请问 websockt 如何封装 html 是主流，有相关demo吗。
js websocket 客户端通信demo吗，因为我服务器用 jkcp 但是客户端用的html

下面是日志。收到客户端准备完成的消息，就给两个客户端都发送战斗开始。
但是会出现有消息取不到数据，提示 消息已经被free 的情况。
下面 add send bb 打印是在 KcpOnUdp 的send方法

public synchronized void send(ByteBuf bb)
{
if (!closed)
{
this.sendList.add(bb);
System.err.println("add send bb:"+ Thread.currentThread().getName()+" hash:"+bb.hashCode()+" bb:"+bb+" ref:"+bb.refCnt());
this.needUpdate = true;
}
}

send fail 是在 Kcp 的send 方法

public int send(ByteBuf buffer) {
System.err.println(" sending buffer :" +Thread.currentThread().getName()+" hash:"+ buffer.hashCode() + " buffer:" + buffer+" ref:"+buffer.refCnt());
if (buffer.readableBytes() == 0) {
System.err.println(" send fail, buffer :" + Thread.currentThread().getName()+" hash:" + buffer.hashCode() + " buffer:" + buffer+ " ref:"+buffer.refCnt());
return -1;
}

下面是出错日志：

2018-06-27 14:45:38][kcp thread 0] INFO - at com.test.fightserver.room.Room.clientReady(Room.java:60) -[ kcp ready:local: /0.0.0.0:2222 remote: /192.168.1.105:60305]
2018-06-27 14:45:38][kcp thread 0] INFO - at com.test.fightserver.room.Room.clientReady(Room.java:74) -[ all ready ]
2018-06-27 14:45:38][kcp thread 0] INFO - at com.test.fightserver.room.Room.clientReady(Room.java:90) -[ send player enter to local: /0.0.0.0:2222 remote: /192.168.1.103:60354 msg: UnpooledByteBufAllocator$InstrumentedUnpooledUnsafeHeapByteBuf(ridx: 0, widx: 30, cap: 30) readableBytes :30]
2018-06-27 14:45:38][kcp thread 0] INFO - at com.test.fightserver.room.Room.clientReady(Room.java:90) -[ send player enter to local: /0.0.0.0:2222 remote: /192.168.1.105:60305 msg: UnpooledByteBufAllocator$InstrumentedUnpooledUnsafeHeapByteBuf(ridx: 0, widx: 30, cap: 30) readableBytes :30]
2018-06-27 14:45:38][kcp thread 0] INFO - at com.test.fightserver.room.Room.clientReady(Room.java:118) -[ send battle start to local: /0.0.0.0:2222 remote: /192.168.1.103:60354 msg: UnpooledByteBufAllocator$InstrumentedUnpooledUnsafeHeapByteBuf(ridx: 0, widx: 12, cap: 12) readableBytes :12]
2018-06-27 14:45:38][kcp thread 0] INFO - at com.test.fightserver.room.Room.clientReady(Room.java:118) -[ send battle start to local: /0.0.0.0:2222 remote: /192.168.1.105:60305 msg: UnpooledByteBufAllocator$InstrumentedUnpooledUnsafeHeapByteBuf(ridx: 0, widx: 12, cap: 12) readableBytes :12]
add send bb:kcp thread 0 hash:1304880682 bb:UnpooledByteBufAllocator$InstrumentedUnpooledUnsafeHeapByteBuf(ridx: 0, widx: 12, cap: 12) ref:1
sending buffer :kcp thread 0 hash:1128752854 buffer:UnpooledByteBufAllocator$InstrumentedUnpooledUnsafeHeapByteBuf(ridx: 0, widx: 30, cap: 30) ref:1
sending buffer :kcp thread 0 hash:1304880682 buffer:UnpooledByteBufAllocator$InstrumentedUnpooledUnsafeHeapByteBuf(ridx: 0, widx: 12, cap: 12) ref:1
sending buffer :kcp thread 0 hash:1128752854 buffer:UnpooledByteBufAllocator$InstrumentedUnpooledUnsafeHeapByteBuf(ridx: 0, widx: 30, cap: 30) ref:1
sending buffer :kcp thread 0 hash:1304880682 buffer:UnpooledByteBufAllocator$InstrumentedUnpooledUnsafeHeapByteBuf(ridx: 0, widx: 12, cap: 12) ref:1
add send bb:pool-3-thread-1 hash:-168963106 bb:UnpooledByteBufAllocator$InstrumentedUnpooledUnsafeHeapByteBuf(ridx: 0, widx: 16, cap: 16) ref:1
add send bb:pool-3-thread-1 hash:-168963106 bb:UnpooledByteBufAllocator$InstrumentedUnpooledUnsafeHeapByteBuf(ridx: 0, widx: 16, cap: 16) ref:1
2018-06-27 14:45:38][kcp thread 0] INFO - at com.test.fightserver.server.UdpServer.handleReceive(UdpServer.java:48) -[ client msg:3012 kcp:local: /0.0.0.0:2222 remote: /192.168.1.103:60354]
sending buffer :kcp thread 0 hash:-168963106 buffer:UnpooledByteBufAllocator$InstrumentedUnpooledUnsafeHeapByteBuf(ridx: 0, widx: 16, cap: 16) ref:1
sending buffer :kcp thread 0 hash:1 buffer:UnpooledByteBufAllocator$InstrumentedUnpooledUnsafeHeapByteBuf(freed) ref:0
send fail, buffer :kcp thread 0 hash:1 buffer:UnpooledByteBufAllocator$InstrumentedUnpooledUnsafeHeapByteBuf(freed) ref:0
2018-06-27 14:45:38][kcp thread 0] ERROR - at com.test.fightserver.server.UdpServer.handleException(UdpServer.java:62) -[ ]
java.lang.IllegalStateException: send error : -1
at org.beykery.jkcp.KcpOnUdp.update(KcpOnUdp.java:198) [classes/:?]
at org.beykery.jkcp.KcpThread.run(KcpThread.java:154) [classes/:?]
2018-06-27 14:45:38][kcp thread 0] INFO - at com.test.fightserver.server.UdpServer.handleClose(UdpServer.java:68) -[ client close................local: /0.0.0.0:2222 remote: /192.168.1.105:60305]

