_size() don't access self and could be functions. 16# pylint: disable=no-self-use 17# 18# _ChromeLogin and _TradefedLogCollector have no public methods. 19# pylint: disable=too-few-public-methods 20 21import contextlib 22import errno 23import glob 24import hashlib 25import lockfile 26import logging 27import os 28import pipes 29import random 30import re 31import shutil 32import stat 33import tempfile 34import urlparse 35 36from autotest_lib.client.bin import utils as client_utils 37from autotest_lib.client.common_lib import utils as common_utils 38from autotest_lib.client.common_lib import error 39from autotest_lib.client.common_lib.cros import dev_server 40from autotest_lib.server import autotest 41from autotest_lib.server import test 42from autotest_lib.server import utils 43 44# TODO(ihf): If akeshet doesn't fix crbug.com/691046 delete metrics again. 45try: 46 from chromite.lib import metrics 47except ImportError: 48 metrics = utils.metrics_mock 49 50# TODO(ihf): Find a home for all these paths. This is getting out of hand. 51_SDK_TOOLS_DIR_M = 'gs://chromeos-arc-images/builds/git_mnc-dr-arc-dev-linux-static_sdk_tools/3554341' 52_SDK_TOOLS_FILES = ['aapt'] 53# To stabilize adb behavior, we use dynamically linked adb. 54_ADB_DIR_M = 'gs://chromeos-arc-images/builds/git_mnc-dr-arc-dev-linux-cheets_arm-user/3554341' 55_ADB_FILES = ['adb'] 56 57_ADB_POLLING_INTERVAL_SECONDS = 1 58_ADB_READY_TIMEOUT_SECONDS = 60 59_ANDROID_ADB_KEYS_PATH = '/data/misc/adb/adb_keys' 60 61_ARC_POLLING_INTERVAL_SECONDS = 1 62_ARC_READY_TIMEOUT_SECONDS = 60 63 64_TRADEFED_PREFIX = 'autotest-tradefed-install_' 65_TRADEFED_CACHE_LOCAL = '/tmp/autotest-tradefed-cache' 66_TRADEFED_CACHE_CONTAINER = '/usr/local/autotest/results/shared/cache' 67_TRADEFED_CACHE_CONTAINER_LOCK = '/usr/local/autotest/results/shared/lock' 68 69# According to dshi a drone has 500GB of disk space. It is ok for now to use 70# 10GB of disk space, as no more than 10 tests should run in parallel. 71# TODO(ihf): Investigate tighter cache size. 72_TRADEFED_CACHE_MAX_SIZE = (10 * 1024 * 1024 * 1024) 73 74 75class _ChromeLogin(object): 76 """Context manager to handle Chrome login state.""" 77 78 def __init__(self, host, cts_helper_kwargs): 79 self._host = host 80 self._cts_helper_kwargs = cts_helper_kwargs 81 82 def __enter__(self): 83 """Logs in to the Chrome.""" 84 logging.info('Ensure Android is running...') 85 # If we can't login to Chrome and launch Android we want this job to 86 # die roughly after 5 minutes instead of hanging for the duration. 87 autotest.Autotest(self._host).run_timed_test('cheets_StartAndroid', 88 timeout=300, 89 check_client_result=True, 90 **self._cts_helper_kwargs) 91 92 def __exit__(self, exc_type, exc_value, traceback): 93 """On exit, to wipe out all the login state, reboot the machine. 94 95 @param exc_type: Exception type if an exception is raised from the 96 with-block. 97 @param exc_value: Exception instance if an exception is raised from 98 the with-block. 99 @param traceback: Stack trace info if an exception is raised from 100 the with-block. 101 @return None, indicating not to ignore an exception from the with-block 102 if raised. 103 """ 104 logging.info('Rebooting...') 105 try: 106 self._host.reboot() 107 except Exception: 108 if exc_type is None: 109 raise 110 # If an exception is raise from the with-block, just record the 111 # exception for the rebooting to avoid ignoring the original 112 # exception. 113 logging.exception('Rebooting failed.') 114 115 116@contextlib.contextmanager 117def lock(filename): 118 """Prevents other autotest/tradefed instances from accessing cache. 119 120 @param filename: The file to be locked. 121 """ 122 filelock = lockfile.FileLock(filename) 123 # It is tempting just to call filelock.acquire(3600). But the implementation 124 # has very poor temporal granularity (timeout/10), which is unsuitable for 125 # our needs. See /usr/lib64/python2.7/site-packages/lockfile/ 126 attempts = 0 127 while not filelock.i_am_locking(): 128 try: 129 attempts += 1 130 logging.info('Waiting for cache lock...') 131 filelock.acquire(random.randint(1, 5)) 132 except (lockfile.AlreadyLocked, lockfile.LockTimeout): 133 if attempts > 1000: 134 # Normally we should aqcuire the lock in a few seconds. Once we 135 # wait on the order of hours either the dev server IO is 136 # overloaded or a lock didn't get cleaned up. Take one for the 137 # team, break the lock and report a failure. This should fix 138 # the lock for following tests. If the failure affects more than 139 # one job look for a deadlock or dev server overload. 140 logging.error('Permanent lock failure. Trying to break lock.') 141 filelock.break_lock() 142 raise error.TestFail('Error: permanent cache lock failure.') 