Beispiel #1
0
    def __init__(self, port_obj, output_dir):
        super(WPTServe, self).__init__(port_obj, output_dir)
        # These ports must match wpt_support/wpt.config.json
        http_port, http_alt_port, https_port = (8001, 8081, 8444)
        ws_port, wss_port = (9001, 9444)
        self._name = 'wptserve'
        self._log_prefixes = ('access_log', 'error_log')
        self._mappings = [{
            'port': http_port
        }, {
            'port': http_alt_port
        }, {
            'port': https_port,
            'sslcert': True
        }, {
            'port': ws_port
        }, {
            'port': wss_port,
            'sslcert': True
        }]

        # TODO(burnik): We can probably avoid PID files for WPT in the future.
        fs = self._filesystem
        self._pid_file = fs.join(self._runtime_path, '%s.pid' % self._name)

        finder = PathFinder(fs)
        path_to_pywebsocket = finder.path_from_chromium_base(
            'third_party', 'pywebsocket', 'src')
        path_to_wpt_support = finder.path_from_blink_tools(
            'blinkpy', 'third_party', 'wpt')
        path_to_wpt_root = fs.join(path_to_wpt_support, 'wpt')
        path_to_wpt_tests = fs.abspath(
            fs.join(self._port_obj.layout_tests_dir(), 'external', 'wpt'))
        path_to_ws_handlers = fs.join(path_to_wpt_tests, 'websockets',
                                      'handlers')
        self._config_file = self._prepare_wptserve_config(path_to_wpt_support)
        wpt_script = fs.join(path_to_wpt_root, 'wpt')
        start_cmd = [
            self._port_obj.host.executable, '-u', wpt_script, 'serve',
            '--config', self._config_file, '--doc_root', path_to_wpt_tests
        ]

        # TODO(burnik): Merge with default start_cmd once we roll in websockets.
        if self._port_obj.host.filesystem.exists(path_to_ws_handlers):
            start_cmd += ['--ws_doc_root', path_to_ws_handlers]

        self._stdout = self._stderr = self._executive.DEVNULL
        # TODO(burnik): We should stop setting the CWD once WPT can be run without it.
        self._cwd = path_to_wpt_root
        self._env = port_obj.host.environ.copy()
        self._env.update({'PYTHONPATH': path_to_pywebsocket})
        self._start_cmd = start_cmd

        expiration_date = datetime.date(2025, 1, 4)
        if datetime.date.today() > expiration_date - datetime.timedelta(30):
            _log.error(
                'Pre-generated keys and certificates are going to be expired at %s.'
                ' Please re-generate them by following steps in %s/README.chromium.',
                expiration_date.strftime('%b %d %Y'), path_to_wpt_support)
Beispiel #2
0
    def generate_manifest(host, dest_path):
        """Generates MANIFEST.json on the specified directory."""
        finder = PathFinder(host.filesystem)
        wpt_exec_path = finder.path_from_blink_tools('blinkpy', 'third_party', 'wpt', 'wpt', 'wpt')
        cmd = ['python', wpt_exec_path, 'manifest', '--no-download', '--tests-root', dest_path]

        # ScriptError will be raised if the command fails.
        host.executive.run_command(
            cmd,
            return_stderr=True  # This will also include stderr in the exception message.
        )
Beispiel #3
0
 def run_pylint(self, path):
     finder = PathFinder(FileSystem())
     executive = Executive()
     env = os.environ.copy()
     env['PYTHONPATH'] = os.pathsep.join([
         get_blink_tools_dir(),
         finder.path_from_blink_source('build', 'scripts'),
         get_blinkpy_thirdparty_dir(),
         finder.path_from_blink_source('bindings', 'scripts'),
         finder.path_from_chromium_base('build', 'android'),
         finder.path_from_chromium_base('third_party', 'catapult', 'devil'),
         finder.path_from_chromium_base('third_party', 'pymock'),
     ])
     return executive.run_command([
         sys.executable,
         finder.path_from_depot_tools_base('pylint.py'),
         '--output-format=parseable',
         '--rcfile=' + finder.path_from_blink_tools('blinkpy', 'pylintrc'),
         path,
     ],
                                  env=env,
                                  error_handler=executive.ignore_error)
class WPTExpectationsUpdater(object):
    def __init__(self, host):
        self.host = host
        self.port = self.host.port_factory.get()
        self.git_cl = GitCL(host)
        self.finder = PathFinder(self.host.filesystem)
        self.ports_with_no_results = set()
        self.ports_with_all_pass = set()
        self.patchset = None

    def run(self, args=None):
        parser = argparse.ArgumentParser(description=__doc__)
        parser.add_argument(
            '--patchset',
            default=None,
            help='Patchset number to fetch new baselines from.')
        parser.add_argument('-v',
                            '--verbose',
                            action='store_true',
                            help='More verbose logging.')
        args = parser.parse_args(args)

        log_level = logging.DEBUG if args.verbose else logging.INFO
        configure_logging(logging_level=log_level, include_time=True)

        self.patchset = args.patchset
        self.update_expectations()

        return 0

    def update_expectations(self):
        """Downloads text new baselines and adds test expectations lines.

        Returns:
            A pair: A set of tests that are rebaselined, and a dictionary
            mapping tests that couldn't be rebaselined to lists of expectation
            lines written to TestExpectations.
        """
        issue_number = self.get_issue_number()
        if issue_number == 'None':
            raise ScriptError('No issue on current branch.')

        build_to_status = self.get_latest_try_jobs()
        _log.debug('Latest try jobs: %r', build_to_status)
        if not build_to_status:
            raise ScriptError('No try job information was collected.')

        # The manifest may be used below to do check which tests are reference tests.
        WPTManifest.ensure_manifest(self.host)

        # Here we build up a dict of failing test results for all platforms.
        test_expectations = {}
        for build, job_status in build_to_status.iteritems():
            if job_status.result == 'SUCCESS':
                self.ports_with_all_pass.add(self.port_name(build))
            port_results = self.get_failing_results_dict(build)
            test_expectations = self.merge_dicts(test_expectations,
                                                 port_results)

        # And then we merge results for different platforms that had the same results.
        for test_name, platform_result in test_expectations.iteritems():
            # platform_result is a dict mapping platforms to results.
            test_expectations[test_name] = self.merge_same_valued_keys(
                platform_result)

        # At this point, test_expectations looks like: {
        #     'test-with-failing-result': {
        #         ('port-name1', 'port-name2'): SimpleTestResult,
        #         'port-name3': SimpleTestResult
        #     }
        # }

        rebaselined_tests, test_expectations = self.download_text_baselines(
            test_expectations)
        test_expectation_lines = self.create_line_dict(test_expectations)
        self.write_to_test_expectations(test_expectation_lines)
        return rebaselined_tests, test_expectation_lines

    def get_issue_number(self):
        """Returns current CL number. Can be replaced in unit tests."""
        return self.git_cl.get_issue_number()

    def get_latest_try_jobs(self):
        """Returns the latest finished try jobs as Build objects."""
        return self.git_cl.latest_try_jobs(self._get_try_bots(),
                                           patchset=self.patchset)

    def get_failing_results_dict(self, build):
        """Returns a nested dict of failing test results.

