4th attempt at landing this.

It caused several tests to fails previously.
Cannot repro any of the failure. I am suspecting a clobber would have probably fixed it. (there is a change in one of the IPC messages that probably messed up the build somehow).
No code change.

Enabling language detection on page load.
A memory error has been fixed in the CLD library in the meantime.
This should hopefully fixes the crashers in the reliability tests.
Note that this version is actually simpler than the original review since the detection is now performed in the renderer. (So the CLD code runs sandboxed.)

Original review:
http://codereview.chromium.org/492024/show

BUG=30662
TEST=Run the unit-tests. 
TBR=brettw
Review URL: http://codereview.chromium.org/548057

git-svn-id: svn://svn.chromium.org/chrome/trunk/src@36442 0039d316-1c4b-4281-b951-d872f2087c98

1 files changed, 21 insertions, 32 deletions

diff --git a/chrome/renderer/render_view.cc b/chrome/renderer/render_view.cc
index 649542f..a43c011 100644
--- a/chrome/renderer/render_view.cc
+++ b/chrome/renderer/render_view.cc
@@ -304,7 +304,6 @@ RenderView::RenderView(RenderThreadBase* render_thread,
       send_preferred_size_changes_(false),
       ALLOW_THIS_IN_INITIALIZER_LIST(
           notification_provider_(new NotificationProvider(this))),
-      determine_page_text_after_loading_stops_(false),
       view_type_(ViewType::INVALID),
       browser_window_id_(-1),
       last_top_level_navigation_page_id_(-1),
@@ -495,7 +494,6 @@ void RenderView::OnMessageReceived(const IPC::Message& message) {
     IPC_MESSAGE_HANDLER(ViewMsg_CopyImageAt, OnCopyImageAt)
     IPC_MESSAGE_HANDLER(ViewMsg_ExecuteEditCommand, OnExecuteEditCommand)
     IPC_MESSAGE_HANDLER(ViewMsg_Find, OnFind)
-    IPC_MESSAGE_HANDLER(ViewMsg_DeterminePageLanguage, OnDeterminePageLanguage)
     IPC_MESSAGE_HANDLER(ViewMsg_Zoom, OnZoom)
     IPC_MESSAGE_HANDLER(ViewMsg_SetZoomLevelForLoadingHost,
                         OnSetZoomLevelForLoadingHost)
@@ -666,24 +664,35 @@ void RenderView::CapturePageInfo(int load_id, bool preliminary_capture) {
   if (!preliminary_capture)
     last_indexed_page_id_ = load_id;
 
-  // get the URL for this page
+  // Get the URL for this page.
   GURL url(main_frame->url());
   if (url.is_empty())
     return;
 
-  // full text
+  // Retrieve the frame's full text.
   std::wstring contents;
   CaptureText(main_frame, &contents);
   if (contents.size()) {
-    // Send the text to the browser for indexing.
-    Send(new ViewHostMsg_PageContents(url, load_id, contents));
-  }
+    std::string language = "unknown";
+#if defined(OS_WIN)
+    int num_languages = 0;
+    bool is_reliable = false;
+    base::TimeTicks begin_time = base::TimeTicks::Now();
+    Language cld_lang =
+        DetectLanguageOfUnicodeText(NULL, contents.c_str(), true,
+                                    &is_reliable, &num_languages, NULL);
+    if (cld_lang != NUM_LANGUAGES && cld_lang != UNKNOWN_LANGUAGE &&
+        cld_lang != TG_UNKNOWN_LANGUAGE) {
+      language = LanguageCodeISO639_1(cld_lang);
+    }
+    UMA_HISTOGRAM_MEDIUM_TIMES("Renderer4.LanguageDetection",
+                               base::TimeTicks::Now() - begin_time);
+#endif
 
-  // Now that we have the contents, we can determine the language if necessary.
-  if (determine_page_text_after_loading_stops_) {
-    determine_page_text_after_loading_stops_ = false;
-    Send(new ViewHostMsg_PageLanguageDetermined(
-        routing_id_, DetermineTextLanguage(contents)));
+    // Send the text to the browser for indexing (the browser might decide not
+    // to index, if the URL is HTTPS for instance) and language discovery.
+    Send(new ViewHostMsg_PageContents(routing_id_, url, load_id, contents,
+                                      language));
   }
 
   // thumbnail
@@ -695,15 +704,6 @@ void RenderView::CaptureText(WebFrame* frame, std::wstring* contents) {
   if (!frame)
     return;
 
-  // Don't index any https pages. People generally don't want their bank
-  // accounts, etc. indexed on their computer, especially since some of these
-  // things are not marked cachable.
-  // TODO(brettw) we may want to consider more elaborate heuristics such as
-  // the cachability of the page. We may also want to consider subframes (this
-  // test will still index subframes if the subframe is SSL).
-  if (GURL(frame->url()).SchemeIsSecure())
-    return;
-
 #ifdef TIME_TEXT_RETRIEVAL
   double begin = time_util::GetHighResolutionTimeNow();
 #endif
@@ -3067,17 +3067,6 @@ void RenderView::OnFind(int request_id, const string16& search_text,
   }
 }
 
-void RenderView::OnDeterminePageLanguage() {
-  if (is_loading_) {
-    // Wait for the page to finish loading before trying to determine the
-    // language.
-    determine_page_text_after_loading_stops_ = true;
-    return;
-  }
-
-  Send(new ViewHostMsg_PageLanguageDetermined(routing_id_, DetectLanguage()));
-}
-
 std::string RenderView::DetectLanguage() {
   if (!webview() || is_loading_)
     return kUnknownLanguageCode;