Message10928

The relevant Netty change is https://github.com/netty/netty/issues/2166, which happened in 4.0.16; but most likely 4.1 made this more prominent. My recollection is that I don't believe it's possible to have Python socket closing semantics without attempting to close the pool as soon as possible - as it is, we wait up to 100ms. See the _realsocket._finish_closing method in Lib/socket.py for more on what is being done to close the socket.

Therefore we need to take advantage of the rejected logger, which is under the name io.netty.util.concurrent.DefaultPromise.rejectedExecution; see https://github.com/netty/netty/commit/2ba00cd358de2f4fbfef12b0be63605d49826f1a

Assuming SL4J + Logback is still being used, this advice on StackOverflow should be relevant: http://stackoverflow.com/a/11852808/423006. In particular we can put a suitably configured logback.xml in jython.jar or jython-dev.jar to remove this logging: http://logback.qos.ch/manual/configuration.html#auto_configuration