[icedtea-web] RFC: Patch to double-buffer applets

[Deepak Bhole]

* Omair Majid <omajid at redhat.com> [2010-11-03 09:32]:
> Hi Deepak,
> 
> On 11/02/2010 05:37 PM, Deepak Bhole wrote:
> >* Deepak Bhole<dbhole at redhat.com>  [2010-11-02 17:31]:
> >>Hi,
> >>
> >>IcedTea plugin currently does not double-buffer the applet panel. As a
> >>result, some applets may flicker depending on the machine.
> >>
> >>Attached patch buffers the frame offscreen and then draws it, thereby
> >>addressing the issue.
> >>
> 
> Is there a specific test case that this fixes? The only flickering
> applet that I know of is at
> http://download.oracle.com/javase/1.4.2/docs/guide/misc/applet.html,
> and I dont see any change in that applet with the patch applied.
> 
> >>ChangeLog:
> >>2010-11-02  Deepak Bhole<dbhole at redhat.com>
> >>
> >>    * plugin/icedteanp/java/sun/applet/PluginAppletViewer.java (update):
> >>    Override method and implement double-buffering.
> >>
> >>Cheers,
> >>Deepak
> >
> >
> >Doh! There shouldn't be a println in there. New patch attached. I also
> >fixed the variable names for the image and its graphics.
> >
> >
> >>diff -r 6c2527d42900 ChangeLog
> >>--- a/ChangeLog	Mon Nov 01 11:44:15 2010 -0400
> >>+++ b/ChangeLog	Tue Nov 02 17:29:03 2010 -0700
> >>@@ -1,3 +1,8 @@
> >>+2010-11-02  Deepak Bhole<dbhole at redhat.com>
> >>+
> >>+	* plugin/icedteanp/java/sun/applet/PluginAppletViewer.java (update):
> >>+	Override method and implement double-buffering.
> >>+
> >>  2010-11-01  Omair Majid<omajid at redhat.com>
> >>
> >>  	* Makefile.am (clean-IcedTeaPlugin): Only delete launcher directory if it
> >>diff -r 6c2527d42900 NEWS
> >>--- a/NEWS	Mon Nov 01 11:44:15 2010 -0400
> >>+++ b/NEWS	Tue Nov 02 17:29:03 2010 -0700
> >>@@ -12,3 +12,4 @@
> >>
> >>  * Initial release of IcedTea-Web
> >>  * PR542: Plugin fails with NPE on http://www.openprocessing.org/visuals/iframe.php?visualID=2615
> >>+* Applets are now double-buffered to eliminate flicker in ones that do heavy drawing
> >>diff -r 6c2527d42900 plugin/icedteanp/java/sun/applet/PluginAppletViewer.java
> >>--- a/plugin/icedteanp/java/sun/applet/PluginAppletViewer.java	Mon Nov 01 11:44:15 2010 -0400
> >>+++ b/plugin/icedteanp/java/sun/applet/PluginAppletViewer.java	Tue Nov 02 17:29:03 2010 -0700
> >>@@ -359,6 +366,9 @@
> >>
> >>       private static Long requestIdentityCounter = 0L;
> >>
> >>+     private Image bufferedPanelImage;
> >>+     private Graphics bufferedImageGraphics;
> >>+
> >>       /**
> >>        * Null constructor to allow instantiation via newInstance()
> >>        */
> >>@@ -2163,4 +2178,24 @@
> >>          }
> >>      }
> >>       }
> >>+
> >>+     /**
> >>+      * {@inheritDoc}
> >>+      */
> >>+     public void update(Graphics g) {
> >>+
> >>+    	 System.err.println("Buffering and drawing");
> >>+    	
> >>+    	 // If the image or the graphics don't exist, create new ones
> >>+    	 if (bufferedPanelImage == null || bufferedImageGraphics == null) {
> >>+    		 bufferedPanelImage = createImage(getWidth(), getHeight());
> >>+    		 bufferedImageGraphics = bufferedPanelImage.getGraphics ();
> >>+    	 }
> >>+
> >>+    	 // Paint off-screen
> >>+    	 paint(bufferedImageGraphics);
> 
> The overridden method is update, but paint is called here instead.
> Is there a reason for that? As far as I know, the semantics of
> update() are slightly different from paint() - update() clears the
> background and then calls paint(). Perhaps super.update() might be
> more appropriate?
>

super.update() would just call paint on the Graphics instance passed to
it. Container's (super in this case) update() methods do a couple of
additional checks, but we don't need them for applets, which are known
lightweight containers.

I can definitely add the call if you feel it would be cleaner though --
just that it doesn't make a difference in this case.

Cheers,
Deepak
 
> >>+
> >>+    	 // Draw the painted image
> >>+    	 g.drawImage(bufferedPanelImage, 0, 0, this);
> >>+     }
> >>   }
> >
> 
> Thanks,
> Omair