A quick fix for Issue 15971.

This is another my stupid mistake that my r19238 sends a Char event even when its text[] member does not contains any characters.
I mis-understoold I needed to send a Char event every time when gtk_keyval_to_unicode() returns 0 to avoid Issue 15024. But this is wrong.
(Page Down, Page Up, Arrow Keys work without Char events. Only return keys need Char events.)

BUG=15971 "drop down list moves 2 positions when using the keyboard".
TEST=Open the attached test-case
Review URL: http://codereview.chromium.org/155207

git-svn-id: svn://svn.chromium.org/chrome/trunk/src@20370 0039d316-1c4b-4281-b951-d872f2087c98

1 files changed, 4 insertions, 3 deletions

diff --git a/chrome/browser/renderer_host/render_widget_host_view_gtk.cc b/chrome/browser/renderer_host/render_widget_host_view_gtk.cc
index 4fb20ba..e8ce708 100644
--- a/chrome/browser/renderer_host/render_widget_host_view_gtk.cc
+++ b/chrome/browser/renderer_host/render_widget_host_view_gtk.cc
@@ -135,15 +135,16 @@ class RenderWidgetHostViewGtkWidget {
       // 2. The given key event is not a control-key event but printable
       //    characters aren't assigned to the event, (e.g. alt+d, etc.)
       // Create a Char event manually from this key event and send it to the
-      // renderer only when this event is a control-key event because
-      // control-key events should be processed by WebKit.
+      // renderer when this Char event contains a printable character which
+      // should be processed by WebKit.
       // TODO(hbono): Windows Chrome sends a Char event with its isSystemKey
       // value true for the above case 2. We should emulate this behavior?
       if (event->type == GDK_KEY_PRESS &&
           !gdk_keyval_to_unicode(event->keyval)) {
         NativeWebKeyboardEvent wke(event);
         wke.type = WebKit::WebInputEvent::Char;
-        host_view->GetRenderWidgetHost()->ForwardKeyboardEvent(wke);
+        if (wke.text[0])
+          host_view->GetRenderWidgetHost()->ForwardKeyboardEvent(wke);
       }
     }