Fixed the inconsistent behavior when long pressing on editable area in Chrome.

content/renderer/render_widget.cc

Index: content/renderer/render_widget.cc
diff --git a/content/renderer/render_widget.cc b/content/renderer/render_widget.cc
index 10d87e16fe98629050057dce80ae3b6bc182a69d..96718bd79072eb735d1eb95b26a9692c02a945c2 100644
--- a/content/renderer/render_widget.cc
+++ b/content/renderer/render_widget.cc
@@ -1967,9 +1967,14 @@ void RenderWidget::didHandleGestureEvent(
 #if defined(OS_ANDROID) || defined(USE_AURA)
   if (event_cancelled)
     return;
-  if (event.type == WebInputEvent::GestureTap ||
-      event.type == WebInputEvent::GestureLongPress) {
+  if (event.type == WebInputEvent::GestureTap) {
     UpdateTextInputState(SHOW_IME_IF_NEEDED, FROM_NON_IME);
+  } else if (event.type == WebInputEvent::GestureLongPress) {
+    blink::WebTextInputInfo new_info;
+    if (webwidget_)
+      new_info = webwidget_->textInputInfo();
+    if (!new_info.value.isEmpty())
+      UpdateTextInputState(SHOW_IME_IF_NEEDED, FROM_NON_IME);

[aurimas (slooooooooow), 2014/08/18 16:05:24]

Conditionally calling UpdateTextInputState will cause all sorts of issues as
browser side IME state will not be initialized (AdapterInputConnection.java).
There are many callsites of UpdateTextInputState and the next one to call it
will push the update to Java defeating your conditional call.

[AKVT, 2014/08/18 16:14:30]

Thanks. In that case, how do we overcome this problem ? We need to handle from
Webkit side itself ? or let the call goes to AdapterInputConnection and prevent
IME showing from there ?
Please suggest your opinion

   }
 #endif
 }