        Retrieves a full list of layout test results from a builder result URL.
        Collects the builder name, platform and a list of tests that did not
        run as expected.

        Args:
            build: A Build object.

        Returns:
            A dictionary with the structure: {
                'test-with-failing-result': {
                    'full-port-name': SimpleTestResult
                }
            }
            If results could be fetched but none are failing,
            this will return an empty dictionary.
        """
        port_name = self.port_name(build)
        if port_name in self.ports_with_all_pass:
            # All tests passed, so there should be no failing results.
            return {}
        layout_test_results = self.host.buildbot.fetch_results(build)
        if layout_test_results is None:
            _log.warning('No results for build %s', build)
            self.ports_with_no_results.add(self.port_name(build))
            return {}
        failing_test_results = [
            result
            for result in layout_test_results.didnt_run_as_expected_results()
            if not result.did_pass()
        ]
        return self.generate_results_dict(self.port_name(build),
                                          failing_test_results)

    @memoized
    def port_name(self, build):
        return self.host.builders.port_name_for_builder_name(
            build.builder_name)

    def generate_results_dict(self, full_port_name, layout_test_results):
        """Makes a dict with results for one platform.

        Args:
            full_port_name: The fully-qualified port name, e.g. "win-win10".
            layout_test_results: A list of LayoutTestResult objects.

        Returns:
            A dictionary with the structure: {
                'test-name': {
                    'full-port-name': SimpleTestResult
                }
            }
        """
        test_dict = {}
        for result in layout_test_results:
            test_name = result.test_name()

            if not self.port.is_wpt_test(test_name):
                continue

            test_dict[test_name] = {
                full_port_name:
                SimpleTestResult(expected=result.expected_results(),
                                 actual=result.actual_results(),
                                 bug=UMBRELLA_BUG)
            }
        return test_dict

    def merge_dicts(self, target, source, path=None):
        """Recursively merges nested dictionaries.

        Args:
            target: First dictionary, which is updated based on source.
            source: Second dictionary, not modified.
            path: A list of keys, only used for making error messages.

        Returns:
            The updated target dictionary.
        """
        path = path or []
        for key in source:
            if key in target:
                if (isinstance(target[key], dict)) and isinstance(
                        source[key], dict):
                    self.merge_dicts(target[key], source[key],
                                     path + [str(key)])
                elif target[key] == source[key]:
                    pass
                else:
                    raise ValueError(
                        'The key: %s already exist in the target dictionary.' %
                        '.'.join(path))
            else:
                target[key] = source[key]
        return target

    def merge_same_valued_keys(self, dictionary):
        """Merges keys in dictionary with same value.

        Traverses through a dict and compares the values of keys to one another.
        If the values match, the keys are combined to a tuple and the previous
        keys are removed from the dict.

        Args:
            dictionary: A dictionary with a dictionary as the value.

        Returns:
            A new dictionary with updated keys to reflect matching values of keys.
            Example: {
                'one': {'foo': 'bar'},
                'two': {'foo': 'bar'},
                'three': {'foo': 'bar'}
            }
            is converted to a new dictionary with that contains
            {('one', 'two', 'three'): {'foo': 'bar'}}
        """
        merged_dict = {}
        matching_value_keys = set()
        keys = sorted(dictionary.keys())
        while keys:
            current_key = keys[0]
            found_match = False
            if current_key == keys[-1]:
                merged_dict[current_key] = dictionary[current_key]
                keys.remove(current_key)
                break

            for next_item in keys[1:]:
                if dictionary[current_key] == dictionary[next_item]:
                    found_match = True
                    matching_value_keys.update([current_key, next_item])

                if next_item == keys[-1]:
                    if found_match:
                        merged_dict[tuple(
                            matching_value_keys)] = dictionary[current_key]
                        keys = [
                            k for k in keys if k not in matching_value_keys
                        ]
                    else:
                        merged_dict[current_key] = dictionary[current_key]
                        keys.remove(current_key)
            matching_value_keys = set()
        return merged_dict

    def get_expectations(self, result, test_name=''):
        """Returns a set of test expectations based on the result of a test.

        Returns a set of one or more test expectations based on the expected
        and actual results of a given test name. This function is to decide
        expectations for tests that could not be rebaselined.

        Args:
            result: A SimpleTestResult.
            test_name: The test name string (optional).

        Returns:
            A set of one or more test expectation strings with the first letter
            capitalized. Example: {'Failure', 'Timeout'}.
        """
        actual_results = set(result.actual.split())
        # If the result is MISSING, this implies that the test was not
        # rebaselined and has an actual result but no baseline. We can't
        # add a Missing expectation (this is not allowed), but no other
        # expectation is correct.
        # We also want to skip any new manual tests that are not automated;
        # see crbug.com/708241 for context.
        if ('MISSING' in actual_results
                or '-manual.' in test_name and 'TIMEOUT' in actual_results):
            return {'Skip'}
        expectations = set()
        failure_types = {'TEXT', 'IMAGE+TEXT', 'IMAGE', 'AUDIO'}
        other_types = {'TIMEOUT', 'CRASH', 'PASS'}
        for actual in actual_results:
            if actual in failure_types:
                expectations.add('Failure')
            if actual in other_types:
                expectations.add(actual.capitalize())
        return expectations

    def create_line_dict(self, merged_results):
        """Creates list of test expectations lines.

        Traverses through the given |merged_results| dictionary and parses the
        value to create one test expectations line per key.

        Test expectation lines have the following format:
            ['BUG_URL [PLATFORM(S)] TEST_NAME [EXPECTATION(S)]']

        Args:
            merged_results: A dictionary with the format:
                {
                    'test-with-failing-result': {
                        ('port-name1', 'port-name2'): SimpleTestResult,
                        'port-name3': SimpleTestResult
                    }
                }

        Returns:
            A dictionary from test names to a list of test expectation lines
            (each SimpleTestResult turns into a line).
        """
        line_dict = defaultdict(list)
        for test_name, port_results in sorted(merged_results.iteritems()):
            if not self.port.is_wpt_test(test_name):
                _log.warning(
                    'Non-WPT test "%s" unexpectedly passed to create_line_dict.',
                    test_name)
                continue
            for port_names, result in sorted(port_results.iteritems()):
                line_dict[test_name].append(
                    self._create_line(test_name, port_names, result))
        return line_dict

    def _create_line(self, test_name, port_names, result):
        """Constructs a test expectation line string.

        Args:
            test_name: The test name string.
            port_names: A list of full port names that the line should apply to.
            result: A SimpleTestResult.