143 else: 144 logging.info('Acquired cache lock after %d attempts.', attempts) 145 try: 146 yield 147 finally: 148 filelock.release() 149 logging.info('Released cache lock.') 150 151 152@contextlib.contextmanager 153def adb_keepalive(target, extra_paths): 154 """A context manager that keeps the adb connection alive. 155 156 AdbKeepalive will spin off a new process that will continuously poll for 157 adb's connected state, and will attempt to reconnect if it ever goes down. 158 This is the only way we can currently recover safely from (intentional) 159 reboots. 160 161 @param target: the hostname and port of the DUT. 162 @param extra_paths: any additional components to the PATH environment 163 variable. 164 """ 165 from autotest_lib.client.common_lib.cros import adb_keepalive as module 166 # |__file__| returns the absolute path of the compiled bytecode of the 167 # module. We want to run the original .py file, so we need to change the 168 # extension back. 169 script_filename = module.__file__.replace('.pyc', '.py') 170 job = common_utils.BgJob([script_filename, target], 171 nickname='adb_keepalive', stderr_level=logging.DEBUG, 172 stdout_tee=common_utils.TEE_TO_LOGS, 173 stderr_tee=common_utils.TEE_TO_LOGS, 174 extra_paths=extra_paths) 175 176 try: 177 yield 178 finally: 179 # The adb_keepalive.py script runs forever until SIGTERM is sent. 180 common_utils.nuke_subprocess(job.sp) 181 common_utils.join_bg_jobs([job]) 182 183 184@contextlib.contextmanager 185def pushd(d): 186 """Defines pushd. 187 @param d: the directory to change to. 188 """ 189 current = os.getcwd() 190 os.chdir(d) 191 try: 192 yield 193 finally: 194 os.chdir(current) 195 196 197def parse_tradefed_v2_result(result, waivers=None): 198 """Check the result from the tradefed-v2 output. 199 200 @param result: The result stdout string from the tradefed command. 201 @param waivers: a set() of tests which are permitted to fail. 202 @return 5-tuple (tests, passed, failed, notexecuted, waived) 203 """ 204 # Regular expressions for start/end messages of each test-run chunk. 205 abi_re = r'armeabi-v7a|x86' 206 # TODO(kinaba): use the current running module name. 207 module_re = r'\S+' 208 start_re = re.compile(r'(?:Start|Continu)ing (%s) %s with' 209 r' (\d+(?:,\d+)?) test' % (abi_re, module_re)) 210 end_re = re.compile(r'(%s) %s (?:complet|fail)ed in .*\.' 211 r' (\d+) passed, (\d+) failed, (\d+) not executed' 212 % (abi_re, module_re)) 213 214 # Records the result per each ABI. 215 total_test = dict() 216 total_pass = dict() 217 total_fail = dict() 218 last_notexec = dict() 219 220 # ABI and the test count for the current chunk. 221 abi = None 222 ntest = None 223 prev_npass = prev_nfail = prev_nnotexec = None 224 225 for line in result.splitlines(): 226 # Beginning of a chunk of tests. 227 match = start_re.search(line) 228 if match: 229 if abi: 230 raise error.TestFail('Error: Unexpected test start: ' + line) 231 abi = match.group(1) 232 ntest = int(match.group(2).replace(',','')) 233 prev_npass = prev_nfail = prev_nnotexec = None 234 else: 235 # End of the current chunk. 236 match = end_re.search(line) 237 if not match: 238 continue 239 240 npass, nfail, nnotexec = map(int, match.group(2,3,4)) 241 if abi != match.group(1): 242 # When the last case crashed during teardown, tradefed emits two 243 # end-messages with possibly increased fail count. Ignore it. 244 if (prev_npass == npass and (prev_nfail == nfail or 245 prev_nfail == nfail - 1) and prev_nnotexec == nnotexec): 246 continue 247 raise error.TestFail('Error: Unexpected test end: ' + line) 248 prev_npass, prev_nfail, prev_nnotexec = npass, nfail, nnotexec 249 250 # When the test crashes too ofen, tradefed seems to finish the 251 # iteration by running "0 tests, 0 passed, ...". Do not count 252 # that in. 253 if ntest > 0: 254 total_test[abi] = (total_test.get(abi, 0) + ntest - 255 last_notexec.get(abi, 0)) 256 total_pass[abi] = total_pass.get(abi, 0) + npass 257 total_fail[abi] = total_fail.get(abi, 0) + nfail 258 last_notexec[abi] = nnotexec 259 abi = None 260 261 if abi: 262 raise error.TestFail('Error: No end message for the last chunk.') 263 264 # TODO(rohitbm): make failure parsing more robust by extracting the list 265 # of failing tests instead of searching in the result blob. As well as 266 # only parse for waivers for the running ABI. 267 waived = 0 268 if waivers: 269 abis = total_test.keys() 270 for testname in waivers: 271 # TODO(dhaddock): Find a more robust way to apply waivers. 