Please report new issues athttps://github.com/DOCGroup
If a process initializes an ORB with a specific listen endpoint, it doesn't seem to be able to use that same endpoint again if it goes through a destroy()/ORB_init() cycle. We want to be able to restart the ORB without shutting down the process, and can only do so by changing the listen endpoint. An empty orbID string is being passed to the ORB_init() call, and destroy() is definitely being called between ORB_init() calls. Is the ORB keeping some socket resources open, or possibly the OS is doing it? I have reproduced the problem using the TAO Hello test and can attach that if needed. Thanks
Could you please post your modifications as a patch, please?
Created attachment 638 [details] Changes to Hello/server.cpp
Created attachment 639 [details] Changes to Hello/client.cpp
That should be it.
I am sorry, but your changes seem to be too extensive for me to grok, at the moment. I would suggest trying to somehow limit them, while preserving the essence of what you want to demonstrate. Also, use "unified" diff _and_ include all changed files in a single patch. Otherwise, it might take more time to understand your changes than one can spare (for free, that is :)) ... HTH, --Iliyan
Created attachment 640 [details] Unified patch without white-space changes
Joe, Thanks for the patch. Could you please, also submit the output you are receiving on the console from running your modified test? Also, if you could run with "-ORBDebugLevel 10 -ORBVerboseLogging 2" for both the client and the server. Thanks for your help, --Iliyan
Created attachment 641 [details] Output of test w/ ORB debugging