Fix some flaky tests.

content/test/content_browser_test_utils_internal.cc

Index: content/test/content_browser_test_utils_internal.cc
diff --git a/content/test/content_browser_test_utils_internal.cc b/content/test/content_browser_test_utils_internal.cc
index f7fcfe4d17208802fec37e0d8b66770990e86a1b..392439525ba2bc8451d4dc5720447e0382d6d94b 100644
--- a/content/test/content_browser_test_utils_internal.cc
+++ b/content/test/content_browser_test_utils_internal.cc
@@ -379,7 +379,9 @@ UrlCommitObserver::UrlCommitObserver(FrameTreeNode* frame_tree_node,
                                        ->GetAsWebContents()),
       frame_tree_node_id_(frame_tree_node->frame_tree_node_id()),
       url_(url),
-      message_loop_runner_(new MessageLoopRunner) {}
+      message_loop_runner_(
+          new MessageLoopRunner(MessageLoopRunner::QuitMode::IMMEDIATE)) {

[nasko, 2016/11/23 17:48:29]

I imagine all test helpers waiting for commit will likely expect this type of
behavior.

[Alexander Semashko, 2016/11/23 18:38:08]

I'm planning to look into other helper classes a bit later, once we agree on
what to do with MessageLoopRunner globally (please see the discussion on
chromium-dev).
In this CL I only focused on tests that have shown flakiness on our bots.

[nasko, 2016/11/24 00:00:44]

Acknowledged.

+}
 
 UrlCommitObserver::~UrlCommitObserver() {}