272 fail_count = (result.count(testname + ' FAIL') + 273 result.count(testname + ' fail')) 274 if fail_count: 275 if fail_count > len(abis): 276 # This should be an error.TestFail, but unfortunately 277 # tradefed has a bug that emits "fail" twice when a 278 # test failed during teardown. It will anyway causes 279 # a test count inconsistency and visible on the dashboard. 280 logging.error('Found %d failures for %s ' 281 'but there are only %d abis: %s', 282 fail_count, testname, len(abis), abis) 283 waived += fail_count 284 logging.info('Waived failure for %s %d time(s)', 285 testname, fail_count) 286 counts = tuple(sum(count_per_abi.values()) for count_per_abi in 287 (total_test, total_pass, total_fail, last_notexec)) + (waived,) 288 msg = ('tests=%d, passed=%d, failed=%d, not_executed=%d, waived=%d' % 289 counts) 290 logging.info(msg) 291 if counts[2] - waived < 0: 292 raise error.TestFail('Error: Internal waiver bookkeeping has ' 293 'become inconsistent (%s)' % msg) 294 return counts 295 296 297def select_32bit_java(): 298 """Switches to 32 bit java if installed (like in lab lxc images) to save 299 about 30-40% server/shard memory during the run.""" 300 if utils.is_in_container() and not client_utils.is_moblab(): 301 java = '/usr/lib/jvm/java-8-openjdk-i386' 302 if os.path.exists(java): 303 logging.info('Found 32 bit java, switching to use it.') 304 os.environ['JAVA_HOME'] = java 305 os.environ['PATH'] = (os.path.join(java, 'bin') + os.pathsep + 306 os.environ['PATH']) 307 308 309class TradefedTest(test.test): 310 """Base class to prepare DUT to run tests via tradefed.""" 311 version = 1 312 313 # TODO(ihf): Remove _ABD_DIR_M/_SDK_TOOLS_DIR_M defaults once M is dead. 314 def initialize(self, host=None, adb_dir=_ADB_DIR_M, 315 sdk_tools_dir=_SDK_TOOLS_DIR_M): 316 """Sets up the tools and binary bundles for the test.""" 317 logging.info('Hostname: %s', host.hostname) 318 self._host = host 319 self._install_paths = [] 320 # Tests in the lab run within individual lxc container instances. 321 if utils.is_in_container(): 322 cache_root = _TRADEFED_CACHE_CONTAINER 323 else: 324 cache_root = _TRADEFED_CACHE_LOCAL 325 326 # TODO(ihf): reevaluate this again when we run out of memory. We could 327 # for example use 32 bit java on the first run but not during retries. 328 # b/62895114. If select_32bit_java gets deleted for good also remove it 329 # from the base image. 330 # Try to save server memory (crbug.com/717413). 331 # select_32bit_java() 332 333 # Quick sanity check and spew of java version installed on the server. 334 utils.run('java', args=('-version',), ignore_status=False, verbose=True, 335 stdout_tee=utils.TEE_TO_LOGS, stderr_tee=utils.TEE_TO_LOGS) 336 # The content of the cache survives across jobs. 337 self._safe_makedirs(cache_root) 338 self._tradefed_cache = os.path.join(cache_root, 'cache') 339 self._tradefed_cache_lock = os.path.join(cache_root, 'lock') 340 # The content of the install location does not survive across jobs and 341 # is isolated (by using a unique path)_against other autotest instances. 342 # This is not needed for the lab, but if somebody wants to run multiple 343 # TradedefTest instance. 344 self._tradefed_install = tempfile.mkdtemp(prefix=_TRADEFED_PREFIX) 345 # Under lxc the cache is shared between multiple autotest/tradefed 346 # instances. We need to synchronize access to it. All binaries are 347 # installed through the (shared) cache into the local (unshared) 348 # lxc/autotest instance storage. 349 # If clearing the cache it must happen before all downloads. 350 self._clear_download_cache_if_needed() 351 # Set permissions (rwxr-xr-x) to the executable binaries. 352 permission = (stat.S_IRWXU | stat.S_IRGRP | stat.S_IXGRP | stat.S_IROTH 353 | stat.S_IXOTH) 354 self._install_files(adb_dir, _ADB_FILES, permission) 355 self._install_files(sdk_tools_dir, _SDK_TOOLS_FILES, permission) 356 357 def cleanup(self): 358 """Cleans up any dirtied state.""" 359 # Kill any lingering adb servers. 360 self._run('adb', verbose=True, args=('kill-server',)) 361 logging.info('Cleaning up %s.', self._tradefed_install) 362 shutil.rmtree(self._tradefed_install) 363 364 def _login_chrome(self, **cts_helper_kwargs): 365 """Returns Chrome log-in context manager. 366 367 Please see also cheets_StartAndroid for details about how this works. 368 """ 369 return _ChromeLogin(self._host, cts_helper_kwargs) 370 371 def _get_adb_target(self): 372 return '{}:{}'.format(self._host.hostname, self._host.port) 373 374 def _try_adb_connect(self): 375 """Attempts to connect to adb on the DUT. 376 377 @return boolean indicating if adb connected successfully. 378 """ 379 # This may fail return failure due to a race condition in adb connect 380 # (b/29370989). If adb is already connected, this command will 381 # immediately return success. 