https://bz.apache.org/bugzilla/show_bug.cgi?id=60560
--- Comment #7 from Coty Sutherland <csuth...@redhat.com> --- Comments from a review of the latest provided patch: * Your patch adds an unnecessary space before the TimeUnit import in NioEndpoint * There is a typo, 'automatically', in http.xml * The useInheritedChannel documentation addition in http.xml specifies type (boolean) which is inconsistent with other (bool) attrs * If you specify useInheritedChannel on a Connector it shouldn't fall back to the default behavior when starting without an OS provided socket. If you specifically designate that tomcat should use an OS socket, then it should fail to start the Connector (like the behavior that causes an SSL Connector to fail to start when it can't find the keystore). * If you're borrowing a socket from the OS for tomcat to use (by skipping the bind), then you shouldn't be closing the socket in NioEndpoint.unbind(). After applying this patch the behavior seems to work (mostly, systemd is behaving a bit odd, but that's not tomcat's fault), but there's something that may be confusing for users: the thread name (i.e. Starting ProtocolHandler ["http-nio-8080"]). If you specify useInheritedChannel on the Connector then you're no longer using the specified port. We should probably address that somehow. -- You are receiving this mail because: You are the assignee for the bug. --------------------------------------------------------------------- To unsubscribe, e-mail: dev-unsubscr...@tomcat.apache.org For additional commands, e-mail: dev-h...@tomcat.apache.org
