Some further poking on the Android Developers page lead me to the Lunar Lander example which looked right what I wanted to do. I installed it on my HTC and played around with it a bit when I suddenly received this error message:

What happened? To put it simple, there is a bug in Google’s Lunar Lander example. Let’s look at the implemenation of the thread that draws the game in LunarView.java and how it’s started:

1
2
3
4

public void surfaceCreated(SurfaceHolder holder) {
	thread.setRunning(true);
	thread.start();
}

The method surfaceCreated() is a callback method from the SurfaceHolder class. It is called whenever the Surface we want to draw on is created, for example when the application starts or when we return from a pause. Respectivly the method surfaceDestroyed() is called whenever the surface is destroyed, for example when the application terminates or another application gets the focus.

1
2
3
4
5
6
7
8
9
10
11

public void surfaceDestroyed(SurfaceHolder holder) {
	boolean retry = true;
	thread.setRunning(false);
	while (retry) {
		try {
			thread.join();
			retry = false;
		} catch (InterruptedException e) {
		}
	}
}

In the Lunar Lander implemention the drawing thread is terminated on the first time the surface is destroyed. If we then try to come back to the application, for example with a long click on the home button, the application throws this exception:

java.lang.IllegalThreadStateException: Thread already started.

We reached an illegal state because a thread is only allowed to be started a single time!

To fix this issue I don’t stop the drawing thread in my application if the app looses focus. Instead I just stop drawing and let the thread sleep for a while (and thus save the CPU resources etc). When the app gets focus again, the drawing starts over. In the example below this behaviour is controlled by the paused variable, running is the variable that is set to true when the game starts and is set to false when the app is terminated and the thread really needs to stop:

1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21

@Override
public void run() {
	while (running) {
		while (paused) {
			try {
				Thread.sleep(50L);
			} catch (InterruptedException ignore) {
			}
		}
 
		Canvas canvas = null;
		try {
			canvas = holder.lockCanvas(null);
			draw(canvas);
		} finally {
			if (canvas != null) {
				holder.unlockCanvasAndPost(canvas);
			}
		}
	}
}

I’ve implemented this gameloop in a small example that displays Conway’s Game of Life:

Of course my implementation has a fixed timestep as suggested by the famous article Fix Your Timestep!

Download my example sourcecode here: android-gameloop-example-v1.zip
App in the Android Market: Gameloop Tutorial

1
2

System.setOut(PrintStream stream);
System.setErr(PrintStream stream);

So we simply have to add our own PrintStream implementation that calls the logger and we are done:

1
2
3
4
5

new PrintStream(realPrintStream) {
    public void print(final String message) {
        LOG.info(message);
    }
}

LOG would be our log4j Logger.

It suggests not just to put the classes name there but also use a @link tag so one can navigate the Javadoc sources later on. The finished comment:

Unfortunately this functionality is not available in JetBrains’ IntelliJ Idea. There is (of course) auto completion for classnames in Idea but it simply doesn’t put them in a @link tag. A simple trick saves the day: simply create a live template with context ‘Java comment’ like here:

EDIT: The screenshot is missing the word ‘link’. The complete template should be: ‘{@link $LINK$} ‘.

Don’t forget to click on ‘Edit variables’ and fill in ‘complete()’ as Expression. This activates the auto completion for classnames.

Now everytime a @link in the Javadocs is needed we invoke the Live Templates with Alt + J and enter a classname. The @link tag will be build around that name. It’s not as simple as with Eclipse, but works pretty well.

static void glVertexPointer(int size, int type, int stride, long pointer_buffer_offset);

With the last offset one can specify where the vertex information starts in the buffer given to the VBO management. Well, I could/should have guessed that this offset has to be in bytes since I am dealing with a buffer here but instead I used the number of floats.. So next time I see an operation dealing with buffers that takes a long argument I’ll try the byte-count from the beginning..

Für die volle Pracht auf den HD Button drücken!

avrdude: stk500_getsync(): not in sync: resp=0×00
avrdude: stk500_disable(): protocol error, expect=0×14, resp=0×51

The error manifest while uploading a new sketch to Meggy. If you pull out the USB cable and plug it in again you can upload a sketch successful. But on the next time you try you’ll get the error again. Luckily there is a pretty simple solution. I found it while digging through the Evil Mad Scientist Forums. It involves setting the “Set RTS on close” flag of the COM Port. You can do this in the Device Manager. Find the COM Port that was installed when you first plugged in the USB-to-Serial cable. For me it’s COM3. In the connection settings click on details and mark the flag selected. That’s it. I tried this on Windows XP with Service Pack 3, but it should work similiarily on other Windows versions.

RTS Flag