382 hostport = self._get_adb_target() 383 result = self._run( 384 'adb', 385 args=('connect', hostport), 386 verbose=True, 387 ignore_status=True) 388 logging.info('adb connect {}:\n{}'.format(hostport, result.stdout)) 389 if result.exit_status != 0: 390 return False 391 392 result = self._run('adb', args=('devices',)) 393 logging.info('adb devices:\n' + result.stdout) 394 if not re.search( 395 r'{}\s+(device|unauthorized)'.format(re.escape(hostport)), 396 result.stdout): 397 return False 398 399 # Actually test the connection with an adb command as there can be 400 # a race between detecting the connected device and actually being 401 # able to run a commmand with authenticated adb. 402 result = self._run('adb', args=('shell', 'exit'), ignore_status=True) 403 return result.exit_status == 0 404 405 def _android_shell(self, command): 406 """Run a command remotely on the device in an android shell 407 408 This function is strictly for internal use only, as commands do not run 409 in a fully consistent Android environment. Prefer adb shell instead. 410 """ 411 self._host.run('android-sh -c ' + pipes.quote(command)) 412 413 def _write_android_file(self, filename, data): 414 """Writes a file to a location relative to the android container. 415 416 This is an internal function used to bootstrap adb. 417 Tests should use adb push to write files. 418 """ 419 android_cmd = 'echo %s > %s' % (pipes.quote(data), 420 pipes.quote(filename)) 421 self._android_shell(android_cmd) 422 423 def _connect_adb(self): 424 """Sets up ADB connection to the ARC container.""" 425 logging.info('Setting up adb connection.') 426 # Generate and push keys for adb. 427 # TODO(elijahtaylor): Extract this code to arc_common and de-duplicate 428 # code in arc.py on the client side tests. 429 key_path = os.path.join(self.tmpdir, 'test_key') 430 pubkey_path = key_path + '.pub' 431 self._run('adb', verbose=True, args=('keygen', pipes.quote(key_path))) 432 with open(pubkey_path, 'r') as f: 433 self._write_android_file(_ANDROID_ADB_KEYS_PATH, f.read()) 434 self._android_shell('restorecon ' + pipes.quote(_ANDROID_ADB_KEYS_PATH)) 435 os.environ['ADB_VENDOR_KEYS'] = key_path 436 437 # Kill existing adb server to ensure that the env var is picked up. 438 self._run('adb', verbose=True, args=('kill-server',)) 439 440 # This starts adbd. 441 self._android_shell('setprop sys.usb.config mtp,adb') 442 443 # Also let it be automatically started upon reboot. 444 self._android_shell('setprop persist.sys.usb.config mtp,adb') 445 446 # adbd may take some time to come up. Repeatedly try to connect to adb. 447 utils.poll_for_condition(lambda: self._try_adb_connect(), 448 exception=error.TestFail( 449 'Error: Failed to set up adb connection'), 450 timeout=_ADB_READY_TIMEOUT_SECONDS, 451 sleep_interval=_ADB_POLLING_INTERVAL_SECONDS) 452 453 logging.info('Successfully setup adb connection.') 454 455 def _wait_for_arc_boot(self): 456 """Wait until ARC is fully booted. 457 458 Tests for the presence of the intent helper app to determine whether ARC 459 has finished booting. 460 """ 461 def _intent_helper_running(): 462 result = self._run('adb', args=('shell', 'pgrep', '-f', 463 'org.chromium.arc.intent_helper')) 464 return bool(result.stdout) 465 utils.poll_for_condition( 466 _intent_helper_running, 467 exception=error.TestFail( 468 'Error: Timed out waiting for intent helper.'), 469 timeout=_ARC_READY_TIMEOUT_SECONDS, 470 sleep_interval=_ARC_POLLING_INTERVAL_SECONDS) 471 472 def _disable_adb_install_dialog(self): 473 """Disables a dialog shown on adb install execution. 474 475 By default, on adb install execution, "Allow Google to regularly check 476 device activity ... " dialog is shown. It requires manual user action 477 so that tests are blocked at the point. 478 This method disables it. 479 """ 480 logging.info('Disabling the adb install dialog.') 481 result = self._run( 482 'adb', 483 verbose=True, 484 args=( 485 'shell', 486 'settings', 487 'put', 488 'global', 489 'verifier_verify_adb_installs', 490 '0')) 491 logging.info('Disable adb dialog: %s', result.stdout) 492 493 def _ready_arc(self): 494 """Ready ARC and adb for running tests via tradefed.""" 495 self._connect_adb() 496 self._disable_adb_install_dialog() 497 self._wait_for_arc_boot() 498 499 def _safe_makedirs(self, path): 500 """Creates a directory at |path| and its ancestors. 501 502 Unlike os.makedirs(), ignore errors even if directories exist. 503 """ 504 try: 505 os.makedirs(path) 506 except OSError as e: 507 if not (e.errno == errno.EEXIST and os.path.isdir(path)): 508 raise 509 510 def _unzip(self, filename): 511 """Unzip the file. 512 513 The destination directory name will be the stem of filename. 