Temporarily avoid process swaps for renderer navigations away from an app.

chrome/renderer/render_view.cc

Index: chrome/renderer/render_view.cc
diff --git a/chrome/renderer/render_view.cc b/chrome/renderer/render_view.cc
index f62a2e97fdbee4ef76cdf022aa77debd66836845..ed043ae8446287d65ddd0a035dc1fd7b28652990 100644
--- a/chrome/renderer/render_view.cc
+++ b/chrome/renderer/render_view.cc
@@ -391,6 +391,20 @@ static bool CrossesExtensionExtents(WebFrame* frame, const GURL& new_url) {
   return !ExtensionRendererInfo::InSameExtent(old_url, new_url);
 }
 
+// Returns true if the frame is in an extension app's extent.
+static bool InExtensionExtent(WebFrame* frame) {
+  if (!RenderThread::current())

[darin (slow to review), 2010/12/14 17:17:40]

what is the purpose of this null check?  i don't see RenderThread::current()
dereferenced below.

[Charlie Reis, 2010/12/14 18:52:29]

We don't need this function anymore, but the same check is in
CrossesExtensionExtents above.  Erik, do you know why it's there?

+    return false;
+
+  // If the URL is still empty, this is a window.open navigation. Check the
+  // opener's URL.
+  GURL old_url(frame->url());
+  if (old_url.is_empty() && frame->opener())

[darin (slow to review), 2010/12/14 17:17:40]

Why do you bother checking frame->opener() here?  At the callsite, you check
frame->opener() as well.  If the opener is non-null here, then the result of
InExtensionExtent will be ignored anyways.

[Charlie Reis, 2010/12/14 18:52:29]

No longer needed here, but still used in CrossesExtensionExtents.

+    old_url = frame->opener()->url();
+
+  return ExtensionRendererInfo::GetByURL(old_url) != NULL;
+}
+
 // Returns the ISO 639_1 language code of the specified |text|, or 'unknown'
 // if it failed.
 static std::string DetermineTextLanguage(const string16& text) {
@@ -3001,7 +3015,13 @@ WebNavigationPolicy RenderView::decidePolicyForNavigation(
     // TODO(erikkay) This is happening inside of a check to is_content_initiated
     // which means that things like the back button won't trigger it.  Is that
     // OK?
-    if (CrossesExtensionExtents(frame, url)) {
+    if (CrossesExtensionExtents(frame, url) &&
+        // Temporary workaround for crbug.com/65953: Only swap processes if we

[darin (slow to review), 2010/12/14 17:17:40]

nit: Comments like this should have a TODO(creis)

[Charlie Reis, 2010/12/14 18:52:29]

Done.

+        // are not already in an app, and only if window.opener is null, since
+        // we do not yet restore window.opener if the window navigates back to
+        // this process later.
+        !InExtensionExtent(frame) &&
+        frame->opener() == NULL) {

[darin (slow to review), 2010/12/14 17:17:40]

style nit: prefer "!frame->opener()"

[Charlie Reis, 2010/12/14 18:52:29]

Simplified by removing this part of the check and modifying
CrossesExtensionExtents instead.

       // Include the referrer in this case since we're going from a hosted web
       // page. (the packaged case is handled previously by the extension
       // navigation test)