Preventing touch events from registering when they occur outside the desktop image

remoting/android/java/src/org/chromium/chromoting/TouchInputHandler.java

Index: remoting/android/java/src/org/chromium/chromoting/TouchInputHandler.java
diff --git a/remoting/android/java/src/org/chromium/chromoting/TouchInputHandler.java b/remoting/android/java/src/org/chromium/chromoting/TouchInputHandler.java
index 3f518a6f0d8d75bfe3d4fb541530d93a5dde9bed..4cbd35ed60803a8945067e13135d52748f44e497 100644
--- a/remoting/android/java/src/org/chromium/chromoting/TouchInputHandler.java
+++ b/remoting/android/java/src/org/chromium/chromoting/TouchInputHandler.java
@@ -570,6 +570,9 @@ public class TouchInputHandler {
             }
 
             if (!mInputStrategy.isIndirectInputMode()) {
+                if (!mapScreenPointToImage(x, y)) {
+                    return false;
+                }
                 moveCursorToScreenPoint(x, y);
             }
 
@@ -592,6 +595,9 @@ public class TouchInputHandler {
             }
 
             if (!mInputStrategy.isIndirectInputMode()) {
+                if (!mapScreenPointToImage(x, y)) {
+                    return;
+                }
                 moveCursorToScreenPoint(x, y);
             }
 
@@ -619,5 +625,22 @@ public class TouchInputHandler {
                     return InputStub.BUTTON_UNDEFINED;
             }
         }
+
+        /** Verifies the given point maps to a valid location within the desktop image. */
+        private boolean mapScreenPointToImage(float screenX, float screenY) {

[Lambros, 2016/07/12 23:19:50]

Looks like we have this logic a few times in this file already. Maybe have this
function take a mutable PointF object to accept the transformed coords? Doesn't
have to block this CL from landing.

[joedow, 2016/07/13 00:04:50]

Ack.  I was thinking this kind of functionality may make sense to move in
RenderData (if it continues to exist with the refactoring).  Leaving for this CL
but will consider it in a future CL.

+            float[] mappedPoints = {screenX, screenY};
+            int imageWidth;
+            int imageHeight;
+            Matrix screenToImage = new Matrix();
+            synchronized (mRenderData) {
+                mRenderData.transform.invert(screenToImage);
+                imageWidth = mRenderData.imageWidth;
+                imageHeight = mRenderData.imageHeight;
+            }
+            screenToImage.mapPoints(mappedPoints);
+
+            return (mappedPoints[0] >= 0 && mappedPoints[0] <= imageWidth)

[Lambros, 2016/07/12 23:19:50]

Is it worth adding a little bit of slop margin here? If the user taps at the
extreme edge, it might be a bit frustrating for the tap to sometimes go through
and sometimes not?

[joedow, 2016/07/13 00:04:50]

I considered that, but this method is only used when the display is sufficiently
shrunk that the desktop image has a border.  If they are having difficulty
tapping an edge target (or any target really), I'd think it would be more
reliable for them to zoom in a bit.

+                    && (mappedPoints[1] >= 0 && mappedPoints[1] <= imageHeight);
+        }
     }
 }