514 E.g., _unzip('foo/bar/baz.zip') will create directory at 515 'foo/bar/baz', and then will inflate zip's content under the directory. 516 If here is already a directory at the stem, that directory will be used. 517 518 @param filename: Path to the zip archive. 519 @return Path to the inflated directory. 520 """ 521 destination = os.path.splitext(filename)[0] 522 if os.path.isdir(destination): 523 return destination 524 self._safe_makedirs(destination) 525 utils.run('unzip', args=('-d', destination, filename)) 526 return destination 527 528 def _dir_size(self, directory): 529 """Compute recursive size in bytes of directory.""" 530 size = 0 531 for root, _, files in os.walk(directory): 532 size += sum(os.path.getsize(os.path.join(root, name)) 533 for name in files) 534 return size 535 536 def _clear_download_cache_if_needed(self): 537 """Invalidates cache to prevent it from growing too large.""" 538 # If the cache is large enough to hold a working set, we can simply 539 # delete everything without thrashing. 540 # TODO(ihf): Investigate strategies like LRU. 541 with lock(self._tradefed_cache_lock): 542 size = self._dir_size(self._tradefed_cache) 543 if size > _TRADEFED_CACHE_MAX_SIZE: 544 logging.info('Current cache size=%d got too large. Clearing %s.' 545 , size, self._tradefed_cache) 546 shutil.rmtree(self._tradefed_cache) 547 self._safe_makedirs(self._tradefed_cache) 548 else: 549 logging.info('Current cache size=%d of %s.', size, 550 self._tradefed_cache) 551 552 def _download_to_cache(self, uri): 553 """Downloads the uri from the storage server. 554 555 It always checks the cache for available binaries first and skips 556 download if binaries are already in cache. 557 558 The caller of this function is responsible for holding the cache lock. 559 560 @param uri: The Google Storage or dl.google.com uri. 561 @return Path to the downloaded object, name. 562 """ 563 # Split uri into 3 pieces for use by gsutil and also by wget. 564 parsed = urlparse.urlparse(uri) 565 filename = os.path.basename(parsed.path) 566 # We are hashing the uri instead of the binary. This is acceptable, as 567 # the uris are supposed to contain version information and an object is 568 # not supposed to be changed once created. 569 output_dir = os.path.join(self._tradefed_cache, 570 hashlib.md5(uri).hexdigest()) 571 output = os.path.join(output_dir, filename) 572 # Check for existence of file. 573 if os.path.exists(output): 574 logging.info('Skipping download of %s, reusing %s.', uri, output) 575 return output 576 self._safe_makedirs(output_dir) 577 578 if parsed.scheme not in ['gs', 'http', 'https']: 579 raise error.TestFail('Error: Unknown download scheme %s' % 580 parsed.scheme) 581 if parsed.scheme in ['http', 'https']: 582 logging.info('Using wget to download %s to %s.', uri, output_dir) 583 # We are downloading 1 file at a time, hence using -O over -P. 584 utils.run( 585 'wget', 586 args=( 587 '--report-speed=bits', 588 '-O', 589 output, 590 uri), 591 verbose=True) 592 return output 593 594 if not client_utils.is_moblab(): 595 # If the machine can access to the storage server directly, 596 # defer to "gsutil" for downloading. 597 logging.info('Host %s not in lab. Downloading %s directly to %s.', 598 self._host.hostname, uri, output) 599 # b/17445576: gsutil rsync of individual files is not implemented. 600 utils.run('gsutil', args=('cp', uri, output), verbose=True) 601 return output 602 603 # We are in the moblab. Because the machine cannot access the storage 604 # server directly, use dev server to proxy. 605 logging.info('Host %s is in lab. Downloading %s by staging to %s.', 606 self._host.hostname, uri, output) 607 608 dirname = os.path.dirname(parsed.path) 609 archive_url = '%s://%s%s' % (parsed.scheme, parsed.netloc, dirname) 610 611 # First, request the devserver to download files into the lab network. 612 # TODO(ihf): Switch stage_artifacts to honor rsync. Then we don't have 613 # to shuffle files inside of tarballs. 614 info = self._host.host_info_store.get() 615 ds = dev_server.ImageServer.resolve(info.build) 616 ds.stage_artifacts(info.build, files=[filename], 617 archive_url=archive_url) 618 619 # Then download files from the dev server. 620 # TODO(ihf): use rsync instead of wget. Are there 3 machines involved? 621 # Itself, dev_server plus DUT? Or is there just no rsync in moblab? 622 ds_src = '/'.join([ds.url(), 'static', dirname, filename]) 623 logging.info('dev_server URL: %s', ds_src) 624 # Calls into DUT to pull uri from dev_server. 