[navigation] Removing IsPerNavigationMojoInterface in individual tests.

NavigationClient is now on by default, so removing an unused branch in tests.

Also removing a test that does not make sense anymore.

Bug: 784904
Change-Id: Ibf91fdfbdc23d03a9f5a966ca02c82314fcdca97
Reviewed-on: https://chromium-review.googlesource.com/c/chromium/src/+/1821558
Commit-Queue: Arthur Hemery <ahemery@chromium.org>
Reviewed-by: Nasko Oskov <nasko@chromium.org>
Cr-Commit-Position: refs/heads/master@{#708177}

content/browser/security_exploit_browsertest.cc

@@ -1602,17 +1602,11 @@ IN_PROC_BROWSER_TEST_F(SecurityExploitViaDisabledWebSecurityTest,
   RenderProcessHostKillWaiter process_kill_waiter(rfh->GetProcess());
 
   mojo::PendingAssociatedRemote<mojom::NavigationClient> navigation_client;
-  if (IsPerNavigationMojoInterfaceEnabled()) {
-    auto navigation_client_receiver =
-        navigation_client.InitWithNewEndpointAndPassReceiver();
-    rfh->frame_host_receiver_for_testing().impl()->BeginNavigation(
-        std::move(common_params), std::move(begin_params), mojo::NullRemote(),
-        std::move(navigation_client), mojo::NullRemote());
-  } else {
-    rfh->frame_host_receiver_for_testing().impl()->BeginNavigation(
-        std::move(common_params), std::move(begin_params), mojo::NullRemote(),
-        mojo::NullAssociatedRemote(), mojo::NullRemote());
-  }
+  auto navigation_client_receiver =
+      navigation_client.InitWithNewEndpointAndPassReceiver();
+  rfh->frame_host_receiver_for_testing().impl()->BeginNavigation(
+      std::move(common_params), std::move(begin_params), mojo::NullRemote(),
+      std::move(navigation_client), mojo::NullRemote());
   EXPECT_EQ(bad_message::RFH_BASE_URL_FOR_DATA_URL_SPECIFIED,
             process_kill_waiter.Wait());
 

content/browser/site_per_process_browsertest.cc

@@ -11843,109 +11843,6 @@ class CommitMessageOrderReverser : public DidCommitNavigationInterceptor {
 
