I did an hour long presentation and got this today:

E/AndroidRuntime( 2269): java.lang.OutOfMemoryError: bitmap size exceeds VM budget
E/AndroidRuntime( 2269):        at android.graphics.BitmapFactory.nativeDecodeByteArray(Native Method)
E/AndroidRuntime( 2269):        at android.graphics.BitmapFactory.decodeByteArray(BitmapFactory.java:405)
E/AndroidRuntime( 2269):        at android.graphics.BitmapFactory.decodeByteArray(BitmapFactory.java:418)
E/AndroidRuntime( 2269):        at org.libreoffice.impressremote.communication.SlideShow.getImage(SlideShow.java:57)
E/AndroidRuntime( 2269):        at org.libreoffice.impressremote.PresentationFragment$ThumbnailAdapter.createBitmap(PresentationFragment.java:309)
E/AndroidRuntime( 2269):        at pl.polidea.coverflow.AbstractCoverFlowImageAdapter.getItem(AbstractCoverFlowImageAdapter.java:73)


I strongly suspect that this is because:

    public Bitmap getImage(int aSlide) {
        byte[] aImage = mPreviewImages.get(aSlide);
        if (aImage == null) {
            return BitmapFactory.decodeResource(mContext.getResources(),
                            R.drawable.image_loading);
        }
        Bitmap aBitmap = null;
        try {
            aBitmap = BitmapFactory.decodeByteArray(aImage, 0, aImage.length);
        } catch (OutOfMemoryError e) {
            Log.e(Globals.TAG, "Bitmap decoding error byte length: " + aImage.length +
                  "first 4 bytes: " + aImage[0] + " " + aImage[1] + " " + aImage[2] + " " + aImage[3] +
                  "Exception " + e);
        }
        if (aBitmap == null) {
            return BitmapFactory.decodeResource(mContext.getResources(),
                            R.drawable.image_loading);
        }
        return aBitmap;
    }

is called by the coverflow, and constantly re-creates the bitmap from the stored base64 string.

It seems to me that this is really silly :-) we should store the Bitmap and return the same Bitmap pointer for each slide as a Bitmap - which has to be 50% of the size of a base64 encoded image. It would also be the same image each time so as to not kill the GC etc. - and caching that should improve perf.

We should throw away the base64 text too as soon as we successfully have an image in hand :-)