ReplayDecoder and LITTLE_ENDIANness

["Trustin Lee (이희승)"]

Hello,

Sorry for a late reply.

Could you post the source code of TracingFrameDecoder?  I can find only
the one derived from Replayingdecoder.

Trustin

UweH wrote:
> I have found a way to make it work. What I have is a PipelineFactory like
> this:
> 
> public class TracingPipelineFactory implements ChannelPipelineFactory {
> 
> ...
> 	@Override
> 	public ChannelPipeline getPipeline() throws Exception {
> 		final ChannelPipeline pipeline = pipeline();
> 
> 		pipeline.addLast("byteorder_decoder", new ByteOrderDecoder());
> 		pipeline.addLast("tracing_decoder", new TracingDecoder());
> 		//		pipeline.addLast("frame_decoder", new TracingFrameDecoder());
> 		pipeline.addLast("handler", new TracingHandler());
> 		return pipeline;
> 	}
> 
> }
> 
> The ByteOrderDecoder decodes only the first byte (which is
> endianess-independent), and switches the BufferFactory to the correct
> endianess, so that the next decoder (TracingDecoder) gets the buffer set to
> proper endianess already:
> 
> public class ByteOrderDecoder extends FrameDecoder {
> 
> 	@Override
> 	protected Object decode(final ChannelHandlerContext ctx, final Channel
> channel, final ChannelBuffer buffer)
> 			throws Exception {
> 
> 		if (buffer.readableBytes() < 1) {
> 			return null;
> 		}
> 
> 		final short order = buffer.getUnsignedByte(buffer.readerIndex());
> 		if (order != 1 && order != 2) {
> 			buffer.resetReaderIndex();
> 			throw new CorruptedFrameException("Invalid byte order: " + order);
> 		}
> 		final ByteOrder byteOrder = getByteOrder(order);
> 
> 		// switch byte order of buffer factory, so that next decoders get proper
> buffers
> 		if (channel.getConfig().getBufferFactory().getDefaultOrder() != byteOrder)
> {
> 			channel.getConfig().setBufferFactory(new
> HeapChannelBufferFactory(byteOrder));
> 		}
> 
> 		ctx.getPipeline().remove(this); // dont check byte order again, since it
> wont change
> 
> 		// hand off remaining data to signal decoder
> 		return buffer.readBytes(buffer.readableBytes());
> 	}
> 
> Next in line is the TracingDecoder, which does the actual decoding, creates
> a list of business objects and hands them to the TracingHandler.
> 
> public class TracingDecoder extends ReplayingDecoder<DecoderState> {
> 
> 	private final List<TracingSignal> tracingSignals = new
> LinkedList<TracingSignal>();
> 
> 	public TracingDecoder(final TracingManager manager) {
> 		super(DecoderState.READ_BYTEORDER, true); // initial state
> 
> 		this.manager = manager;
> 	}
> 
> 	@Override
> 	protected Object decode(final ChannelHandlerContext ctx, final Channel
> channel, final ChannelBuffer buffer,
> 			final DecoderState state) throws Exception {
> 
> 		System.out.println("converted buffer class: " + buffer.getClass());
> 		System.out.println("converted buffer byte order: " + buffer.order());
> 
> 		switch (state) {
> 			case READ_BYTEORDER:
> 				tracingSignals.clear();
> 				buffer.readUnsignedByte(); // read byte order
> 				......
> 				return Collections.unmodifiableList(tracingSignals);
> 			default:
> 				throw new ServiceException("Error decoding trace!");
> 		}
> 	}
> 
> 
> While this setup works, strange enough, if I replace the TracingDecoder with
> TracingFrameDecoder (which is derived from FrameDecoder), it stops working.
> 
> The FrameDecoder-version doesnt get a proper-endianess buffer, but a
> standard BIG_ENDIAN one.
> 
> Any idea why this is?
> 

-- 
what we call human nature in actuality is human habit
http://gleamynode.net/


-------------- next part --------------
A non-text attachment was scrubbed...
Name: signature.asc
Type: application/pgp-signature
Size: 260 bytes
Desc: OpenPGP digital signature
Url : http://lists.jboss.org/pipermail/netty-users/attachments/20100428/c898d22e/attachment-0001.bin