[omniORB] protocol of communication between javaorb and omniORB

[Brian Wallis]

I haven't been following this thread so this may be on the wrong track. 

Are you using the JDK orb in a version prior to jdk 1.5? If so there is a bug 
in their GIOP implementation described in Sun JDK bugId: 4775833 (see 
http://bugs.sun.com/bugdatabase/view_bug.do?bug_id=4775833 for details).

In essense the problem was that the JDK orb adds padding bytes to the output 
stream to align the request body on a 8 byte boundary. This is correct, but 
the JDK orb was adding the padding even when there wasn't a request body.

This is fixed in JDK 1.5 and we worked around it in 1.4 by using JacORB 
instead.

Hope this helps.

On Friday 26 November 2004 2:09 am, Duncan Grisby wrote:
> On Tuesday 16 November, "Diez B. Roggisch" wrote:
>
> [...]
>
> > omniORB: inputMessage: from giop:tcp:192.168.1.3:35588 66 bytes
> > omniORB:
> > 4749 4f50 0102 0000 0000 0036 0000 0005 GIOP.......6....
> > 0300 0000 0000 0000 0000 000e fed7 0b9a ................
> > 4100 003e f100 0000 0001 0000 0000 0004 A..>............
> > 6765 7400 0000 0000 0000 0005 7465 7374 get.........test
> > 0000                                    ..
> > omniORB: Garbage left at the end of input message from
> > giop:tcp:192.168.1.3:35588
> > omniORB: From endpoint: giop:tcp:192.168.1.3:35588. Detected GIOP 1.2
> > protocol error in input
> > message. ../../../../../src/lib/omniORB/orbcore/giopImpl12.cc:789.
> > Connection is closed.
>
> omniORB thinks there is more data in that message than there should
> be. One possibility is that omniORB and the Java ORB have seen different
> versions of the IDL. Other than that, either the Java ORB is really
> sending too much, or omniORB is wrong in thinking there's too much. To
> find out, what is the IDL definition of the get operation you're
> calling?
>
> Cheers,
>
> Duncan.

-- 
brian wallis...
Open Telecommunications
http://www.ot.com.au/