        Returns:
            A string that contains a line of test expectation.
        """
        port_names = self.tuple_or_value_to_list(port_names)

        # The set of ports with no results is assumed to have have no
        # overlap with the set of port names passed in here.
        assert (set(port_names) & self.ports_with_no_results) == set()

        # The ports with no results are generally ports of builders that
        # failed, maybe for unrelated reasons. At this point, we add ports
        # with no results to the list of platforms because we're guessing
        # that this new expectation might be cross-platform and should
        # also apply to any ports that we weren't able to get results for.
        port_names.extend(self.ports_with_no_results)

        specifier_part = self.specifier_part(test_name, port_names)

        line_parts = [result.bug]
        if specifier_part:
            line_parts.append(specifier_part)
        line_parts.append(test_name)
        line_parts.append('[ %s ]' %
                          ' '.join(self.get_expectations(result, test_name)))

        return ' '.join(line_parts)

    def specifier_part(self, test_name, port_names):
        """Returns the specifier part for a new test expectations line.

        Args:
            test_name: The test name string.
            port_names: A list of full port names that the line should apply to.

        Returns:
            The specifier part of the new expectation line, e.g. "[ Mac ]".
            This will be an empty string if the line should apply to all platforms.
        """
        specifiers = []
        for name in sorted(port_names):
            specifiers.append(
                self.host.builders.version_specifier_for_port_name(name))

        if self.specifiers_can_extend_to_all_platforms(specifiers, test_name):
            return ''

        specifiers = self.simplify_specifiers(
            specifiers, self.port.configuration_specifier_macros())
        if not specifiers:
            return ''
        return '[ %s ]' % ' '.join(specifiers)

    @staticmethod
    def tuple_or_value_to_list(tuple_or_value):
        """Converts a tuple to a list, and a string value to a one-item list."""
        if isinstance(tuple_or_value, tuple):
            return list(tuple_or_value)
        return [tuple_or_value]

    def specifiers_can_extend_to_all_platforms(self, specifiers, test_name):
        """Tests whether a list of specifiers can be extended to all platforms.

        Tries to add skipped platform specifiers to the list and tests if the
        extended list covers all platforms.
        """
        extended_specifiers = specifiers + self.skipped_specifiers(test_name)
        # If the list is simplified to empty, then all platforms are covered.
        return not self.simplify_specifiers(
            extended_specifiers, self.port.configuration_specifier_macros())

    def skipped_specifiers(self, test_name):
        """Returns a list of platform specifiers for which the test is skipped."""
        specifiers = []
        for port in self.all_try_builder_ports():
            if port.skips_test(test_name):
                specifiers.append(
                    self.host.builders.version_specifier_for_port_name(
                        port.name()))
        return specifiers

    @memoized
    def all_try_builder_ports(self):
        """Returns a list of Port objects for all try builders."""
        return [
            self.host.port_factory.get_from_builder_name(name)
            for name in self._get_try_bots()
        ]

    def simplify_specifiers(self, specifiers, specifier_macros):
        """Simplifies the specifier part of an expectation line if possible.

        "Simplifying" means finding the shortest list of platform specifiers
        that is equivalent to the given list of specifiers. This can be done
        because there are "macro specifiers" that stand in for multiple version
        specifiers, and an empty list stands in for "all platforms".

        Args:
            specifiers: A collection of specifiers (case insensitive).
            specifier_macros: A dict mapping "macros" for groups of specifiers
                to lists of version specifiers. e.g. {"win": ["win7", "win10"]}.
                If there are versions in this dict for that have no corresponding
                try bots, they are ignored.

        Returns:
            A shortened list of specifiers (capitalized). For example, ["win7",
            "win10"] would be converted to ["Win"]. If the given list covers
            all supported platforms, then an empty list is returned.
        """
        specifiers = {s.lower() for s in specifiers}
        covered_by_try_bots = self._platform_specifiers_covered_by_try_bots()
        for macro, versions in specifier_macros.iteritems():
            macro = macro.lower()

            # Only consider version specifiers that have corresponding try bots.
            versions = {
                s.lower()
                for s in versions if s.lower() in covered_by_try_bots
            }
            if versions <= specifiers:
                specifiers -= versions
                specifiers.add(macro)
        if specifiers == {macro.lower() for macro in specifier_macros}:
            return []
        return sorted(specifier.capitalize() for specifier in specifiers)

    def _platform_specifiers_covered_by_try_bots(self):
        all_platform_specifiers = set()
        for builder_name in self._get_try_bots():
            all_platform_specifiers.add(
                self.host.builders.platform_specifier_for_builder(
                    builder_name).lower())
        return frozenset(all_platform_specifiers)

    def write_to_test_expectations(self, line_dict):
        """Writes the given lines to the TestExpectations file.

        The place in the file where the new lines are inserted is after a marker
        comment line. If this marker comment line is not found, then everything
        including the marker line is appended to the end of the file.

        Args:
            line_dict: A dictionary from test names to a list of test expectation lines.
        """
        if not line_dict:
            _log.info('No lines to write to TestExpectations.')
            return

        _log.info('Lines to write to TestExpectations:')
        line_list = []
        for lines in line_dict.itervalues():
            for line in lines:
                line_list.append(line)
                _log.info('  %s', line)

        expectations_file_path = self.port.path_to_generic_test_expectations_file(
        )
        file_contents = self.host.filesystem.read_text_file(
            expectations_file_path)

        marker_comment_index = file_contents.find(MARKER_COMMENT)
        if marker_comment_index == -1:
            file_contents += '\n%s\n' % MARKER_COMMENT
            file_contents += '\n'.join(line_list)
        else:
            end_of_marker_line = (file_contents[marker_comment_index:].find(
                '\n')) + marker_comment_index
            file_contents = file_contents[:end_of_marker_line + 1] + '\n'.join(
                line_list) + file_contents[end_of_marker_line:]

        self.host.filesystem.write_text_file(expectations_file_path,
                                             file_contents)

    # TODO(robertma): Unit test this method.
    def download_text_baselines(self, test_results):
        """Fetches new baseline files for tests that should be rebaselined.

        Invokes `blink_tool.py rebaseline-cl` in order to download new baselines
        (-expected.txt files) for testharness.js tests that did not crash or
        time out. Then, the platform-specific test is removed from the overall
        failure test dictionary and the resulting dictionary is returned.

        Args:
            test_results: A dictionary of failing test results, mapping test
                names to lists of platforms to SimpleTestResult.

        Returns:
            A pair: A set of tests that are rebaselined, and a modified copy of
            the test_results dictionary containing only tests that couldn't be
            rebaselined.
        """
        tests_to_rebaseline, test_results = self.get_tests_to_rebaseline(
            test_results)
        if not tests_to_rebaseline:
            _log.info('No tests to rebaseline.')
            return tests_to_rebaseline, test_results
        _log.info('Tests to rebaseline:')
        for test in tests_to_rebaseline:
            _log.info('  %s', test)

        blink_tool = self.finder.path_from_blink_tools('blink_tool.py')
        command = [
            'python',
            blink_tool,
            'rebaseline-cl',
            '--verbose',
            '--no-trigger-jobs',
            '--fill-missing',
        ]
        if self.patchset:
            command.append('--patchset=' + str(self.patchset))
        command += tests_to_rebaseline
        self.host.executive.run_command(command)
        return tests_to_rebaseline, test_results

    def get_tests_to_rebaseline(self, test_results):
        """Filters failing tests that can be rebaselined.