我进行了部分测试，连接上大量客户端后关闭客户端程序，在服务器端的KcpOnUdp对象没有从map中移除，经过查代码发现有这样的问题:
如果一个客户端上来发送的第一个包不完整后面不发送数据则lastTime为0

   if (this.timeout > 0 && lastTime > 0 && System.currentTimeMillis() - this.lastTime > this.timeout)
    {
      this.closed = true;
      this.release();
      this.listerner.handleClose(this);
    }

这里的close状态永远不会被设置为true，这样导致KcpOnUdp永远存在内存里面了

lastTime 是否应该考虑放在接受到消息就设置时间，而不是收到完整包再设置时间？

当发送的源数据在几百k的时候，分成n个1400的包发送，接收端的channelRead中接收一百个1400的包之后就直接走到readComplete了

KCP.java中update方法第一行将Long强转为int类型
this.current = (int) current;
外部传入的是毫秒时间戳，这里强转为int类型，导致current变为负数，进而导致后续的rto计算全部失效。
看了c版本的kcp代码都是long类型的，这改为int是如何考虑的？上述问题应该如何解决？

大佬，如题！

TestServer 和 TestClient 的 handleRecieve 都改为类似
@OverRide
public void handleReceive(ByteBuf bb, KcpOnUdp kcp) {
String content = bb.toString(Charset.forName("utf-8"));
System.out.println("KCP "+ content +System.currentTimeMillis());
kcp.send(bb);// echo
}
互相回发，Client的main函数中 sleep 1s 后调用close() 方法，测试下来Server端可以打印100行左右。
而用Netty直接写UDP测试1s可以打印出近2000行消息，TCP 1s 也可以打印近1800行消息。

造成这种差异的原因，是KCP算法本身调度，和增加了流量大小所致，还是别的原因？

测试DUP 用例：
public class Server {
private static Logger logger = LoggerFactory.getLogger(Server.class);

