> On Nov. 15, 2012, 4:51 p.m., Alan Conway wrote: > >
Well because of the way the code is structured it is _possible_ (but admittedly unlikely) that some other protocol might still be able to run. As a simple example rdma could still run even if TCP couldn't. In point of fact because of the way the code is currently structured this case can't actually happen anyway! (It should be possible when refactored though). - Andrew ----------------------------------------------------------- This is an automatically generated e-mail. To reply, visit: https://reviews.apache.org/r/8076/#review13473 ----------------------------------------------------------- On Nov. 15, 2012, 3:22 p.m., Andrew Stitcher wrote: > > ----------------------------------------------------------- > This is an automatically generated e-mail. To reply, visit: > https://reviews.apache.org/r/8076/ > ----------------------------------------------------------- > > (Updated Nov. 15, 2012, 3:22 p.m.) > > > Review request for qpid. > > > Description > ------- > > This change introduces a new --interface option to the qpid c++ broker which > allows you to specify either: > * An interface name > * A literal IP address - IPv6 literal addresses should be enclosed in > '['...']' > * A resolvable name > If one (or more) of these options are specified they will restrict what the > broker listens to only be those interfaces/addresses. > > If the option is not specified the current behaviour is unchanged. > > This current change is the first part of the work and only restricts > listening to the existing --port and --ssl-port on the specified interfaces. > There is no way to additionally specify extra ports to listen for, that will > come in the next part of the work and requires some refactoring of the > existing code. > > This current patch is designed for minimal impact and no refactoring. Because > of this there is some element of duplicated code which should really be > factored into a new file. This will happen in the next phase of work. > > Limitation of this current change: You can't specify interfaces by name under > windows, although you can specify individual interface addresses. > > [This change also incidentally removes ans unused function call from > SystemInfo] > > A broken down version of this change can be found on github: > http://github.com/astitcher/qpid/commits/netinterfaces > > > This addresses bug qpid-3351. > https://issues.apache.org/jira/browse/qpid-3351 > > > Diffs > ----- > > /trunk/qpid/cpp/include/qpid/sys/SystemInfo.h 1408983 > /trunk/qpid/cpp/src/qpid/broker/Broker.h 1408983 > /trunk/qpid/cpp/src/qpid/broker/Broker.cpp 1408983 > /trunk/qpid/cpp/src/qpid/broker/windows/SslProtocolFactory.cpp 1408983 > /trunk/qpid/cpp/src/qpid/sys/SslPlugin.cpp 1408983 > /trunk/qpid/cpp/src/qpid/sys/TCPIOPlugin.cpp 1408983 > /trunk/qpid/cpp/src/qpid/sys/posix/SystemInfo.cpp 1408983 > /trunk/qpid/cpp/src/qpid/sys/windows/SystemInfo.cpp 1408983 > /trunk/qpid/cpp/src/tests/SystemInfo.cpp 1408983 > > Diff: https://reviews.apache.org/r/8076/diff/ > > > Testing > ------- > > Tested manually by restricting the listening to either loopback, ethernet, or > a virtual bridge in various combinations using regular and SSL connections > and seeing that connections are correctly refused. > > > Thanks, > > Andrew Stitcher > >
