RuntimeException in WebSocketWorker should not stop server

[sippykup]

Here is the method I'm referring to, in WebSocketServer.java:

        public void run() {
            WebSocketImpl ws = null;
            try {
                while ( true ) {
                    ByteBuffer buf = null;
                    ws = iqueue.take();
                    buf = ws.inQueue.poll();
                    assert ( buf != null );
                    try {
                        ws.decode( buf );
                    } finally {
                        pushBuffer( buf );
                    }
                }
            } catch ( InterruptedException e ) {
            } catch ( RuntimeException e ) {
                handleFatal( ws, e );
            }
        }
    }

handleFatal() is what calls stop(). My argument is that you shouldn't treat this RuntimeException as a fatal error that you cannot recover from. Here's the stack trace that caused our production server to shut down today:

java.lang.NullPointerException
at org.java_websocket.WebSocketAdapter.getFlashPolicy(WebSocketAdapter.java:87)
at org.java_websocket.WebSocketImpl.decodeHandshake(WebSocketImpl.java:193)
at org.java_websocket.WebSocketImpl.decode(WebSocketImpl.java:161)
at org.java_websocket.server.WebSocketServer$WebSocketWorker.run(WebSocketServer.java:663)

That operation is just trying to decode a handshake from a single client. Anything unplanned that goes wrong while dealing with one client should not have the effect of shutting down the entire server. It should log an error, close the client connection, and move on.

[Davidiusdadi]

handleFatal will only be called if there happened an unrecoverable error like the one you posted.
If the method gets called there is a bug in the library!

Even if the method would not be called there is no way that the server remained operational.
Therefore calling stop in handleFatal does not really initiate the stop operation - it's rather an attempt to free allocated resources.

I realize however that the users may want to overload the method...

[marci4]

Hello @sippykup,

is this still an issue for you?
Do you want me to look into it again?

Greetings
marci4

[marci4]

No update on this issue for months.

Closing it!
Greetings
marci4