        Creates a list of tests to rebaseline depending on the tests' platform-
        specific results. In general, this will be non-ref tests that failed
        due to a baseline mismatch (rather than crash or timeout).

        Args:
            test_results: A dictionary of failing test results, mapping test
                names to lists of platforms to SimpleTestResult.

        Returns:
            A pair: A set of tests to be rebaselined, and a modified copy of
            the test_results dictionary. The tests to be rebaselined should
            include testharness.js tests that failed due to a baseline mismatch.
        """
        new_test_results = copy.deepcopy(test_results)
        tests_to_rebaseline = set()
        for test_name in test_results:
            for platforms, result in test_results[test_name].iteritems():
                if self.can_rebaseline(test_name, result):
                    del new_test_results[test_name][platforms]
                    tests_to_rebaseline.add(test_name)
        return sorted(tests_to_rebaseline), new_test_results

    def can_rebaseline(self, test_name, result):
        """Checks if a test can be rebaselined.

        Args:
            test_name: The test name string.
            result: A SimpleTestResult.
        """
        if self.is_reference_test(test_name):
            return False
        if any(x in result.actual for x in ('CRASH', 'TIMEOUT', 'MISSING')):
            return False
        return True

    def is_reference_test(self, test_name):
        """Checks whether a given test is a reference test."""
        return bool(self.port.reference_files(test_name))

    def _get_try_bots(self):
        return self.host.builders.all_try_builder_names()
Beispiel #5
0
    def __init__(self, port_obj, output_dir):
        super(WPTServe, self).__init__(port_obj, output_dir)
        # These ports must match wpt_support/wpt.config.json
        http_port, http_alt_port, https_port, https_alt_port = (8001, 8081,
                                                                8444, 8445)
        ws_port, wss_port = (9001, 9444)
        self._name = 'wptserve'
        self._log_prefixes = ('wptserve_stderr', )
        self._mappings = [{
            'port': http_port,
            'scheme': 'http'
        }, {
            'port': http_alt_port,
            'scheme': 'http'
        }, {
            'port': https_port,
            'scheme': 'https',
            'sslcert': True
        }, {
            'port': https_alt_port,
            'scheme': 'https',
            'sslcert': True
        }, {
            'port': ws_port,
            'scheme': 'ws'
        }, {
            'port': wss_port,
            'scheme': 'wss',
            'sslcert': True
        }]

        # TODO(burnik): We can probably avoid PID files for WPT in the future.
        fs = self._filesystem
        self._pid_file = fs.join(self._runtime_path, '%s.pid' % self._name)
        self._config_file = fs.join(self._runtime_path, 'wpt.config.json')

        finder = PathFinder(fs)
        path_to_pywebsocket = finder.path_from_chromium_base(
            'third_party', 'pywebsocket3', 'src')
        self.path_to_wpt_support = finder.path_from_blink_tools(
            'blinkpy', 'third_party', 'wpt')
        path_to_wpt_root = fs.join(self.path_to_wpt_support, 'wpt')
        path_to_wpt_tests = fs.abspath(
            fs.join(self._port_obj.web_tests_dir(), 'external', 'wpt'))
        path_to_ws_handlers = fs.join(path_to_wpt_tests, 'websockets',
                                      'handlers')
        wpt_script = fs.join(path_to_wpt_root, 'wpt')
        start_cmd = [
            self._port_obj.host.executable,
            '-u',
            wpt_script,
            'serve',
            '--config',
            self._config_file,
            '--doc_root',
            path_to_wpt_tests,
            '--no-h2',
        ]

        # Some users (e.g. run_webdriver_tests.py) do not need WebSocket
        # handlers, so we only add the flag if the directory exists.
        if self._port_obj.host.filesystem.exists(path_to_ws_handlers):
            start_cmd += ['--ws_doc_root', path_to_ws_handlers]

        # TODO(burnik): We should stop setting the CWD once WPT can be run without it.
        self._cwd = path_to_wpt_root
        self._env = port_obj.host.environ.copy()
        self._env.update({'PYTHONPATH': path_to_pywebsocket})
        self._start_cmd = start_cmd

        self._error_log_path = self._filesystem.join(output_dir,
                                                     'wptserve_stderr.txt')

        expiration_date = datetime.date(2025, 1, 4)
        if datetime.date.today() > expiration_date - datetime.timedelta(30):
            _log.error(
                'Pre-generated keys and certificates are going to be expired at %s.'
                ' Please re-generate them by following steps in %s/README.chromium.',
                expiration_date.strftime('%b %d %Y'), self.path_to_wpt_support)
Beispiel #6
0
class WPTExpectationsUpdater(object):
    MARKER_COMMENT = '# ====== New tests from wpt-importer added here ======'
    UMBRELLA_BUG = 'crbug.com/626703'

    def __init__(self, host, args=None):
        self.host = host
        self.port = self.host.port_factory.get()
        self.finder = PathFinder(self.host.filesystem)
        self.git_cl = GitCL(host)
        self.git = self.host.git(self.finder.chromium_base())
        self.configs_with_no_results = []
        self.configs_with_all_pass = []
        self.patchset = None

        # Get options from command line arguments.
        parser = argparse.ArgumentParser(description=__doc__)
        self.add_arguments(parser)
        self.options = parser.parse_args(args or [])

        # Set up TestExpectations instance which contains all
        # expectations files associated with the platform.
        expectations_dict = {p: self.host.filesystem.read_text_file(p)
                             for p in self.expectations_files()}
        self._test_expectations = TestExpectations(
            self.port, expectations_dict=expectations_dict)

    def expectations_files(self):
        """Returns list of expectations files.

        Each expectation file in the list will be cleaned of expectations
        for tests that were removed and will also have test names renamed
        for tests that were renamed. Also the files may have their expectations
        updated using builder results.
        """
        return (self.port.all_expectations_dict().keys() +
                PRODUCTS_TO_EXPECTATION_FILE_PATHS.values())

    def run(self):
        """Does required setup before calling update_expectations().