625 utils.run( 626 'wget', 627 args=( 628 '--report-speed=bits', 629 '-O', 630 output, 631 ds_src), 632 verbose=True) 633 return output 634 635 def _instance_copy(self, cache_path): 636 """Makes a copy of a file from the (shared) cache to a wholy owned 637 local instance. Also copies one level of cache directoy (MD5 named). 638 """ 639 filename = os.path.basename(cache_path) 640 dirname = os.path.basename(os.path.dirname(cache_path)) 641 instance_dir = os.path.join(self._tradefed_install, dirname) 642 # Make sure destination directory is named the same. 643 self._safe_makedirs(instance_dir) 644 instance_path = os.path.join(instance_dir, filename) 645 shutil.copyfile(cache_path, instance_path) 646 return instance_path 647 648 def _install_bundle(self, gs_uri): 649 """Downloads a zip file, installs it and returns the local path.""" 650 if not gs_uri.endswith('.zip'): 651 raise error.TestFail('Error: Not a .zip file %s.', gs_uri) 652 # Atomic write through of file. 653 with lock(self._tradefed_cache_lock): 654 cache_path = self._download_to_cache(gs_uri) 655 local = self._instance_copy(cache_path) 656 657 unzipped = self._unzip(local) 658 self._abi = 'x86' if 'x86-x86' in unzipped else 'arm' 659 return unzipped 660 661 def _install_files(self, gs_dir, files, permission): 662 """Installs binary tools.""" 663 for filename in files: 664 gs_uri = os.path.join(gs_dir, filename) 665 # Atomic write through of file. 666 with lock(self._tradefed_cache_lock): 667 cache_path = self._download_to_cache(gs_uri) 668 local = self._instance_copy(cache_path) 669 os.chmod(local, permission) 670 # Keep track of PATH. 671 self._install_paths.append(os.path.dirname(local)) 672 673 def _copy_media(self, media): 674 """Calls copy_media to push media files to DUT via adb.""" 675 logging.info('Copying media to device. This can take a few minutes.') 676 copy_media = os.path.join(media, 'copy_media.sh') 677 with pushd(media): 678 try: 679 self._run('file', args=('/bin/sh',), verbose=True, 680 ignore_status=True, timeout=60, 681 stdout_tee=utils.TEE_TO_LOGS, 682 stderr_tee=utils.TEE_TO_LOGS) 683 self._run('sh', args=('--version',), verbose=True, 684 ignore_status=True, timeout=60, 685 stdout_tee=utils.TEE_TO_LOGS, 686 stderr_tee=utils.TEE_TO_LOGS) 687 except: 688 logging.warning('Could not obtain sh version.') 689 self._run( 690 'sh', 691 args=('-e', copy_media, 'all'), 692 timeout=7200, # Wait at most 2h for download of media files. 693 verbose=True, 694 ignore_status=False, 695 stdout_tee=utils.TEE_TO_LOGS, 696 stderr_tee=utils.TEE_TO_LOGS) 697 698 def _verify_media(self, media): 699 """Verify that the local media directory matches the DUT. 700 Used for debugging b/32978387 where we may see file corruption.""" 701 # TODO(ihf): Remove function once b/32978387 is resolved. 702 # Find all files in the bbb_short and bbb_full directories, md5sum these 703 # files and sort by filename, both on the DUT and on the local tree. 704 logging.info('Computing md5 of remote media files.') 705 remote = self._run('adb', args=('shell', 706 'cd /sdcard/test; find ./bbb_short ./bbb_full -type f -print0 | ' 707 'xargs -0 md5sum | grep -v "\.DS_Store" | sort -k 2')) 708 logging.info('Computing md5 of local media files.') 709 local = self._run('/bin/sh', args=('-c', 710 ('cd %s; find ./bbb_short ./bbb_full -type f -print0 | ' 711 'xargs -0 md5sum | grep -v "\.DS_Store" | sort -k 2') % media)) 712 713 # 'adb shell' terminates lines with CRLF. Normalize before comparing. 714 if remote.stdout.replace('\r\n','\n') != local.stdout: 715 logging.error('Some media files differ on DUT /sdcard/test vs. local.') 716 logging.info('media=%s', media) 717 logging.error('remote=%s', remote) 718 logging.error('local=%s', local) 719 # TODO(ihf): Return False. 720 return True 721 logging.info('Media files identical on DUT /sdcard/test vs. local.') 722 return True 723 724 def _push_media(self, CTS_URI): 725 """Downloads, caches and pushed media files to DUT.""" 726 media = self._install_bundle(CTS_URI['media']) 727 base = os.path.splitext(os.path.basename(CTS_URI['media']))[0] 728 cts_media = os.path.join(media, base) 729 # TODO(ihf): this really should measure throughput in Bytes/s. 730 m = 'chromeos/autotest/infra_benchmark/cheets/push_media/duration' 731 fields = {'success': False, 732 'dut_host_name': self._host.hostname} 733 with metrics.SecondsTimer(m, fields=fields) as c: 734 self._copy_media(cts_media) 735 c['success'] = True 736 if not self._verify_media(cts_media): 737 raise error.TestFail('Error: saw corruption pushing media files.') 738 739 def _run(self, *args, **kwargs): 740 """Executes the given command line. 741 742 To support SDK tools, such as adb or aapt, this adds _install_paths 743 to the extra_paths. Before invoking this, ensure _install_files() has 744 been called. 