[Experiment] purge-and-suspend

tools/perf/page_sets/memory_test.py

+# Copyright 2014 The Chromium Authors. All rights reserved.

[bashi, 2016/07/08 04:16:22]

Are you going to add this?

[tasak, 2016/07/08 06:47:23]

No. This code is just for explaining how I measured renderer memory usages.

+# Use of this source code is governed by a BSD-style license that can be
+# found in the LICENSE file.
+import logging
+import os
+
+from page_sets.login_helpers import dropbox_login
+from page_sets.login_helpers import google_login
+
+from telemetry.page import page as page_module
+from telemetry.page import shared_page_state
+from telemetry import story
+
+DUMP_WAIT_TIME = 10
+DUMP_COUNT = 10
+
+class MemoryTestPage(page_module.Page):
+    def __init__(self, story_set, group, url, page_spec):
+        name = 'load:%s:%s' % (group, page_spec.name)
+        super(MemoryTestPage, self).__init__(
+            page_set=story_set,
+            name=name,
+            url=url,
+            shared_page_state_class=shared_page_state.SharedPageState,
+            credentials_path='data/credentials.json')
+        self._page_spec = page_spec
+
+    @property
+    def platform(self):
+        return self.story_set.PLATFORM
+
+    def _dumpWithInteraction(self, action_runner, text, do_gc=False):
+        with action_runner.CreateInteraction(text):
+            action_runner.Wait(3)
+            if do_gc:
+                action_runner.ForceGarbageCollection()
+            action_runner.SimulateMemoryPressureNotification('critical')
+            action_runner.Wait(3)
+            if not action_runner.tab.browser.DumpMemory():
+                logging.error('Unable to get a memory dump for %s.', self.name)
+
+    def _RunPageInteractions(self, action_runner):
+        action_runner.Wait(60)
+        if not action_runner.tab.browser.DumpMemory():
+            logging.error('Unable to get a memory dump for %s.', self.name)
+
+    def _RunPageAndMoveToBackground(self, action_runner):
+        action_runner.Wait(60)
+        # Force V8 GC.
+        action_runner.ForceGarbageCollection()
+        action_runner.Wait(10)
+        # Dump before moving the page to background.
+        if not action_runner.tab.browser.DumpMemory():
+            logging.error('Unable to get a memory dump for %s.', self.name)
+        action_runner.Wait(3)
+        # Open blank tab (Move the page to background).
+        action_runner.tab.browser.tabs.New()
+        # Wait until purge-and-suspend
+        # Needs "--purge-and-suspend-time=X and X < 20"
+        action_runner.Wait(120)
+        if not action_runner.tab.browser.DumpMemory():
+            logging.error('Unable to get a memory dump for %s.', self.name)
+
+    def _RunPageAndMoveToBackgroundWithNoDump(self, action_runner):
+        action_runner.Wait(60)
+        action_runner.ForceGarbageCollection()
+        action_runner.Wait(10)
+        action_runner.tab.browser.tabs.New()
+        action_runner.Wait(540)
+
+
+    def RunPageInteractions(self, action_runner):
+        action_runner.tab.WaitForDocumentReadyStateToBeComplete()
+        if self._page_spec.post_load_hook:
+            self._page_spec.post_load_hook(self.platform, action_runner)
+        self._RunPageAndMoveToBackground(action_runner)
+        #self._RunPageAndMoveToBackgroundWithNoDump(action_runner)
+
+    def RunNavigateSteps(self, action_runner):
+        if self._page_spec.login_hook:
+            self._page_spec.login_hook(action_runner, self.credentials_path)
+        super(MemoryTestPage, self).RunNavigateSteps(action_runner)
+
+
+def _LogIntoGoogleAccount(action_runner, credentials_path):
+  google_login.LoginGoogleAccount(action_runner, 'googletest', credentials_path)
+
+
+def _LogIntoGoogleAccountAndSetUpGmailSession(action_runner, credentials_path):
+  _LogIntoGoogleAccount(action_runner, credentials_path)
+
+  # Navigating to https://mail.google.com immediately leads to an infinite
+  # redirection loop due to a bug in WPR (see
+  # https://github.com/chromium/web-page-replay/issues/70). We therefore first
+  # navigate to a sub-URL to set up the session and hit the resulting
+  # redirection loop. Afterwards, we can safely navigate to
+  # https://mail.google.com.
+  action_runner.Navigate(
+      'https://mail.google.com/mail/mu/mp/872/trigger_redirection_loop')
+  action_runner.tab.WaitForDocumentReadyStateToBeComplete()
+
+
+def _LogIntoDropboxAccount(action_runner, credentials_path):
+  dropbox_login.LoginAccount(action_runner, 'dropbox', credentials_path)
+
+
+def _CloseInboxInterstitialAndWaitUntilGmailReady(platform, action_runner):
+  if platform == 'desktop':
+    # Wait until the UI loads.
+    action_runner.WaitForJavaScriptCondition(
+        'document.getElementById("loading").style.display === "none"')
+  elif platform == 'mobile':
+    # Close the "Get Inbox by Gmail" interstitial.