        Do not override this function!
        """
        log_level = logging.DEBUG if self.options.verbose else logging.INFO
        configure_logging(logging_level=log_level, include_time=True)

        self.patchset = self.options.patchset

        # Remove expectations for deleted tests and rename tests in expectations
        # for renamed tests.
        self.cleanup_test_expectations_files()

        if not self.options.cleanup_test_expectations_only:
            # Use try job results to update expectations and baselines
            self.update_expectations()

        return 0

    def add_arguments(self, parser):
        parser.add_argument(
            '--patchset',
            default=None,
            help='Patchset number to fetch new baselines from.')
        parser.add_argument(
            '-v',
            '--verbose',
            action='store_true',
            help='More verbose logging.')
        parser.add_argument(
            '--clean-up-affected-tests-only',
            action='store_true',
            help='Only cleanup expectations deleted or renamed in current CL. '
                 'If flag is not used then a full cleanup of deleted or '
                 'renamed tests will be done in expectations.')
        parser.add_argument(
            '--cleanup-test-expectations-only',
            action='store_true',
            help='Cleanup test expectations files and then exit script.')
        # TODO(rmhasan): Move this argument to the
        # AndroidWPTExpectationsUpdater add_arguments implementation.
        # Also look into using sub parsers to separate android and
        # desktop specific arguments.
        parser.add_argument(
            '--android-product', action='append', default=[],
            help='Android products whose baselines will be updated.',
            choices=PRODUCTS)

    def update_expectations(self):
        """Downloads text new baselines and adds test expectations lines.

        Returns:
            A pair: A set of tests that are rebaselined, and a dictionary
            mapping tests that couldn't be rebaselined to lists of expectation
            lines written to TestExpectations.
        """
        issue_number = self.get_issue_number()
        if issue_number == 'None':
            raise ScriptError('No issue on current branch.')

        build_to_status = self.get_latest_try_jobs()
        _log.debug('Latest try jobs: %r', build_to_status)
        if not build_to_status:
            raise ScriptError('No try job information was collected.')

        # Here we build up a dict of failing test results for all platforms.
        test_expectations = {}
        for build, job_status in build_to_status.iteritems():
            if job_status.result == 'SUCCESS':
                self.configs_with_all_pass.extend(
                    self.get_builder_configs(build))
                continue
            result_dicts = self.get_failing_results_dicts(build)
            for result_dict in result_dicts:
                test_expectations = self.merge_dicts(
                    test_expectations, result_dict)

        # At this point, test_expectations looks like: {
        #     'test-with-failing-result': {
        #         config1: SimpleTestResult,
        #         config2: SimpleTestResult,
        #         config3: AnotherSimpleTestResult
        #     }
        # }
        # And then we merge results for different platforms that had the same results.
        for test_name, platform_result in test_expectations.iteritems():
            # platform_result is a dict mapping platforms to results.
            test_expectations[test_name] = self.merge_same_valued_keys(
                platform_result)

        # At this point, test_expectations looks like: {
        #     'test-with-failing-result': {
        #         (config1, config2): SimpleTestResult,
        #         (config3,): AnotherSimpleTestResult
        #     }
        # }

        rebaselined_tests, test_expectations = self.download_text_baselines(
            test_expectations)
        exp_lines_dict = self.write_to_test_expectations(test_expectations)
        return rebaselined_tests, exp_lines_dict

    def get_issue_number(self):
        """Returns current CL number. Can be replaced in unit tests."""
        return self.git_cl.get_issue_number()

    def get_latest_try_jobs(self):
        """Returns the latest finished try jobs as Build objects."""
        return self.git_cl.latest_try_jobs(
            builder_names=self._get_try_bots(), patchset=self.patchset)

    def get_failing_results_dicts(self, build):
        """Returns a list of nested dicts of failing test results.

        Retrieves a full list of web test results from a builder result URL.
        Collects the builder name, platform and a list of tests that did not
        run as expected.

        Args:
            build: A Build object.

        Returns:
            A list of dictionaries that have the following structure.

            {
                'test-with-failing-result': {
                    config: SimpleTestResult
                }
            }

            If results could be fetched but none are failing,
            this will return an empty list.
        """

        test_results_list = self._get_web_test_results(build)

        has_webdriver_tests = self.host.builders.has_webdriver_tests_for_builder(
            build.builder_name)

        if has_webdriver_tests:
            master = self.host.builders.master_for_builder(build.builder_name)
            test_results_list.append(
                self.host.results_fetcher.fetch_webdriver_test_results(
                    build, master))

        test_results_list = filter(None, test_results_list)
        if not test_results_list:
            _log.warning('No results for build %s', build)
            self.configs_with_no_results.extend(self.get_builder_configs(build))
            return []

        failing_test_results = []
        for results_set in test_results_list:
            results_dict = self.generate_failing_results_dict(
                build, results_set)
            if results_dict:
                failing_test_results.append(results_dict)
        return failing_test_results

    def _get_web_test_results(self, build):
        """Gets web tests results for a builder.

        Args:
            build: Named tuple containing builder name and number

        Returns:
            List of web tests results for each web test step
            in build.
        """
        return [self.host.results_fetcher.fetch_results(build)]

    def get_builder_configs(self, build, *_):
        return [DesktopConfig(port_name=self.port_name(build))]

    @memoized
    def port_name(self, build):
        return self.host.builders.port_name_for_builder_name(
            build.builder_name)

    def generate_failing_results_dict(self, build, web_test_results):
        """Makes a dict with results for one platform.

        Args:
            builder: Builder instance containing builder information..
            web_test_results: A list of WebTestResult objects.

        Returns:
            A dictionary with the structure: {
                'test-name': {
                    ('full-port-name',): SimpleTestResult
                }
            }
        """
        test_dict = {}
        configs = self.get_builder_configs(build, web_test_results)
        _log.debug('Getting failing results dictionary'
                   ' for %s step in latest %s build' % (
                       web_test_results.step_name(), build.builder_name))

        if len(configs) > 1:
            raise ScriptError('More than one configs were produced for'
                              ' builder and web tests step combination')
        if not configs:
            raise ScriptError('No configuration was found for builder and web test'
                              ' step combination ')
        config = configs[0]
        if config in self.configs_with_all_pass:
            return {}

        for result in web_test_results.didnt_run_as_expected_results():
            # TODO(rmhasan) If a test fails unexpectedly then it runs multiple
            # times until, it passes or a retry limit is reached. Even though
            # it passed we there are still flaky failures that we are not
            # creating test expectations for. Maybe we should add a mode
            # which creates expectations for tests that are flaky but still
            # pass in a web test step.
            if result.did_pass():
                continue

            test_name = result.test_name()
            if not self._is_wpt_test(test_name):
                continue
            test_dict[test_name] = {
                config:
                SimpleTestResult(
                    expected=result.expected_results(),
                    actual=result.actual_results(),
                    bug=self.UMBRELLA_BUG)
            }
        return test_dict

    def _is_wpt_test(self, test_name):
        """Check if a web test is a WPT tests.

        In blink web tests results, each test name is relative to
        the web_tests directory instead of the wpt directory. We
        need to use the port.is_wpt_test() function to find out if a test
        is from the WPT suite.

        Returns: True if a test is in the external/wpt subdirectory of
            the web_tests directory."""
        return self.port.is_wpt_test(test_name)

    def merge_dicts(self, target, source, path=None):
        """Recursively merges nested dictionaries.

