This forum has been archived. Please start a new discussion on GitHub.

Problems with user-defined types


I am having problems when I use an operation with a parameter like the following (TDFAgentDescription):

struct TDFServiceDescription
string Name;
string Type;

sequence <TDFServiceDescription> TDFServiceDescriptions;

struct TDFAgentDescription
TAID Name;
TDFServiceDescriptions Services;

If I launch a client written in Python that stores a value in Services and that calls an operation which uses the TDFAgentDescription parameter later, the server written in C++ will raise an Ice::ConnectionLostException.

I have reviewed the problem many times, but I have not found the problem yet. I know that the connection is only losed when the parameter Services has at least one element, but if I initialize Services = [] there is no problem.

Thank you very much!!


  • mes
    mes California
    Hi David,

    It's not quite clear from your description whether it's the Python client or the C++ server that reports ConnectionLostException. Is one of them crashing with a segmentation fault? If so, examining the stack trace might provide useful information.

    Have you tried writing a client in a different language that sends the same data to the server?

    If you can write a small test case that reproduces the problem, we'd be happy to take a look at it.

    Please also specify your versions of Ice, operating system, and compiler.

    Take care,
    - Mark
  • Hello!!

    I use Debian, Ice 3.0.1, and g++-4.0.

    I use IceGrid for the server side, and the message I obtain from icegridnode is the following:

    [ icegridnode: Activator: detected termination of server `DirectoryFacilitator'
    signal = SIGSEGV ]

    On the other hand, I initialize the client with the next code (which causes the error):

    class AgentI (FIPA.Agent, Ice.Application):
    """La clase AgentI representa un agente genérico"""

    def __init__ (self, name):
    """Crea un objeto del tipo AgentI"""

    # AgentIdentifier representa la identificación del agente: name y addresses.
    self.AgentIdentifier = FIPA.TAID(name + '@MASYRO', [])
    # AgentDescription representa la descripción del agente en el Directory Facilitator.
    service = FIPA.TDFServiceDescription('render', 'render')
    self.AgentDescription = FIPA.TDFAgentDescription(self.AgentIdentifier, [service])
    # AgentDescriptions representa la lista de objetos que describen a otros agentes del Directory Facilitator.
    self.AgentDescriptions = []
    # State representa el estado del agente en la plataforma de agentes.
    self.State = FIPA.EState.Initiated

    # ServiceRoot representa la lista de servicios básicos de la plataforma.
    self.ServiceRoot = []
    # Proxies a los servicios básicos de la plataforma de agentes.
    self.StartService, self.Ams, self.Df, self.Acc = None, None, None, None

    And this is the call that causes the error:

    explanation = self.getDf().register(self.getAgentDescription())

    where the getDf() operation returns a proxy to the object which is implemented by the DirectoryFacilitator server.

    The error which I obtain from the client side is the following:

    warning: connection exception:
    TcpTransceiver.cpp:217: Ice::ConnectionLostException:
    connection lost: recv() returned zero
    local address =
    remote address =
    Traceback (most recent call last):
    File "/usr/lib/site-python/", line 487, in main
    status =
    File "", line 496, in run
    topic, proxy = self.subscribeDF(adapter)
    File "", line 453, in subscribeDF
    explanation = self.getDf().register(self.getAgentDescription())
    File "../FIPA/", line 862, in register
    ConnectionLostException: Ice.ConnectionLostException:
    recv() returned zero

    I am going to write a test client in another language as you have advised me.

    Thank you very much!!
  • mes
    mes California

    Since your server appears to be crashing, it shouldn't matter which language you use in your client; the server should receive the same data over the network. I suspect you have a bug in your server that causes the crash, so you should check the stack trace of the segmentation fault for more information.

    Good luck!
    - Mark