This may have more to do with my limited understanding of valuetypes
than with mico, but I've been banging my head against this for a couple
of days and need some more expert input (read: I'm almost out of
aspirin). I ran this test on three development environments, all with
the same results:
- micoo 2.3.12RC2, gcc 3.4.4, RedHat Linux 9
- mico 2.3.7, gcc 3.2.1, RedHat Linux 7
- mico 2.3.11, VC++7, Win2k
I've attached my test code, but here's a summary of what's happening:
Using this idl:
module test {
valuetype testvalue {
public long ival;
long foo();
};
interface testinterface {
testvalue getValue();
any getAny();
long dofoo();
};
};
The server side is straightforward. I'm trying to do this on the
client side:
(Get a CORBA::Object_var obj..)
testinterface_var intfc = testinterface::_narrow(obj);
if (CORBA::is_nil(intfc)) {
cerr << argv[0] << ": can't narrow testinterface" << endl;
return -1;
}
CORBA::Any_var any = intfc->getAny();
testvalue_impl * pCValue = new testvalue_impl(999);
test::testvalue_var tv = pCValue;
if (any >>= tv) { <<<<<<<<<<<<<<<<<<<<< ERROR!
printf("%s: Extracted the any. ival = %d, foo = %d\n",
argv[0],
pCValue->ival(),
pCValue->foo());
} else {
cerr << argv[0] << ": Failed to extract the any" << endl;
return -1;
}
return 0;
When I try to extract the testvalue from the any, I get a
CORBA::Marshal exception. I've tried using DynAny to parse the any,
and that works but is very clumsy. So I know the correct information
is in the any but I can't extract it.
Using gdb, I got a backtrace of the error. Everything looks fine up to
this point in the c++ code generated from idl:
::CORBA::Boolean _Marshaller_testvalue::demarshal( ::CORBA::DataDecoder
&dc, StaticValueType v ) const
{
::CORBA::ValueBase* vb = NULL;
if (!::CORBA::ValueBase::_demarshal (dc, vb, "IDL:testvalue:1.0")) {
<<<<<< ERROR!
return FALSE;
}
::CORBA::remove_ref (*(_MICO_T *)v);
*(_MICO_T *)v = ::testvalue::_downcast (vb);
if (vb && !*(_MICO_T *)v) {
::CORBA::remove_ref (vb);
return FALSE;
}
return TRUE;
}
The CORBA::ValueBase::_demarshal function takes a const std::string
reference as the third parameter. But according to gdb, the third
parameter becomes a reference to a null pointer. Is gdb lying to me,
or is something else happening here? I admit I have minimal experience
with gdb, but I also tried creating my own test function like this:
long strtestfn(const std::string& str) {
printf("In strtestfn: %s\n", str.c_str());
return 42;
}
int main(int argc, char * argv[]) {
long result = strtestfn("Hello");
}
.. and this worked without problems. So I'm stumped. Please help!
(Note: Here's the big picture: I'd like to deliver valuetypes using
the event service, so I could have all the data passed like a struct,
and in rare cases also have functions I could call to get more info
from the server. Maybe I could accomplish the same thing by having an
object reference in my struct, and only call methods on that object in
those rare cases. Thoughts?)
Thanks, and keep up the great work on mico!
