I'm testing that if you try to use MessageTransport.executeRequestAsync(..) and the server side of the network has closed its input stream, then an exception will be thrown (an ExecutionException with an underlying cause of IllegalStateException).
I don't think shutdownInput(..) would work, because it says that any input coming in will be silently discarded. I could probably use network.getClientSocket(..).shutdownOutput() to achieve the same effect, though I was really trying to test the case where the server side shut down the input stream. I don't know if there is a better way to do this than polling. On Mon, Mar 15, 2010 at 4:28 PM, <k...@google.com> wrote: > Now we understand the mysterious failure. Thanks for finding it! > > > http://gwt-code-reviews.appspot.com/111808/diff/1/2 > File > dev/core/test/com/google/gwt/dev/shell/remoteui/MessageTransportTest.java > (right): > > http://gwt-code-reviews.appspot.com/111808/diff/1/2#newcode135 > Line 135: sleepCycles++; > We should fail with a clear message if the socket cannot be closed, > rather than a misleading message down stream. > > > http://gwt-code-reviews.appspot.com/111808 > -- http://groups.google.com/group/Google-Web-Toolkit-Contributors