VideoLayerImpl: subtract 1 from visible size when calculating tex stretch.

cc/layers/video_layer_impl.cc

Index: cc/layers/video_layer_impl.cc
diff --git a/cc/layers/video_layer_impl.cc b/cc/layers/video_layer_impl.cc
index d173cc6c2b6d0f511a802fb95d167aed98f9a1d1..8f3d70af76000b1538d521a4f80e393b1ce64e42 100644
--- a/cc/layers/video_layer_impl.cc
+++ b/cc/layers/video_layer_impl.cc
@@ -186,9 +186,9 @@ void VideoLayerImpl::AppendQuads(RenderPass* render_pass,
 
   // Pixels for macroblocked formats.
   const float tex_width_scale =
-      static_cast<float>(visible_rect.width()) / coded_size.width();
+      static_cast<float>(visible_rect.width() - 1) / coded_size.width();

[Pawel Osciak, 2016/02/01 00:40:58]

I'm wondering if we should be checking for 0,0 (and perhaps 1,1) visible rect
before subtracting?

   const float tex_height_scale =
-      static_cast<float>(visible_rect.height()) / coded_size.height();
+      static_cast<float>(visible_rect.height() - 1) / coded_size.height();
 
   switch (frame_resource_type_) {
     // TODO(danakj): Remove this, hide it in the hardware path.