Make content text retrieval in ChromeRenderViewObserver never force layout.

chrome/renderer/chrome_render_view_observer.cc

 // Copyright (c) 2012 The Chromium Authors. All rights reserved.
 // Use of this source code is governed by a BSD-style license that can be
 // found in the LICENSE file.
 
 #include "chrome/renderer/chrome_render_view_observer.h"
 
 #include "base/bind.h"
 #include "base/bind_helpers.h"
 #include "base/command_line.h"
 #include "base/debug/crash_logging.h"
@@ skipping 62 matching lines @@
 using blink::WebSecurityOrigin;
 using blink::WebSize;
 using blink::WebString;
 using blink::WebTouchEvent;
 using blink::WebURL;
 using blink::WebURLRequest;
 using blink::WebView;
 using blink::WebVector;
 using blink::WebWindowFeatures;
 
-// Delay in milliseconds that we'll wait before capturing the page contents
-// and thumbnail.
+// Delay in milliseconds that we'll wait before capturing the page contents.
 static const int kDelayForCaptureMs = 500;
 
-// Typically, we capture the page data once the page is loaded.
-// Sometimes, the page never finishes to load, preventing the page capture
-// To workaround this problem, we always perform a capture after the following
-// delay.
-static const int kDelayForForcedCaptureMs = 6000;
-
 // define to write the time necessary for thumbnail/DOM text retrieval,
 // respectively, into the system debug log
 // #define TIME_TEXT_RETRIEVAL
 
 // maximum number of characters in the document to index, any text beyond this
 // point will be clipped
 static const size_t kMaxIndexChars = 65535;
 
 // Constants for UMA statistic collection.
 static const char kTranslateCaptureText[] = "Translate.CaptureText";
@@ skipping 162 matching lines @@
   if (main_frame->isWebRemoteFrame())
     return;
 
   GURL osdd_url = main_frame->document().openSearchDescriptionURL();
   if (!osdd_url.is_empty()) {
     Send(new ChromeViewHostMsg_PageHasOSDD(
         routing_id(), main_frame->document().url(), osdd_url,
         search_provider::AUTODETECTED_PROVIDER));
   }
 
-  // Don't capture pages including refresh meta tag.
-  if (HasRefreshMetaTag(main_frame))
-    return;
-
   CapturePageInfoLater(
       false,  // preliminary_capture
       base::TimeDelta::FromMilliseconds(
           render_view()->GetContentStateImmediately() ?
               0 : kDelayForCaptureMs));
 }
 
 void ChromeRenderViewObserver::DidCommitProvisionalLoad(
     WebLocalFrame* frame, bool is_new_navigation) {
   // Don't capture pages being not new, or including refresh meta tag.
-  if (!is_new_navigation || HasRefreshMetaTag(frame))
+  if (!is_new_navigation)
     return;
 
   base::debug::SetCrashKeyValue(
       crash_keys::kViewCount,
       base::SizeTToString(content::RenderView::GetRenderViewCount()));
-
-  CapturePageInfoLater(
-      true,  // preliminary_capture
-      base::TimeDelta::FromMilliseconds(kDelayForForcedCaptureMs));
 }
 
 void ChromeRenderViewObserver::CapturePageInfoLater(bool preliminary_capture,
                                                     base::TimeDelta delay) {
   capture_timer_.Start(
       FROM_HERE,
       delay,
       base::Bind(&ChromeRenderViewObserver::CapturePageInfo,
                  base::Unretained(this),
                  preliminary_capture));
 }
 
+void ChromeRenderViewObserver::DidFirstLayoutAfterFinishedParsing() {
+  CapturePageInfo(true);
+}
+
 void ChromeRenderViewObserver::CapturePageInfo(bool preliminary_capture) {
   if (!render_view()->GetWebView())
     return;
 
   WebFrame* main_frame = render_view()->GetWebView()->mainFrame();
   if (!main_frame)
     return;
 
   // TODO(creis): Refactor WebFrame::contentAsText to handle RemoteFrames,
   // likely by moving it to the browser process.  For now, only capture page
   // info from main frames that are LocalFrames, and ignore their RemoteFrame
   // children.
   if (main_frame->isWebRemoteFrame())
     return;
 
+  WebLocalFrame* local_frame = main_frame->toWebLocalFrame();
+  if (local_frame->isNavigationScheduled())
+    return;
+
   // Don't index/capture pages that are in view source mode.
   if (main_frame->isViewSourceModeEnabled())
     return;
 
   // Don't index/capture pages that failed to load.  This only checks the top
   // level frame so the thumbnail may contain a frame that failed to load.
   WebDataSource* ds = main_frame->dataSource();
   if (ds && ds->hasUnreachableURL())
     return;
 
@@ skipping 45 matching lines @@
 
   // When the contents are clipped to the maximum, we don't want to have a
   // partial word indexed at the end that might have been clipped. Therefore,
   // terminate the string at the last space to ensure no words are clipped.
   if (contents->size() == kMaxIndexChars) {
     size_t last_space_index = contents->find_last_of(base::kWhitespaceUTF16);
     if (last_space_index != base::string16::npos)
       contents->resize(last_space_index);
   }
 }
-
-bool ChromeRenderViewObserver::HasRefreshMetaTag(WebFrame* frame) {
-  if (!frame)
-    return false;
-  WebElement head = frame->document().head();
-  if (head.isNull() || !head.hasChildNodes())
-    return false;
-
-  const WebString tag_name(base::ASCIIToUTF16("meta"));
-  const WebString attribute_name(base::ASCIIToUTF16("http-equiv"));
-
-  WebNodeList children = head.childNodes();
-  for (size_t i = 0; i < children.length(); ++i) {
-    WebNode node = children.item(i);
-    if (!node.isElementNode())
-      continue;
-    WebElement element = node.to<WebElement>();
-    if (!element.hasHTMLTagName(tag_name))
-      continue;
-    WebString value = element.getAttribute(attribute_name);
-    if (value.isNull() ||
-        !base::LowerCaseEqualsASCII(base::StringPiece16(value), "refresh"))
-      continue;
-    return true;
-  }
-  return false;
-}