/**
 * udp服务端，接受客户端发送的广播
 */
public static void initServer() {
	EventLoopGroup group = new NioEventLoopGroup();
	try {
		Bootstrap bootstrap = new Bootstrap();
		bootstrap.group(group).channel(NioDatagramChannel.class).option(ChannelOption.SO_BROADCAST, true)
				.handler(new UdpServerHandler());
		Channel channel = bootstrap.bind(AppConstants.SEARCH_PORT).sync().channel();
		channel.closeFuture().await();
	} catch (InterruptedException e) {
		e.printStackTrace();
	} finally {
		group.shutdownGracefully();
	}
}

private static class UdpServerHandler extends SimpleChannelInboundHandler<DatagramPacket> {
	@Override
	protected void channelRead0(ChannelHandlerContext ctx, DatagramPacket msg) throws Exception {
		System.out.println("UDP "+ req +System.currentTimeMillis());
		ctx.writeAndFlush(new DatagramPacket(Unpooled.copiedBuffer(req, CharsetUtil.UTF_8), msg.sender()));
	}
}

public static void main(String args[]) {
	Server.initServer();
}

}

public class Client {
private static Logger logger = LoggerFactory.getLogger(Client.class);
private int scanPort;

public Client(int scanPort) {
	this.scanPort = scanPort;
}

private static class ClientHandler extends SimpleChannelInboundHandler<DatagramPacket> {

	@Override
	protected void channelRead0(ChannelHandlerContext ctx, DatagramPacket msg) throws Exception {
		String response = msg.content().toString(CharsetUtil.UTF_8);
		System.out.println("UDP "+ response +System.currentTimeMillis());
		ctx.writeAndFlush(new DatagramPacket(Unpooled.copiedBuffer(response, CharsetUtil.UTF_8), msg.sender()));
	}
}

public void sendPackage() {
	EventLoopGroup group = new NioEventLoopGroup();
	try {
		Bootstrap b = new Bootstrap();
		b.group(group).channel(NioDatagramChannel.class).option(ChannelOption.SO_BROADCAST, true)
				.handler(new ClientHandler());

		Channel ch = b.bind(0).sync().channel();

		ch.writeAndFlush(new DatagramPacket(Unpooled.copiedBuffer("hello!!!", CharsetUtil.UTF_8),
				new InetSocketAddress("255.255.255.255", scanPort))).sync();
	} catch (Exception e) {
		e.printStackTrace();
	} finally {
		group.shutdownGracefully();
	}
}

public static void main(String args[]) {
	Client client = new Client(AppConstants.SEARCH_PORT);
	client.sendPackage();
}

}

public class AppConstants {

public static final int SEARCH_PORT = 22023;

}

[nioEventLoopGroup-2-1] WARN io.netty.util.ReferenceCountUtil - Failed to release a message: DatagramPacket(/0.0.0.0:6000 => /127.0.0.1:64272, PooledUnsafeDirectByteBuf(freed))
io.netty.util.IllegalReferenceCountException: refCnt: 0, decrement: 1

非常感谢您提供这个开源代码，对我很有帮助。

如果加入自定义编解码，我自定义的编解码器应该放在哪里呢，目前看来似乎是要在TestServer的handleReceive中对收到的Bytebuf消息进行编解码处理。
这样以来就没办法利用pipline()来进行消息流处理。
/ch.pipeline().addLast(new UdpMessageDecoder()); // 消息加解密:消息ID(4字节)/

不知道我又没有把问题描述清楚，感谢！

@OverRide
public void exceptionCaught(ChannelHandlerContext ctx, Throwable cause) throws Exception
{
KcpClient.this.handleException(cause, null);
KcpClient.this.close();
}

KcpClient.this.close(); 是否调整为KcpClient.this.handleClose()更为合理

stream模式下当前包的数据大于上个包可发送剩余数据则会丢弃部分数据，应该把buff扩容

在echo测试下，当发送大于1376大小的数据而分包时，偶尔出现类似丢包现象：收到124之类大小的包。

https://github.com/xtaci/kcp-go/blob/master/fec.go

冗余包来减少丢包率。

客户端建立连接后，发送
ByteBuf bb = Unpooled.buffer(128);
int N = 9000;
for (int i = N; i > 0; i-=4) {
bb.writeInt(i);
}
tc.send(bb);

调试发现 server 的 Kcp.java 的 peekSize 总是走到
if (rcv_queue.size() < seq.frg + 1)
{
return -1;
}

用Kcp进行发送，发送数据不成功

我觉得，因为kcp是纯UDP的。
完全可以用java原生的API，单线程就可以实现。最多把发包和收包各一个线程。一共三个线程就够了。
不需要启动那么多的线程。

接收和发送数据缓存是不是用 环数组要好点呀，java 有提供的 ArrayDeque ; 数组会比链表快好多把