BAD_INV_ORDER_ORBHasShutdown when accessing an external proxy from a device server

[vxgmichel]

It can easily be reproduced using the following pytango server:

import tango.server

proxy = tango.DeviceProxy('sys/database/2')

class Test(tango.server.Device):

    @tango.server.attribute(dtype=str)
    def database_status(self):
        return proxy.status()

if __name__ == '__main__':
    Test.run_server()

Exception:

DevFailed: DevFailed[
DevError[
    desc = BAD_INV_ORDER CORBA system exception: BAD_INV_ORDER_ORBHasShutdown
  origin = DeviceProxy::status()
  reason = API_CorbaException
severity = ERR]

DevError[
    desc = Failed to execute status() on device (CORBA exception)
  origin = DeviceProxy::status()
  reason = API_CommunicationFailed
severity = ERR]

DevError[
    desc = Failed to read_attribute on device test/nodb/test, attribute database_status
  origin = DeviceProxy::read_attribute()
  reason = API_AttributeFailed
severity = ERR]
]

[bourtemb]

Thanks for the report Vincent.

After talking with @taurel, it seems like your problem is linked to issue #145. The comments in this ticket explain well what you are seeing.

It all comes because omniORB manages the ORB object as a singleton.

To solve #145, the "client" ORB is destroyed when the device server is started, to get a "server" ORB. By doing so, it invalidates all the CORBA objects which were created before the "client" ORB destruction.

[mguijarr]

Thanks for the quick help.

If it cannot be fixed, it would be great to display an error message if some client objects were alive at the moment of the ORB destruction, so we would be warned.