[omniORB] heap corruption in _CORBA_Sequence destructor

[Andrew Buza]

I posted a thread about problems I was having using CORBA sequences
back in June ( http://www.omniorb-support.com/pipermail/omniorb-list/2008-June/029546.html
). Following Martin's suggestion, I double-checked my allocations, use
of _var types, etc., but was unable to find anything that (to my eyes)
was wrong. Coming back to that problem now, I have a sample typical of
where I'm having difficulty. I have the following (abridged) IDL:

struct Node
{
    string kind;
    string key;
};

typedef sequence<Node> NodeList;

interface Hierarchy
{
    Node getRootNode();
    NodeList getChildren(in Node n);
};

My client attempts to get a list of all nodes with kind == "Queue" like so:

{
    std::list<std::string> queues;
    //... get Hierarchy object reference...
    Node_var root = hierarchy->getRootNode();
    visit_node(hierarchy, root, queues);
}

void visit_node(Hierarchy_ptr hierarchy, Node const & node,
list<string> &queues)
{
    if(strcmp(node.kind,"Queue") == 0)
    {
        queues.push_back(string(node.key));
    }

    NodeList_var children = hierarchy->getChildren(node);
    CORBA::ULong nchildren = children->length();
    for(CORBA::ULong c = 0; c < nchildren; c++)
    {
        Node const & node = children[c];
        visit_node(hierarchy, node, queues);
    }
}

When this is run I'll get a message indicating heap corruption and the
IDE (MSVC8.0) will break in the _CORBA_Sequence destructor.

Is the listing above correct, or is there something I'm not
understanding about memory management in CORBA?