        Args:
            target: First dictionary, which is updated based on source.
            source: Second dictionary, not modified.
            path: A list of keys, only used for making error messages.

        Returns:
            The updated target dictionary.
        """
        path = path or []
        for key in source:
            if key in target:
                if (isinstance(target[key], dict)) and isinstance(
                        source[key], dict):
                    self.merge_dicts(target[key], source[key],
                                     path + [str(key)])
                elif target[key] == source[key]:
                    pass
                else:
                    raise ValueError(
                        'The key: %s already exist in the target dictionary.' %
                        '.'.join(path))
            else:
                target[key] = source[key]
        return target

    def merge_same_valued_keys(self, dictionary):
        """Merges keys in dictionary with same value.

        Traverses through a dict and compares the values of keys to one another.
        If the values match, the keys are combined to a tuple and the previous
        keys are removed from the dict.

        Args:
            dictionary: A dictionary with a dictionary as the value.

        Returns:
            A new dictionary with updated keys to reflect matching values of keys.
            Example: {
                'one': {'foo': 'bar'},
                'two': {'foo': 'bar'},
                'three': {'foo': 'bar'}
            }
            is converted to a new dictionary with that contains
            {('one', 'two', 'three'): {'foo': 'bar'}}
        """
        merged_dict = {}
        matching_value_keys = set()
        keys = sorted(dictionary.keys())
        while keys:
            current_key = keys[0]
            found_match = False
            if current_key == keys[-1]:
                merged_dict[tuple([current_key])] = dictionary[current_key]
                keys.remove(current_key)
                break

            for next_item in keys[1:]:
                if dictionary[current_key] == dictionary[next_item]:
                    found_match = True
                    matching_value_keys.update([current_key, next_item])

                if next_item == keys[-1]:
                    if found_match:
                        merged_dict[
                            tuple(matching_value_keys)] = dictionary[current_key]
                        keys = [
                            k for k in keys if k not in matching_value_keys
                        ]
                    else:
                        merged_dict[tuple([current_key])] = dictionary[current_key]
                        keys.remove(current_key)
            matching_value_keys = set()
        return merged_dict

    def get_expectations(self, result, test_name=''):
        """Returns a set of test expectations based on the result of a test.

        Returns a set of one or more test expectations based on the expected
        and actual results of a given test name. This function is to decide
        expectations for tests that could not be rebaselined.

        Args:
            result: A SimpleTestResult.
            test_name: The test name string (optional).

        Returns:
            A set of one or more test expectation strings with the first letter
            capitalized. Example: {'Failure', 'Timeout'}.
        """
        actual_results = set(result.actual.split())

        # If the result is MISSING, this implies that the test was not
        # rebaselined and has an actual result but no baseline. We can't
        # add a Missing expectation (this is not allowed), but no other
        # expectation is correct.
        # We also want to skip any new manual tests that are not automated;
        # see crbug.com/708241 for context.
        if 'MISSING' in actual_results:
            return {'Skip'}
        if '-manual.' in test_name and 'TIMEOUT' in actual_results:
            return {'Skip'}
        expectations = set()
        failure_types = {'TEXT', 'IMAGE+TEXT', 'IMAGE', 'AUDIO', 'FAIL'}
        other_types = {'TIMEOUT', 'CRASH', 'PASS'}
        for actual in actual_results:
            if actual in failure_types:
                expectations.add('Failure')
            if actual in other_types:
                expectations.add(actual.capitalize())
        return expectations

    def create_line_dict(self, merged_results):
        """Creates list of test expectations lines.

        Traverses through the given |merged_results| dictionary and parses the
        value to create one test expectations line per key.

        Test expectation lines have the following format:
            ['BUG_URL [PLATFORM(S)] TEST_NAME [EXPECTATION(S)]']

        Args:
            merged_results: A dictionary with the format:
                {
                    'test-with-failing-result': {
                        (config1, config2): SimpleTestResult,
                        (config3,): SimpleTestResult
                    }
                }

        Returns:
            A dictionary from test names to a list of test expectation lines
            (each SimpleTestResult turns into a line).
        """
        line_dict = defaultdict(list)
        for test_name, test_results in sorted(merged_results.iteritems()):
            if not self._is_wpt_test(test_name):
                _log.warning(
                    'Non-WPT test "%s" unexpectedly passed to create_line_dict.',
                    test_name)
                continue
            for configs, result in sorted(test_results.iteritems()):
                line_dict[test_name].extend(
                    self._create_lines(test_name, configs, result))
        return line_dict

    def _create_lines(self, test_name, configs, result):
        """Constructs test expectation line strings.

        Args:
            test_name: The test name string.
            configs: A list of full configs that the line should apply to.
            result: A SimpleTestResult.

        Returns:
            A list of strings which each is a line of test expectation for given
            |test_name|.
        """
        lines = []
        # The set of ports with no results is assumed to have have no
        # overlap with the set of port names passed in here.
        assert set(configs) & set(self.configs_with_no_results) == set()

        # The ports with no results are generally ports of builders that
        # failed, maybe for unrelated reasons. At this point, we add ports
        # with no results to the list of platforms because we're guessing
        # that this new expectation might be cross-platform and should
        # also apply to any ports that we weren't able to get results for.
        configs = tuple(list(configs) + self.configs_with_no_results)

        expectations = '[ %s ]' % \
            ' '.join(self.get_expectations(result, test_name))
        for specifier in self.normalized_specifiers(test_name, configs):
            line_parts = []
            if specifier:
                line_parts.append('[ %s ]' % specifier)
            # Escape literal asterisks for typ (https://crbug.com/1036130).
            line_parts.append(test_name.replace('*', '\\*'))
            line_parts.append(expectations)

            # Only add the bug link if the expectations do not include WontFix.
            if 'WontFix' not in expectations and result.bug:
                line_parts.insert(0, result.bug)

            lines.append(' '.join(line_parts))
        return lines

    def normalized_specifiers(self, test_name, configs):
        """Converts and simplifies ports into platform specifiers.

        Args:
            test_name: The test name string.
            configs: A list of full configs that the line should apply to.

        Returns:
            A list of specifier string, e.g. ["Mac", "Win"].
            [''] will be returned if the line should apply to all platforms.
        """
        specifiers = []
        for config in configs:
            specifiers.append(
                self.host.builders.version_specifier_for_port_name(
                    config.port_name))

        if self.specifiers_can_extend_to_all_platforms(specifiers, test_name):
            return ['']

        specifiers = self.simplify_specifiers(
            specifiers, self.port.configuration_specifier_macros())
        if not specifiers:
            return ['']
        return specifiers

    def specifiers_can_extend_to_all_platforms(self, specifiers, test_name):
        """Tests whether a list of specifiers can be extended to all platforms.

