Add Start/Stop event signalling on Prerenders.

chrome/browser/prerender/prerender_browsertest.cc

Index: chrome/browser/prerender/prerender_browsertest.cc
diff --git a/chrome/browser/prerender/prerender_browsertest.cc b/chrome/browser/prerender/prerender_browsertest.cc
index 97f7be961289c398b93fe7edfa6ae2875b05b596..3b0f871d96861c8a9bbe998b969bf14880f4fddf 100644
--- a/chrome/browser/prerender/prerender_browsertest.cc
+++ b/chrome/browser/prerender/prerender_browsertest.cc
@@ -2276,12 +2276,44 @@ IN_PROC_BROWSER_TEST_F(PrerenderBrowserTest, PrerenderCancelAll) {
                    1);
   // Post a task to cancel all the prerenders.
   MessageLoop::current()->PostTask(
-      FROM_HERE,
-      base::Bind(&CancelAllPrerenders, GetPrerenderManager()));
+      FROM_HERE, base::Bind(&CancelAllPrerenders, GetPrerenderManager()));
   content::RunMessageLoop();
   EXPECT_TRUE(GetPrerenderContents() == NULL);
 }
 
+IN_PROC_BROWSER_TEST_F(PrerenderBrowserTest, PrerenderEvents) {

[mmenke, 2012/12/07 17:11:48]

Suggest a test where the page itself cancels the prerender, and we make sure
stopped is set, or not, depending on the other CL.

[mmenke, 2012/12/07 19:12:47]

That test could actually just be added to the corresponding RemovingLink test.

In fact, I think that we should add event testing to all the other RemovingLink
tests, if we can reasonably manage to do so.  I'm not sure that we can
reasonably make a test for the
"PrerenderPageRemovingLinkWithTwoLinksRemovingOne" test (Though I've love to
have a test if we can manage it in a way that is results in non-flaky failures
on regression).

I'd also suggest some others:
* Make sure the events work for a prerender that starts life pending prerender,
which we end up starting and then cancelling (Due to visiting the source
prerender).
* Cancel a pending prerender while it's still pending, and make sure we never
start the prerender or create a prerender contents.  Should also make sure we
never send the send event for the prerender.

I think these make the most sense as browser_tests, to make sure everything
hooked in all the way through, and shouldn't be too hard to manage, but if you
feel some make more sense as unit tests, that's fine.

[gavinp, 2012/12/14 21:48:52]

Done.

+  PrerenderTestURL("files/prerender/prerender_page.html",
+                   FINAL_STATUS_CANCELLED, 1);
+  WebContents* web_contents = chrome::GetActiveWebContents(browser());
+  bool received_prerender_started = false;
+  bool received_prerender_stopped = false;
+  ASSERT_TRUE(content::ExecuteJavaScriptAndExtractBool(
+      web_contents->GetRenderViewHost(), L"",
+      L"window.domAutomationController.send(receivedPrerenderStartEvent)",
+      &received_prerender_started));
+  ASSERT_TRUE(content::ExecuteJavaScriptAndExtractBool(
+      web_contents->GetRenderViewHost(), L"",
+      L"window.domAutomationController.send(receivedPrerenderStopEvent)",
+      &received_prerender_stopped));
+  ASSERT_TRUE(received_prerender_started);
+  ASSERT_FALSE(received_prerender_stopped);
+
+  MessageLoop::current()->PostTask(
+      FROM_HERE, base::Bind(&CancelAllPrerenders, GetPrerenderManager()));
+  content::RunMessageLoop();
+
+  ASSERT_TRUE(content::ExecuteJavaScriptAndExtractBool(
+      web_contents->GetRenderViewHost(), L"",
+      L"window.domAutomationController.send(receivedPrerenderStartEvent)",
+      &received_prerender_started));

[mmenke, 2012/12/07 17:11:48]

Is there a reason for this?  I don't object to it, just wondering if there's
some motivation I'm missing.

[gavinp, 2012/12/14 21:48:52]

Nope, none. I think some of these ASSERTs should be EXPECTS though.

+  ASSERT_TRUE(content::ExecuteJavaScriptAndExtractBool(
+      web_contents->GetRenderViewHost(), L"",
+      L"window.domAutomationController.send(receivedPrerenderStopEvent)",
+      &received_prerender_stopped));
+  ASSERT_TRUE(received_prerender_started);
+  ASSERT_TRUE(received_prerender_stopped);
+}
+
 // Prerendering and history tests.
 // The prerendered page is navigated to in several ways [navigate via
 // omnibox, click on link, key-modified click to open in background tab, etc],