+    action_runner.WaitForJavaScriptCondition(
+        'document.querySelector("#isppromo a") !== null')
+    action_runner.ExecuteJavaScript(
+        'document.querySelector("#isppromo a").click()')
+    # Wait until the UI loads.
+    action_runner.WaitForJavaScriptCondition(
+        'document.getElementById("apploadingdiv").style.height === "0px"')
+
+
+def _WaitUntilBubblesReady(_, action_runner):
+  # The #logo element is removed right before the main menu is displayed.
+  action_runner.WaitForJavaScriptCondition(
+      'document.getElementById("logo") === null')
+
+
+def _WaitUntilSpyChaseReady(_, action_runner):
+  # The background of the game canvas is set when the "Tap screen to play"
+  # caption is displayed.
+  action_runner.WaitForJavaScriptCondition(
+      'document.querySelector("#game canvas").style.background !== ""')
+
+
+def _WaitUntilFlickrReady(_, action_runner):
+  # Wait until the 'Recently tagged' view loads.
+  action_runner.WaitForJavaScriptCondition('''
+      document.querySelector(
+          '.search-photos-everyone-trending-view .photo-list-view') !== null''')
+
+
+class _PageSpec(object):
+
+  def __init__(self, name, url, login_hook=None, post_load_hook=None):
+    assert isinstance(url, (dict, str))
+    self._name = name
+    self._url = url
+    self._login_hook = login_hook
+    self._post_load_hook = post_load_hook
+
+  @property
+  def name(self):
+    return self._name
+
+  @property
+  def url(self):
+    return self._url
+
+  @property
+  def login_hook(self):
+    return self._login_hook
+
+  @property
+  def post_load_hook(self):
+    return self._post_load_hook
+
+
+_SINGLE_PAGE_SPECS = {
+    # Search and e-commerce.
+    'search': [
+        _PageSpec(
+            name='google',
+            url='https://www.google.com/#hl=en&q=science'),
+        _PageSpec(
+            name='baidu',
+            url='https://www.baidu.com/s?word=google'),
+        _PageSpec(
+            name='yahoo',
+            url='https://search.yahoo.com/search;_ylt=?p=google'),
+        _PageSpec(
+            name='amazon',
+            url='https://www.amazon.com/s/?field-keywords=nexus'),
+        # "ali_trackid" in the URL suppresses "Download app" interstitial.
+        _PageSpec(
+            name='taobao',
+            url={'desktop': 'https://world.taobao.com/',
+                 'mobile': 'http://m.intl.taobao.com/?ali_trackid'}),
+        _PageSpec(
+            name='yandex',
+            url='https://yandex.ru/touchsearch?text=science'),
+        # Redirects to the "http://" version.
+        _PageSpec(
+            name='ebay',
+            url='https://www.ebay.com/sch/i.html?_nkw=headphones'),
+    ],
+
+    # Social networks.
+    'social': [
+        # Using Facebook login often causes "404 Not Found" with WPR.
+        _PageSpec(
+            name='facebook',
+            url='https://www.facebook.com/rihanna'),
+        _PageSpec(
+            name='twitter',
+            url='https://www.twitter.com/justinbieber?skip_interstitial=true'),
+        # Due to the deterministic date injected by WPR (February 2008), the
+        # cookie set by https://vk.com immediately expires, so the page keeps
+        # refreshing indefinitely on mobile
+        # (see https://github.com/chromium/web-page-replay/issues/71).
+        _PageSpec(
+            name='vk',
+            url={'desktop': 'https://vk.com/sbeatles'}),
+        _PageSpec(
+            name='instagram',
+            url='https://www.instagram.com/selenagomez/'),
+        _PageSpec(
+            name='pinterest',
+            url='https://uk.pinterest.com/categories/popular/'),
+        # Redirects to the "http://" version.
+        _PageSpec(
+            name='tumblr',
+            url='https://50thousand.tumblr.com/'),
+    ],
+
+    # News, discussion and knowledge portals and blogs.
+    'news': [
+        # Redirects to the "http://" version.
+        _PageSpec(
+            name='bbc',
+            url='https://www.bbc.co.uk/news/world-asia-china-36189636'),
+        # Using "https://" shows "Your connection is not private".
+        _PageSpec(
+            name='cnn',
+            url=(
+                'http://edition.cnn.com/2016/05/02/health/three-habitable-planets-earth-dwarf-star/index.html')),
+        _PageSpec(
+            name='reddit',
+            url={'desktop': (
+                     'https://www.reddit.com/r/AskReddit/comments/4hi90e/whats_your_best_wedding_horror_story/'),
+                 'mobile': (
+                     'https://m.reddit.com/r/AskReddit/comments/4hi90e/whats_your_best_wedding_horror_story/')}),
+        # Using "https://" hangs and shows "This site can't be reached".
+        _PageSpec(
+            name='qq',
+            url='http://news.qq.com/a/20160503/003186.htm'),
+        # Using "https://" leads to missing images and scripts on mobile (due
