[KARAF-7682] Changed handling of the TCP connection to send the stop command

[diedertimmers]

The karaf stop command does not always stop the karaf 4.3.6 framework using the loopback stop option.

After research a number of things stand out:

· There is always a 'java.net.SocketException' in the logging when the karaf framework is stopped · The try with resource statement is used on the Socket · A System.exit(0) call is used inside the try block.

Analysis of these findings shows the following issues:

· The send stopCommand does not include al EOL or EOF character. This results in the read loop ShutdownSocketThread not exiting, and ultimately throwing a 'java.net.SocketException: Connection Reset' in the logging. · The mechanism try-with-resource closes the socket, but this does not guaranty the flushing of the sockets Outputstream. Because this is done asynchronous and after this statement a System.exit() call is made effectively terminating the JVM. · The call to System.exit(0) Terminates the proces without calling the implicit finally clause (which is generated by the java compiler see https://docs.oracle.com/javase/specs/jls/se16/html/jls-14.html#jls-14.20.3 and https://docs.oracle.com/javase/tutorial/essential/exceptions/finally.html ). Because of this calling the close on the Socket is not guaranteed.

Suggested changes:

· Add EOL to the stopcommand before sending · add flush on the outputstream before calling System.exit(0) · Move the System.exit(0) call out of the try block

[mattrpav]

The .flush() here makes good sense. I've seen similar 'hangs' during stop from time-to-time as well. +1