        Tries to add skipped platform specifiers to the list and tests if the
        extended list covers all platforms.
        """
        extended_specifiers = specifiers + self.skipped_specifiers(test_name)
        # If the list is simplified to empty, then all platforms are covered.
        return not self.simplify_specifiers(
            extended_specifiers, self.port.configuration_specifier_macros())

    def skipped_specifiers(self, test_name):
        """Returns a list of platform specifiers for which the test is skipped."""
        specifiers = []
        for port in self.all_try_builder_ports():
            if port.skips_test(test_name):
                specifiers.append(
                    self.host.builders.version_specifier_for_port_name(
                        port.name()))
        return specifiers

    @memoized
    def all_try_builder_ports(self):
        """Returns a list of Port objects for all try builders."""
        return [
            self.host.port_factory.get_from_builder_name(name)
            for name in self._get_try_bots()
        ]

    def simplify_specifiers(self, specifiers, specifier_macros):
        """Simplifies the specifier part of an expectation line if possible.

        "Simplifying" means finding the shortest list of platform specifiers
        that is equivalent to the given list of specifiers. This can be done
        because there are "macro specifiers" that stand in for multiple version
        specifiers, and an empty list stands in for "all platforms".

        Args:
            specifiers: A collection of specifiers (case insensitive).
            specifier_macros: A dict mapping "macros" for groups of specifiers
                to lists of version specifiers. e.g. {"win": ["win7", "win10"]}.
                If there are versions in this dict for that have no corresponding
                try bots, they are ignored.

        Returns:
            A shortened list of specifiers (capitalized). For example, ["win7",
            "win10"] would be converted to ["Win"]. If the given list covers
            all supported platforms, then an empty list is returned.
        """
        specifiers = {s.lower() for s in specifiers}
        covered_by_try_bots = self._platform_specifiers_covered_by_try_bots()
        for macro, versions in specifier_macros.iteritems():
            macro = macro.lower()

            # Only consider version specifiers that have corresponding try bots.
            versions = {
                s.lower()
                for s in versions if s.lower() in covered_by_try_bots
            }
            if len(versions) == 0:
                continue
            if versions <= specifiers:
                specifiers -= versions
                specifiers.add(macro)
        if specifiers == {macro.lower() for macro in specifier_macros}:
            return []
        return sorted(specifier.capitalize() for specifier in specifiers)

    def _platform_specifiers_covered_by_try_bots(self):
        all_platform_specifiers = set()
        for builder_name in self._get_try_bots():
            all_platform_specifiers.add(
                self.host.builders.platform_specifier_for_builder(
                    builder_name).lower())
        return frozenset(all_platform_specifiers)

    def write_to_test_expectations(self, test_expectations):
        """Writes the given lines to the TestExpectations file.

        The place in the file where the new lines are inserted is after a marker
        comment line. If this marker comment line is not found, then everything
        including the marker line is appended to the end of the file.

        All WontFix tests are inserted to NeverFixTests file instead of TextExpectations
        file.

        Args:
            test_expectations: A dictionary mapping test names to a dictionary
            mapping platforms and test results.
        Returns:
            Dictionary mapping test names to lists of test expectation strings.
        """
        line_dict = self.create_line_dict(test_expectations)
        if not line_dict:
            _log.info(
                'No lines to write to TestExpectations,'
                ' WebdriverExpectations or NeverFixTests.'
            )
            return {}

        line_list = []
        wont_fix_list = []
        webdriver_list = []
        for lines in line_dict.itervalues():
            for line in lines:
                if 'Skip' in line and '-manual.' in line:
                    wont_fix_list.append(line)
                elif self.finder.webdriver_prefix() in line:
                    webdriver_list.append(line)
                else:
                    line_list.append(line)

        list_to_expectation = {
            self.port.path_to_generic_test_expectations_file(): line_list,
            self.port.path_to_webdriver_expectations_file(): webdriver_list
        }
        for expectations_file_path, lines in list_to_expectation.iteritems():
            if not lines:
                continue

            _log.info('Lines to write to %s:\n %s', expectations_file_path,
                      '\n'.join(lines))
            # Writes to TestExpectations file.
            file_contents = self.host.filesystem.read_text_file(
                expectations_file_path)

            marker_comment_index = file_contents.find(self.MARKER_COMMENT)
            if marker_comment_index == -1:
                file_contents += '\n%s\n' % self.MARKER_COMMENT
                file_contents += '\n'.join(lines)
            else:
                end_of_marker_line = (file_contents[marker_comment_index:].
                                      find('\n')) + marker_comment_index
                file_contents = (
                    file_contents[:end_of_marker_line + 1] + '\n'.join(lines) +
                    file_contents[end_of_marker_line:])

            self.host.filesystem.write_text_file(expectations_file_path,
                                                 file_contents)

        if wont_fix_list:
            _log.info('Lines to write to NeverFixTests:\n %s',
                      '\n'.join(wont_fix_list))
            # Writes to NeverFixTests file.
            wont_fix_path = self.port.path_to_never_fix_tests_file()
            wont_fix_file_content = self.host.filesystem.read_text_file(
                wont_fix_path)
            if not wont_fix_file_content.endswith('\n'):
                wont_fix_file_content += '\n'
            wont_fix_file_content += '\n'.join(wont_fix_list)
            wont_fix_file_content += '\n'
            self.host.filesystem.write_text_file(wont_fix_path,
                                                 wont_fix_file_content)
        return line_dict

    def cleanup_test_expectations_files(self):
        """Removes deleted tests from expectations files.

        Removes expectations for deleted tests or renames test names in
        expectation files for tests that were renamed. If the
        --clean-up-affected-tests-only command line argument is used then
        only tests deleted in the CL will have their expectations removed
        through this script. If that command line argument is not used then
        expectations for test files that no longer exist will be deleted.
        """
        deleted_test_files = self._list_deleted_test_files()
        renamed_test_files = self._list_renamed_test_files()

        for path in self._test_expectations.expectations_dict:
            _log.info(
                'Updating %s for any removed or renamed tests.' %
                self.host.filesystem.basename(path))
            self._clean_single_test_expectations_file(
                path, deleted_test_files, renamed_test_files)
        self._test_expectations.commit_changes()

    def _clean_single_test_expectations_file(
            self, path, deleted_files, renamed_files):
        """Cleans up a single test expectations file.

        Args:
            path: Path of expectations file that is being cleaned up.
            deleted_files: List of test file paths relative to the web tests
                directory which were deleted.
            renamed_files: Dictionary mapping test file paths to their new file
                name after renaming.
        """
        for line in self._test_expectations.get_updated_lines(path):
            # if a test is a glob type expectation or empty line or comment then
            # add it to the updated expectations file without modifications
            if not line.test or line.is_glob:
                continue
            root_test_file = self._get_root_file(line.test)

            if root_test_file in renamed_files:
                self._test_expectations.remove_expectations(path, [line])
                new_file_name = renamed_files[root_test_file]
                if self.finder.is_webdriver_test_path(root_test_file):
                    _, subtest_suffix = self.port.split_webdriver_test_name(
                        line.test)
                    line.test = self.port.add_webdriver_subtest_suffix(
                        new_file_name, subtest_suffix)
                elif '?' in line.test:
                    line.test = (
                        new_file_name + line.test[line.test.find('?'):])
                else:
                    line.test = new_file_name
                self._test_expectations.add_expectations(
                    path, [line], lineno=line.lineno)
            elif root_test_file in deleted_files:
                self._test_expectations.remove_expectations(
                    path, [line])

    @memoized
    def _get_root_file(self, test_name):
        """Strips arguments from a web test name in order to get the file name.

