Bug in Mac OS X version: flipped bytes

Java Sound & OpenAL
#1

There’s a bug in the OS X version of JOAL that several people have posted about here and other forums where projects (like Xith) use JOAL. As I understand it, OS X uses big-endian words, and so in 16-bit wav files (ogg seems to be affected by this too) JOAL reads the bytes in reverse order, resulting in static sound (8-bit files, since there’s nothing to flip, are read correctly). This seems like it could be easily fixed if JOAL simply checks if the OS is OS X and reads the files in big-endian format. I’ve checked out the CVS source and the bug is still in there.

#2

I patched WAVLoader.java with this:


        ByteBuffer buffer = ByteBuffer.allocateDirect(size);
        aChannel.read(buffer);
      
      // On Mac we need to convert this to big endian
      if (bits == 16 && ByteOrder.nativeOrder() == ByteOrder.BIG_ENDIAN)
      {
            buffer.rewind();
            ShortBuffer tmp = buffer.duplicate().order(ByteOrder.LITTLE_ENDIAN).asShortBuffer();
            while(tmp.hasRemaining())
                  buffer.putShort(tmp.get());
      }

        result = new WAVData(buffer, format, size, freq, false);
        aIn.close();
#3

[quote]I patched WAVLoader.java with this:


        ByteBuffer buffer = ByteBuffer.allocateDirect(size);
        aChannel.read(buffer);
      
      // On Mac we need to convert this to big endian
      if (bits == 16 && ByteOrder.nativeOrder() == ByteOrder.BIG_ENDIAN)
      {
            buffer.rewind();
            ShortBuffer tmp = buffer.duplicate().order(ByteOrder.LITTLE_ENDIAN).asShortBuffer();
            while(tmp.hasRemaining())
                  buffer.putShort(tmp.get());
      }

        result = new WAVData(buffer, format, size, freq, false);
        aIn.close();

[/quote]
Awesome! That works perfectly :slight_smile:

#4

Java programmers will need to be aware that JOAL on Mac expects 16-bit samples in the native (Big Endian) byte order. Loaders for OGG will need similar tweaks.

If you are making samples algorithmically with 16-bit samples be sure to set the ByteBuffer to the native byte order before you stuff it with ‘shorts’.

#5

Yes, it’s perfect! thanks swpalmer!!

But, why is this not actually in a new version??
I cannot understand why this issue still isn’t fixed in the new versions, given that this post dates back to november 2004!! (almost a year ago).

Actually, I will put a new post regarding this… for anyone that has that problem.
Thanks again

#6

Just a heads up that if you do any byte flipping here, because if you tested for MacOS X, make sure you also check for the architecture. I say this since this will break otherwise on the new x86 machines due out in 2006. Truth is, this should be fixed in the native library.

#7

If you take a closer look at my fix… I used “ByteOrder.nativeOrder() == ByteOrder.BIG_ENDIAN”, I chose that on-purpose in the hope that it would work for unforseen platforms (e.g. Linux PPC). So hopefully OS X on Intel will “just work”.

#8

swpalmer: thanks for the patch. I’ll apply something similar to the source base soon.

Regarding the lack of updates on JOAL, the original maintainer left Sun a while back and the project has been somewhat orphaned since then. I’ll post an update in another thread.