745 """ 746 kwargs['extra_paths'] = ( 747 kwargs.get('extra_paths', []) + self._install_paths) 748 return utils.run(*args, **kwargs) 749 750 def _collect_tradefed_global_log(self, result, destination): 751 """Collects the tradefed global log. 752 753 @param result: The result object from utils.run. 754 @param destination: Autotest result directory (destination of logs). 755 """ 756 match = re.search(r'Saved log to /tmp/(tradefed_global_log_.*\.txt)', 757 result.stdout) 758 if not match: 759 logging.error('no tradefed_global_log file is found') 760 return 761 762 name = match.group(1) 763 dest = os.path.join(destination, 'logs', 'tmp') 764 self._safe_makedirs(dest) 765 shutil.copy(os.path.join('/tmp', name), os.path.join(dest, name)) 766 767 def _parse_tradefed_datetime(self, result, summary=None): 768 """Get the tradefed provided result ID consisting of a datetime stamp. 769 770 Unfortunately we are unable to tell tradefed where to store the results. 771 In the lab we have multiple instances of tradefed running in parallel 772 writing results and logs to the same base directory. This function 773 finds the identifier which tradefed used during the current run and 774 returns it for further processing of result files. 775 776 @param result: The result object from utils.run. 777 @param summary: Test result summary from runs so far. 778 @return datetime_id: The result ID chosen by tradefed. 779 Example: '2016.07.14_00.34.50'. 780 """ 781 # This string is show for both 'run' and 'continue' after all tests. 782 match = re.search(r': XML test result file generated at (\S+). Passed', 783 result.stdout) 784 if not (match and match.group(1)): 785 # TODO(ihf): Find out if we ever recover something interesting in 786 # this case. Otherwise delete it. 787 # Try harder to find the remains. This string shows before all 788 # tests but only with 'run', not 'continue'. 789 logging.warning('XML test result file incomplete?') 790 match = re.search(r': Created result dir (\S+)', result.stdout) 791 if not (match and match.group(1)): 792 error_msg = 'Test did not complete due to Chrome or ARC crash.' 793 if summary: 794 error_msg += (' Test summary from previous runs: %s' 795 % summary) 796 raise error.TestFail(error_msg) 797 datetime_id = match.group(1) 798 logging.info('Tradefed identified results and logs with %s.', 799 datetime_id) 800 return datetime_id 801 802 def _parse_tradefed_datetime_v2(self, result, summary=None): 803 """Get the tradefed provided result ID consisting of a datetime stamp. 804 805 Unfortunately we are unable to tell tradefed where to store the results. 806 In the lab we have multiple instances of tradefed running in parallel 807 writing results and logs to the same base directory. This function 808 finds the identifier which tradefed used during the current run and 809 returns it for further processing of result files. 810 811 @param result: The result object from utils.run. 812 @param summary: Test result summary from runs so far. 813 @return datetime_id: The result ID chosen by tradefed. 814 Example: '2016.07.14_00.34.50'. 815 """ 816 # This string is show for both 'run' and 'continue' after all tests. 817 match = re.search(r'(\d\d\d\d.\d\d.\d\d_\d\d.\d\d.\d\d)', result.stdout) 818 if not (match and match.group(1)): 819 error_msg = 'Error: Test did not complete. (Chrome or ARC crash?)' 820 if summary: 821 error_msg += (' Test summary from previous runs: %s' 822 % summary) 823 raise error.TestFail(error_msg) 824 datetime_id = match.group(1) 825 logging.info('Tradefed identified results and logs with %s.', 826 datetime_id) 827 return datetime_id 828 829 def _parse_result(self, result, waivers=None): 830 """Check the result from the tradefed output. 831 832 This extracts the test pass/fail/executed list from the output of 833 tradefed. It is up to the caller to handle inconsistencies. 834 835 @param result: The result object from utils.run. 836 @param waivers: a set() of tests which are permitted to fail. 837 """ 838 # Parse the stdout to extract test status. In particular step over 839 # similar output for each ABI and just look at the final summary. 840 match = re.search(r'(XML test result file generated at (\S+). ' 841 r'Passed (\d+), Failed (\d+), Not Executed (\d+))', 842 result.stdout) 843 if not match: 844 raise error.Test('Test log does not contain a summary.') 845 846 passed = int(match.group(3)) 847 failed = int(match.group(4)) 848 not_executed = int(match.group(5)) 849 match = re.search(r'(Start test run of (\d+) packages, containing ' 850 r'(\d+(?:,\d+)?) tests)', result.stdout) 851 if match and match.group(3): 852 tests = int(match.group(3).replace(',', '')) 853 else: 854 # Unfortunately this happens. Assume it made no other mistakes. 