When there is a problem with handshake, we just close the connection after logging the SSLHanshakeException.

As a result, the other end gets a connection unexpectedly closed when it tries to read or write to the SSLChannel.

The other end will be able to display a better error message if we try to wrap and flush when this occurs, as SSLEngine will apparently communicate the problem to the other end.

This problem is reproducible on Windows when using Undertow with a null cert chain on the client side. See attachments for more info.