Add support for entering/exiting HTML fullscreen from OOPIFs.

third_party/WebKit/Source/web/WebViewImpl.cpp

 /*
  * Copyright (C) 2011, 2012 Google Inc. All rights reserved.
  *
  * Redistribution and use in source and binary forms, with or without
  * modification, are permitted provided that the following conditions are
  * met:
  *
  *     * Redistributions of source code must retain the above copyright
  * notice, this list of conditions and the following disclaimer.
  *     * Redistributions in binary form must reproduce the above
@@ skipping 128 matching lines @@
 #include "public/web/WebFrame.h"
 #include "public/web/WebFrameClient.h"
 #include "public/web/WebHitTestResult.h"
 #include "public/web/WebInputElement.h"
 #include "public/web/WebMeaningfulLayout.h"
 #include "public/web/WebMediaPlayerAction.h"
 #include "public/web/WebNode.h"
 #include "public/web/WebPlugin.h"
 #include "public/web/WebPluginAction.h"
 #include "public/web/WebRange.h"
+#include "public/web/WebScopedUserGesture.h"
 #include "public/web/WebSelection.h"
 #include "public/web/WebTextInputInfo.h"
 #include "public/web/WebViewClient.h"
 #include "public/web/WebWindowFeatures.h"
 #include "web/CompositionUnderlineVectorBuilder.h"
 #include "web/ContextFeaturesClientImpl.h"
 #include "web/ContextMenuAllowedScope.h"
 #include "web/DatabaseClientImpl.h"
 #include "web/DedicatedWorkerGlobalScopeProxyProviderImpl.h"
 #include "web/DevToolsEmulator.h"
@@ skipping 1775 matching lines @@
 
 void WebViewImpl::resize(const WebSize& newSize)
 {
     if (m_shouldAutoResize || m_size == newSize)
         return;
 
     resizeWithTopControls(
         newSize, topControls().height(), topControls().shrinkViewport());
 }
 
+void WebViewImpl::willEnterFullScreen(WebRemoteFrame* fullscreenFrame)
+{
+    FrameOwner* owner = toWebRemoteFrameImpl(fullscreenFrame)->frame()->owner();

[dcheng, 2016/05/19 22:08:39]

In a future patch, let's try moving this to WebRemoteFrame.

[alexmos, 2016/05/19 23:58:19]

Will do, I think that should be possible.

+    HTMLFrameOwnerElement* ownerElement = toHTMLFrameOwnerElement(owner);
+
+    // Let FullscreenController know that |ownerElement| is an ancestor of the
+    // actual fullscreen element, so that it can be treated a little
+    // differently:
+    // - it will need :-webkit-full-screen-ancestor style in addition to
+    //   :-webkit-full-screen.
+    // - it does not need to resend the ToggleFullscreen IPC to the browser
+    //   process.
+    m_fullscreenController->setFullscreenIsForCrossProcessAncestor();

[dcheng, 2016/05/19 22:08:39]

This seems like the main blocker against moving this: is it possible/feasible to
plumb this information through requestFullscreen(), similar to how we're going
to have to plumb the prefix information through?

[alexmos, 2016/05/19 23:58:19]

Yes, I think that should be feasible.  In fact, once we have that information
plumbed into requestFullscreen(), it can skip some work that's already been
performed in the process of the actual fullscreened element (like the user
gesture check) and just set up the fullscreen element stack and the events. 
(Maybe requestFullscreen() can be split into two functions based on this.)  I'm
assuming it's ok to tackle this in a followup per your comment above?  I've
already got a followup in the works for requestFullscreen() (to support A-B-A
hierarchies), so I'll be tinkering with that code anyway.

[dcheng, 2016/05/20 07:01:44]

Yeah, followup is fine. Thanks!

+
+    // Call requestFullScreen() on |ownerElement| to make it the provisional
+    // fullscreen element in FullscreenController, and to prepare
+    // fullscreenchange events that will need to fire on it and its (local)
+    // ancestors. The events will be triggered if/when fullscreen is entered.
+    // Note that requestFullscreen() requires a user gesture.
+    //
+    // TODO(alexmos): currently, WebElement::requestFullScreen assumes prefixed
+    // requests, but in the future, this should plumb in information about
+    // which request type (prefixed or unprefixed) to use for firing fullscreen
+    // events.
+    {
+        WebElement element(ownerElement);

[dcheng, 2016/05/19 22:08:39]

Nit: let's use the core API directly, I looked at this, and I think we're
probably going to remove requestFullScreen() from the public API and move it to
WebPluginContainer.

[alexmos, 2016/05/19 23:58:19]

Done.

+        WebScopedUserGesture userGesture;
+        element.requestFullScreen();
+    }
+}
+
 void WebViewImpl::didEnterFullScreen()
 {
     m_fullscreenController->didEnterFullScreen();
 }
 
 void WebViewImpl::didExitFullScreen()
 {
     m_fullscreenController->didExitFullScreen();
 }
 
+void WebViewImpl::didUpdateFullScreenSize()
+{
+    m_fullscreenController->updateSize();
+}
+
 void WebViewImpl::beginFrame(double lastFrameTimeMonotonic)
 {
     TRACE_EVENT1("blink", "WebViewImpl::beginFrame", "frameTime", lastFrameTimeMonotonic);
     DCHECK(lastFrameTimeMonotonic);
 
     // Create synthetic wheel events as necessary for fling.
     if (m_gestureAnimation) {
         if (m_gestureAnimation->animate(lastFrameTimeMonotonic))
             scheduleAnimation();
         else {
@@ skipping 2560 matching lines @@
 {
     // TODO(oshima): Investigate if this should return the ScreenInfo's scale factor rather than
     // page's scale factor, which can be 1 in use-zoom-for-dsf mode.
     if (!page())
         return 1;
 
     return page()->deviceScaleFactor();
 }
 
 } // namespace blink