Remove unused WebFrame::swap() path for local frames.

Source/web/WebLocalFrameImpl.cpp

Index: Source/web/WebLocalFrameImpl.cpp
diff --git a/Source/web/WebLocalFrameImpl.cpp b/Source/web/WebLocalFrameImpl.cpp
index f48c85566e5f6a66c23438876de38b3076970bd2..fd63993a5853904f8315f022fd7aa279d3b8e236 100644
--- a/Source/web/WebLocalFrameImpl.cpp
+++ b/Source/web/WebLocalFrameImpl.cpp
@@ -1857,7 +1857,11 @@ void WebLocalFrameImpl::loadJavaScriptURL(const KURL& url)
 void WebLocalFrameImpl::initializeToReplaceRemoteFrame(WebRemoteFrame* oldWebFrame)
 {
     Frame* oldFrame = toCoreFrame(oldWebFrame);
-    OwnPtrWillBePersistent<FrameOwner> tempOwner = RemoteBridgeFrameOwner::create(nullptr, SandboxNone);

[dcheng, 2015/02/23 21:05:12]

Btw, is there ever a reason we'd want a stack Persistent? This seems like
something that the Oilpan plugin should warn about.

[haraken, 2015/02/24 00:09:07]

OwnPtrWillBeRawPtr is a right pointer here.

In unit tests, we sometimes need to use persistent handles for on-stack pointers
to protect the pointers from getting collected by
Heap::collectGarbage(NoHeapPointersOnStack) (we use NoHeapPointersOnStack to
make the behavior deterministic (i.e., not conservative)). However, this happens
only in unit tests (*Test.cpp), so I agree that we should improve the plugin to
detect this kind of error.

+    // Note: this *always* temporarily sets a frame owner, even for main frames!
+    // When a core Frame is created with no owner, it attempts to set itself as
+    // the main frame of the Page. However, this is a provisional frame, and may
+    // disappear, so Page::m_mainFrame can't be updated just yet.
+    OwnPtrWillBeRawPtr<FrameOwner> tempOwner = RemoteBridgeFrameOwner::create(nullptr, SandboxNone);
     m_frame = LocalFrame::create(&m_frameLoaderClientImpl, oldFrame->host(), tempOwner.get());
     m_frame->setOwner(oldFrame->owner());
     m_frame->tree().setName(oldFrame->tree().name());