855 logging.warning('Tradefed forgot to print number of tests.') 856 tests = passed + failed + not_executed 857 # TODO(rohitbm): make failure parsing more robust by extracting the list 858 # of failing tests instead of searching in the result blob. As well as 859 # only parse for waivers for the running ABI. 860 waived = 0 861 if waivers: 862 for testname in waivers: 863 # TODO(dhaddock): Find a more robust way to apply waivers. 864 fail_count = result.stdout.count(testname + ' FAIL') 865 if fail_count: 866 if fail_count > 2: 867 raise error.TestFail('Error: There are too many ' 868 'failures found in the output to ' 869 'be valid for applying waivers. ' 870 'Please check output.') 871 waived += fail_count 872 logging.info('Waived failure for %s %d time(s)', 873 testname, fail_count) 874 logging.info( 875 'tests=%d, passed=%d, failed=%d, not_executed=%d, waived=%d', 876 tests, passed, failed, not_executed, waived) 877 if failed < waived: 878 raise error.TestFail('Error: Internal waiver book keeping has ' 879 'become inconsistent.') 880 return (tests, passed, failed, not_executed, waived) 881 882 def _parse_result_v2(self, result, waivers=None): 883 """Check the result from the tradefed-v2 output. 884 885 This extracts the test pass/fail/executed list from the output of 886 tradefed. It is up to the caller to handle inconsistencies. 887 888 @param result: The result object from utils.run. 889 @param waivers: a set() of tests which are permitted to fail. 890 """ 891 return parse_tradefed_v2_result(result.stdout, waivers) 892 893 def _collect_logs(self, repository, datetime, destination): 894 """Collects the tradefed logs. 895 896 It is legal to collect the same logs multiple times. This is normal 897 after 'tradefed continue' updates existing logs with new results. 898 899 @param repository: Full path to tradefeds output on disk. 900 @param datetime: The identifier which tradefed assigned to the run. 901 Currently this looks like '2016.07.14_00.34.50'. 902 @param destination: Autotest result directory (destination of logs). 903 """ 904 logging.info('Collecting tradefed testResult.xml and logs to %s.', 905 destination) 906 repository_results = os.path.join(repository, 'results') 907 repository_logs = os.path.join(repository, 'logs') 908 # Because other tools rely on the currently chosen Google storage paths 909 # we need to keep destination_results in 910 # cheets_CTS.*/results/android-cts/2016.mm.dd_hh.mm.ss(/|.zip) 911 # and destination_logs in 912 # cheets_CTS.*/results/android-cts/logs/2016.mm.dd_hh.mm.ss/ 913 destination_results = destination 914 destination_results_datetime = os.path.join(destination_results, 915 datetime) 916 destination_results_datetime_zip = destination_results_datetime + '.zip' 917 destination_logs = os.path.join(destination, 'logs') 918 destination_logs_datetime = os.path.join(destination_logs, datetime) 919 # We may have collected the same logs before, clean old versions. 920 if os.path.exists(destination_results_datetime_zip): 921 os.remove(destination_results_datetime_zip) 922 if os.path.exists(destination_results_datetime): 923 shutil.rmtree(destination_results_datetime) 924 if os.path.exists(destination_logs_datetime): 925 shutil.rmtree(destination_logs_datetime) 926 shutil.copytree( 927 os.path.join(repository_results, datetime), 928 destination_results_datetime) 929 # Copying the zip file has to happen after the tree so the destination 930 # directory is available. 931 shutil.copy( 932 os.path.join(repository_results, datetime) + '.zip', 933 destination_results_datetime_zip) 934 shutil.copytree( 935 os.path.join(repository_logs, datetime), 936 destination_logs_datetime) 937 938 def _get_expected_failures(self, directory): 939 """Return a list of expected failures. 940 941 @return: a list of expected failures. 942 """ 943 logging.info('Loading expected failures from %s.', directory) 944 expected_fail_dir = os.path.join(self.bindir, directory) 945 expected_fail_files = glob.glob(expected_fail_dir + '/*.' + self._abi) 946 expected_failures = set() 947 for expected_fail_file in expected_fail_files: 948 try: 949 file_path = os.path.join(expected_fail_dir, expected_fail_file) 950 with open(file_path) as f: 951 lines = set(f.read().splitlines()) 952 logging.info('Loaded %d expected failures from %s', 953 len(lines), expected_fail_file) 954 expected_failures |= lines 955 except IOError as e: 956 logging.error('Error loading %s (%s).', file_path, e.strerror) 957 logging.info('Finished loading expected failures: %s', 958 expected_failures) 959 return expected_failures 960