Question

I have an app that plays an mp3 file and I’m trying to update a custom field in synchrony with certain times we have tabulated for the sound playback (kind of like a karaoke effect). I’m using a Handler to schedule these updates. In my custom field class, I define a Runnable that is supposed to run the update at the right time:

private final Runnable mTrigger = new Runnable() {
    @Override
    public void run() {
        int now = mPlayer.getCurrentPosition();
        if (mState == STATE_PLAYING && mUpdateAction != null) {
            if (mTriggerTime - now > MAX_PREMATURE_TRIGGER) {
                // Sound is lagging too much; reschedule this trigger
                mHandler.postDelayed(this, mTriggerTime - now);
            } else {
                // Run the update
                mUpdateAction.run();
            }
        }
    }
};

When I call mPlayer.start() I schedule the first update by calling mHandler.postDelayed(mTrigger, timeToFirstUpdate). Each update action decides what the next update will be and schedules it (by calling mHandler.postDelayed(mTrigger, timeToNextUpdate)). The updates times are typically a few hundred milliseconds apart.

The problem is that, while some updates are happening promptly at the scheduled times, others can be delayed by 200 milliseconds or more, which is quite noticeable to the user. I’m not doing anything in my app between these updates other than playing the sound. (No background worker threads; no other display updates.) The delays appear to be random and vary considerably each time through.

I didn’t think that the timing for postDelayed would be this imprecise! I don’t know if this is an emulator issue or a problem with my approach. Does sound playback screw up the timing of the UI thread loop? Should I move the timing into a background thread (and is it safe to call mPlayer.getCurrentPosition() from a background thread)? Something else?

Answer 1

After much experimenting, it seems like the problem is the emulator. When I ran everything on a speedier workstation, the problem seems to have gone away.