Switch the page-capturing machinery to use the new hooks.

chrome/browser/translate/translate_manager_browsertest.cc

 // Copyright (c) 2012 The Chromium Authors. All rights reserved.
 // Use of this source code is governed by a BSD-style license that can be
 // found in the LICENSE file.
 
 #include "components/translate/core/browser/translate_manager.h"
 
 #include "base/memory/scoped_ptr.h"
 #include "chrome/browser/chrome_notification_types.h"
 #include "chrome/browser/prefs/session_startup_pref.h"
 #include "chrome/browser/translate/chrome_translate_client.h"
@@ skipping 54 matching lines @@
 #endif
 IN_PROC_BROWSER_TEST_F(TranslateManagerBrowserTest,
                        MAYBE_TranslateSessionRestore) {
   scoped_ptr<test::CldDataHarness> cld_data_harness =
       test::CldDataHarnessFactory::Get()->CreateCldDataHarness();
   ASSERT_NO_FATAL_FAILURE(cld_data_harness->Init());
   content::WebContents* current_web_contents =
       browser()->tab_strip_model()->GetActiveWebContents();
   content::Source<content::WebContents> source(current_web_contents);
 
+  // LOG(INFO) << "About to set up the fr_language_detected_signal";
+

[dglazkov, 2015/10/09 23:22:22]

Would love your insight here. After this change, the OnLanguageDetermined timing
is no longer 500ms+ -- it's called as soon as the page content is ready. This
exposes a fun race in this code here: before the change, the "About to set up...
" log comes up before the "Language detected" log above.

After this change, it's the other way around, which means that the
WindowedNotificationObserver thing just sits there waiting forever -- the
notification is issued before we're able to set up the observer.

How do I deal with cases like this? Any ideas?

[jochen (gone - plz use gerrit), 2015/10/12 11:23:50]

I think the browser test harness is created before the session is restored, so
if you create the notification observer as a member of
TranslateManagerBrowserTest it should work

[dglazkov, 2015/10/17 00:29:24]

The problem here is that the observer needs to have a source, and that's unknown
until the InProcBrowserTests starts the loop, and it looks like the
implementation there is pretty simplistic. For situations where the Webcontents
already exists (I am guessing this includes session restore), it assumes that
all important things will happen after load finished:

https://code.google.com/p/chromium/codesearch#chromium/src/chrome/test/base/i...

That's what makes the changed timing no longer work.

   ui_test_utils::WindowedNotificationObserverWithDetails<
       translate::LanguageDetectionDetails>
       fr_language_detected_signal(chrome::NOTIFICATION_TAB_LANGUAGE_DETERMINED,
                                   source);
   fr_language_detected_signal.Wait();
 }