 }  // namespace
 
-// Regression test for https://crbug.com/877239, simulating the following
-// scenario:
-//
-//  1) http://a.com/empty.html is loaded in a main frame.
-//  2) Dynamically by JS, a same-site child frame is added:
-//       <iframe 'src=http://a.com/title1.html'/>.
-//  3) The initial byte of the response for `title1.html` arrives, causing
-//     FrameMsg_CommitNavigation to be sent to the same renderer.
-//  4) Just before processing this message, however, `main.html` navigates
-//     the iframe to http://baz.com/title2.html, which results in mojom::Frame::
-//     BeginNavigation being called on the RenderFrameHost.
-//  5) Suppose that immediately afterwards, `main.html` enters a busy-loop.
-//  6) The cross site navigation in the child frame starts, the first response
-//     byte arrives quickly, and thus the navigation commits quickly.
-//  6.1) FrameTreeNode::has_committed_real_load is set to true for the child.
-//  6.2) The same-site RenderFrame in the child FrameTreeNode is swapped out,
-//       i.e. UnfreezableFrameMsg_SwapOut is sent.
-//  7) The renderer for site instance `a.com` exits from the busy loop,
-//     and starts processing messages in order:
-//  7.1) The first being processed is FrameMsg_CommitNavigation, so a
-//       provisional load is created and immediately committed to
-//       http://a.com/title1.html.
-//  7.2) Because at the time the same-site child RenderFrame was created,
-//       there had been no real load committed in the child frame, and because
-//       the navigation from the initial empty document to the first real
-//       document was same-origin, the global object is reused and the
-//       RemoteInterfaceProvider of the RenderFrame is not rebound.
-//  7.3) The obsoleted load in the same-site child frame commits, calling
-//       mojom::Frame::DidCommitProvisionalLoad, however, with
-//       |interface_provider_request| being null.
-//  8) RenderFrameHostImpl::DidCommitProvisionalLoad sees that a real load was
-//     already committed in the frame, but |interface_provider_request| is
-//     missing. However, it also sees that the frame was waiting for a swap-out
-//     ACK, so ignores the commit, and does not kill the renderer process.
-//
-// In the simulation of this scenario, we simulate (5) not by delaying
-// renderer-side processing of the CommmitNavigation message, but by delaying
-// browser-side processing of the response to it, of DidCommitProvisionalLoad.
-IN_PROC_BROWSER_TEST_F(SitePerProcessBrowserTest,
-                       InterfaceProviderRequestIsOptionalForRaceyFirstCommits) {
-  // This test does not make sense anymore for the new NavigationClient way
-  // of doing navigations. The first iframe navigation will get instantly
-  // destroyed by the NavigationClient disconnection handle
-  // NavigationRequest::OnRendererAbortedNavigation.
-  // The pipe gets closed in the renderer when the javascript starts the
-  // second navigation.
-  // TODO(ahemery): Remove this test when IsPerNavigationMojoInterfaceEnabled
-  // is on by default.
-  if (IsPerNavigationMojoInterfaceEnabled())
-    return;
-
-  const GURL kMainFrameUrl(
-      embedded_test_server()->GetURL("a.com", "/empty.html"));
-  const GURL kSubframeSameSiteUrl(
-      embedded_test_server()->GetURL("a.com", "/title1.html"));
-  const GURL kCrossSiteSubframeUrl(
-      embedded_test_server()->GetURL("baz.com", "/title2.html"));
-
-  const auto kAddSameSiteDynamicSubframe = JsReplace(
-      "var f = document.createElement(\"iframe\");"
-      "f.src=$1;"
-      "document.body.append(f);",
-      kSubframeSameSiteUrl);
-  const auto kNavigateSubframeCrossSite =
-      JsReplace("f.src = $1;", kCrossSiteSubframeUrl);
-  const std::string kExtractSubframeUrl =
-      "window.domAutomationController.send(f.src);";
-
-  ASSERT_TRUE(NavigateToURL(shell(), kMainFrameUrl));
-
-  const auto* main_rfh_site_instance =
-      shell()->web_contents()->GetMainFrame()->GetSiteInstance();
-
-  auto did_start_deferring_commit_callback =
-      base::BindLambdaForTesting([&](RenderFrameHost* subframe_rfh) {
-        // Verify that the subframe starts out as same-process with its parent.
-        ASSERT_EQ(main_rfh_site_instance, subframe_rfh->GetSiteInstance());
-
-        // Trigger the second commit now that we are deferring the first one.
-        ASSERT_TRUE(ExecuteScript(shell(), kNavigateSubframeCrossSite));
-      });
-
-  CommitMessageOrderReverser commit_order_reverser(
-      shell()->web_contents(), kSubframeSameSiteUrl /* deferred_url */,
-      std::move(did_start_deferring_commit_callback));
-
-  ASSERT_TRUE(ExecuteScript(shell(), kAddSameSiteDynamicSubframe));
-  commit_order_reverser.WaitForBothCommits();
-
-  // Verify that:
-  //  - The cross-site navigation in the sub-frame was committed and the
-  //    same-site navigation was ignored.
-  //  - The parent frame thinks so, too.
-  //  - The renderer process corresponding to the sub-frame with the ignored
-  //    commit was not killed. This is verified implicitly: this is the same
-  //    renderer process where the parent RenderFrame lives, so if the call to
-  //    ExecuteScriptAndExtractString succeeds here, the process is still alive.
-  std::string actual_subframe_url;
-  ASSERT_TRUE(ExecuteScriptAndExtractString(shell(), kExtractSubframeUrl,
-                                            &actual_subframe_url));
-  EXPECT_EQ(kCrossSiteSubframeUrl.spec(), actual_subframe_url);
-}
-
 // Create an out-of-process iframe that causes itself to be detached during
 // its layout/animate phase. See https://crbug.com/802932.
 // Disabled on Android due to flakiness, https://crbug.com/809580.