        It also removes the arguments for web driver tests. For instances for
        the test test1/example.html?Hello this function will return
        test1/example.html. For a webdriver test it would include arguments and
        would have the following format, {test file}>>{argument}.

        Args:
            test_name: Test name which may include test arguments.

        Returns:
            Returns the test file which is the root of a test.
        """
        if self.finder.is_webdriver_test_path(test_name):
            root_test_file, _ = (
                self.port.split_webdriver_test_name(test_name))
        elif '?' in test_name:
            root_test_file = test_name[:test_name.find('?')]
        else:
            root_test_file = test_name
        return root_test_file

    def _list_deleted_test_files(self):
        """Returns a list of web tests that have been deleted.

        If --clean-up-affected-tests-only is true then only test files deleted
        in the current CL may be removed from expectations. Otherwise, any test
        file may be removed from expectations if it has been deleted.

        Returns: A list of web test files that have been deleted.
        """
        if self.options.clean_up_affected_tests_only:
            # TODO(robertma): Improve Git.changed_files so that we can use
            # it here.
            paths = set(self.git.run(
                ['diff', 'origin/master', '-M100%', '--diff-filter=D',
                 '--name-only']).splitlines())
            deleted_tests = set()
            for path in paths:
                test = self._relative_to_web_test_dir(path)
                if test:
                    deleted_tests.add(test)
        else:
            # Remove expectations for all test which have files that
            # were deleted. Paths are already relative to the web_tests
            # directory
            deleted_tests = self._deleted_test_files_in_expectations()
        return deleted_tests

    def _list_renamed_test_files(self):
        """Returns a dictionary mapping tests to their new name.

        Regardless of the command line arguments used this test will only
        return a dictionary for tests affected in the current CL.

        Returns a dictionary mapping source name to destination name.
        """
        out = self.git.run(
            ['diff', 'origin/master', '-M100%', '--diff-filter=R',
             '--name-status'])
        renamed_tests = {}
        for line in out.splitlines():
            _, source_path, dest_path = line.split()
            source_test = self._relative_to_web_test_dir(source_path)
            dest_test = self._relative_to_web_test_dir(dest_path)
            if source_test and dest_test:
                renamed_tests[source_test] = dest_test
        return renamed_tests

    def _relative_to_web_test_dir(self, path_relative_to_repo_root):
        """Returns a path that's relative to the web tests directory."""
        abs_path = self.finder.path_from_chromium_base(
            path_relative_to_repo_root)
        if not abs_path.startswith(self.finder.web_tests_dir()):
            return None
        return self.host.filesystem.relpath(
            abs_path, self.finder.web_tests_dir())

    def _deleted_test_files_in_expectations(self):
        """Returns a list of test files that were deleted.

        Returns a list of test file names that are still in the expectations
        files but no longer exists in the web tests directory.
        """
        deleted_files = set()
        existing_files = {
            self._get_root_file(p)
            for p in self.port.tests()}
        for path in self._test_expectations.expectations_dict:
            for line in self._test_expectations.get_updated_lines(path):
                if not line.test or line.is_glob:
                    continue
                root_test_file = self._get_root_file(line.test)
                if root_test_file not in existing_files:
                    deleted_files.add(root_test_file)
        return deleted_files

    # TODO(robertma): Unit test this method.
    def download_text_baselines(self, test_results):
        """Fetches new baseline files for tests that should be rebaselined.

        Invokes `blink_tool.py rebaseline-cl` in order to download new baselines
        (-expected.txt files) for testharness.js tests that did not crash or
        time out. Then, the platform-specific test is removed from the overall
        failure test dictionary and the resulting dictionary is returned.

        Args:
            test_results: A dictionary of failing test results, mapping test
                names to lists of platforms to SimpleTestResult.

        Returns:
            A pair: A set of tests that are rebaselined, and a modified copy of
            the test_results dictionary containing only tests that couldn't be
            rebaselined.
        """
        tests_to_rebaseline, test_results = self.get_tests_to_rebaseline(
            test_results)
        if not tests_to_rebaseline:
            _log.info('No tests to rebaseline.')
            return tests_to_rebaseline, test_results
        _log.info('Tests to rebaseline:')
        for test in tests_to_rebaseline:
            _log.info('  %s', test)

        blink_tool = self.finder.path_from_blink_tools('blink_tool.py')
        command = [
            'python',
            blink_tool,
            'rebaseline-cl',
            '--verbose',
            '--no-trigger-jobs',
            '--fill-missing',
        ]
        if self.patchset:
            command.append('--patchset=' + str(self.patchset))
        command += tests_to_rebaseline
        self.host.executive.run_command(command)
        return tests_to_rebaseline, test_results

    def get_tests_to_rebaseline(self, test_results):
        """Filters failing tests that can be rebaselined.

        Creates a list of tests to rebaseline depending on the tests' platform-
        specific results. In general, this will be non-ref tests that failed
        due to a baseline mismatch (rather than crash or timeout).

        Args:
            test_results: A dictionary of failing test results, mapping test
                names to lists of platforms to SimpleTestResult.

        Returns:
            A pair: A set of tests to be rebaselined, and a modified copy of
            the test_results dictionary. The tests to be rebaselined should
            include testharness.js tests that failed due to a baseline mismatch.
        """
        new_test_results = copy.deepcopy(test_results)
        tests_to_rebaseline = set()
        for test_name in test_results:
            for platforms, result in test_results[test_name].iteritems():
                if self.can_rebaseline(test_name, result):
                    del new_test_results[test_name][platforms]
                    tests_to_rebaseline.add(test_name)
        return sorted(tests_to_rebaseline), new_test_results

    def can_rebaseline(self, test_name, result):
        """Checks if a test can be rebaselined.

        Args:
            test_name: The test name string.
            result: A SimpleTestResult.
        """
        if self.is_reference_test(test_name):
            return False
        if any(x in result.actual for x in ('CRASH', 'TIMEOUT', 'MISSING')):
            return False
        if self.is_webdriver_test(test_name):
            return False
        return True

    def is_reference_test(self, test_name):
        """Checks whether a given test is a reference test."""
        return bool(self.port.reference_files(test_name))

    def is_webdriver_test(self, test_name):
        """Checks whether a given test is a WebDriver test."""
        return self.finder.is_webdriver_test_path(test_name)

    @memoized
    def _get_try_bots(self):
        return self.host.builders.filter_builders(
            is_try=True, exclude_specifiers={'android'})