+        # to mixed content). The desktop page
+        # (http://news.sohu.com/20160503/n447433356.shtml) almost always fails
+        # to completely load due to
+        # https://github.com/chromium/web-page-replay/issues/74.
+        _PageSpec(
+            name='sohu',
+            url={'mobile': 'http://m.sohu.com/n/447433356/'}),
+        _PageSpec(
+            name='wikipedia',
+            url='https://en.wikipedia.org/wiki/Science'),
+    ],
+
+    # Audio and video.
+    'media': [
+        # No way to disable autoplay on desktop.
+        _PageSpec(
+            name='youtube',
+            url='https://www.youtube.com/watch?v=QGfhS1hfTWw&autoplay=false'),
+        # The side panel with related videos doesn't show on desktop due to
+        # https://github.com/chromium/web-page-replay/issues/74.
+        _PageSpec(
+            name='dailymotion',
+            url=(
+                'https://www.dailymotion.com/video/x489k7d_street-performer-shows-off-slinky-skills_fun?autoplay=false')),
+        _PageSpec(
+            name='google_images',
+            url='https://www.google.co.uk/search?tbm=isch&q=love'),
+        # No way to disable autoplay on desktop. Album artwork doesn't load due
+        # to https://github.com/chromium/web-page-replay/issues/73.
+        _PageSpec(
+            name='soundcloud',
+            url='https://soundcloud.com/lifeofdesiigner/desiigner-panda'),
+        _PageSpec(
+            name='9gag',
+            url='https://www.9gag.com/'),
+        _PageSpec(
+            name='flickr',
+            url='https://www.flickr.com/photos/tags/farm',
+            post_load_hook=_WaitUntilFlickrReady),
+    ],
+
+    # Online tools (documents, emails, storage, ...).
+    'tools': [
+        _PageSpec(
+            name='docs',
+            url=(
+                'https://docs.google.com/document/d/1GvzDP-tTLmJ0myRhUAfTYWs3ZUFilUICg8psNHyccwQ/edit?usp=sharing')),
+        _PageSpec(
+            name='gmail',
+            url='https://mail.google.com/mail/',
+            login_hook=_LogIntoGoogleAccountAndSetUpGmailSession,
+            post_load_hook=_CloseInboxInterstitialAndWaitUntilGmailReady),
+        _PageSpec(
+            name='maps',
+            url='https://www.google.com/maps/place/London,+UK/'),
+        _PageSpec(
+            name='stackoverflow',
+            url=(
+                'https://stackoverflow.com/questions/36827659/compiling-an-application-for-use-in-highly-radioactive-environments')),
+        _PageSpec(
+            name='dropbox',
+            url='https://www.dropbox.com',
+            login_hook=_LogIntoDropboxAccount),
+        _PageSpec(
+            name='weather',
+            url='https://weather.com/en-GB/weather/today/l/USCA0286:1:US'),
+        _PageSpec(
+            name='drive',
+            url='https://drive.google.com/drive/my-drive',
+            login_hook=_LogIntoGoogleAccount),
+    ],
+
+    # In-browser games (HTML5 and Flash).
+    'games': [
+        _PageSpec(
+            name='bubbles',
+            url=(
+                'https://games.cdn.famobi.com/html5games/s/smarty-bubbles/v010/?fg_domain=play.famobi.com&fg_uid=d8f24956-dc91-4902-9096-a46cb1353b6f&fg_pid=4638e320-4444-4514-81c4-d80a8c662371&fg_beat=620'),
+            post_load_hook=_WaitUntilBubblesReady),
+        # Using "https://" hangs and shows "This site can't be reached".
+        _PageSpec(
+            name='lazors',
+            url='http://www8.games.mobi/games/html5/lazors/lazors.html'),
+        # Using "https://" shows "Your connection is not private".
+        _PageSpec(
+            name='spychase',
+            url='http://playstar.mobi/games/spychase/index.php',
+            post_load_hook=_WaitUntilSpyChaseReady),
+        # Desktop only (requires Flash). Using "https://" causes
+        # "404 Not Found" during WPR recording.
+        _PageSpec(
+            name='miniclip',
+            url={'desktop': 'http://www.miniclip.com/games/en/'}),
+        # Desktop only (requires Flash).
+        _PageSpec(
+            name='alphabetty',
+            url={'desktop': 'https://king.com/play/alphabetty'}),
+    ],
+}
+
+class MemoryTestPageSet(story.StorySet):
+
+  PLATFORM = NotImplemented
+
+  def __init__(self):
+    super(MemoryTestPageSet, self).__init__()
+
+    param = os.getenv("MEMORY_TEST_PARAM")
+
+    for group_name, page_specs in iter(_SINGLE_PAGE_SPECS.items()):
+        if param is not None and group_name != param:
+            continue
+        for page_spec in page_specs:
+            url = page_spec.url
+            if isinstance(url, dict):
+                url = url.get(self.PLATFORM)
+                if url is None:
+                    continue  # URL not supported on the platform.
+            self.AddStory(MemoryTestPage(self, group_name, url, page_spec))