[omniORB] omniORB and TAO

Андрей Пучин pouchin at gmail.com
Mon Jun 20 18:50:15 BST 2005


Hi!

I've got message from TAO developers.

TAO really doesn't support chunked valuetype encoding  :-((

Friday, June 17, 2005, 4:49:50 PM, you wrote:

> On Friday 17 June, =?Windows-1251?Q?=C0=ED=E4=F0=E5=E9_=CF=F3=F7=E8=ED?= wrote:

>> I've compiled idl with -Wbfactories.

> That registers factories for the Python side. I wanted to make sure you
> have factories registered on the TAO side.

>> May be it's my fault, but i try to patch pyValueType.cc
>> "static void real_marshalPyObjectValue(cdrValueChunkStream& stream, PyObject*
>> d_o, PyObject* a_o)"
>> and set derived=1.

> That should indeed cause omniORBpy to always send at least one repoid,
> which might make TAO happier.

>> op1 method on TAO server can't properly handle recieved arguments
>> despite of lack of MARSHALL exception.
>> 
>> First argument is ok, but the others are garbage.

> What exactly does TAO do?  In what way are the others garbage?

> One thought is that perhaps TAO doesn't properly support the chunked
> valuetype encoding. omniORBpy always uses chunking when it sends values,
> since it's awkward to decide whether it's necessary or not. You could
> try an experiment of sending a value from C++ omniORB to TAO, since that
> only uses chunking when it really has to.

> Cheers,

> Duncan.





More information about the omniORB-list mailing list