[omniORB] ORBit and OMNIORB interoperability

[Duncan Grisby]

On Thursday 15 May, Chris McGee wrote:

> In this case, I'm not too interested in the naming
> service because I can get the IOR from another source.
> I'm having severe problems with GIOP/IIOP. The ORBit
> code just seems to crash all of the time with
> segfaults. I'm not sure if it is the case that ORBit
> has a very poor GIOP implementation or if OMNIORB is
> sending it invalid messages and is causing the
> segfaults.

omniORB interoperates without difficulty with plenty of other ORBs, so
I doubt it is doing anything wrong. Regardless of that, it should not
be possible for an invalid GIOP message to crash an ORB. Even if
omniORB has a bug, ORBit has a bug too in not handling it gracefully.

If you can post a GIOP trace from omniORB (with -ORBtraceLevel 40)
that triggers an ORBit segfault, we can tell you if it's valid
according to the GIOP spec or not.

> If it's ORBit that is the problem, then how could
> ORBit<->ORBit communication work so well in GNOME? It
> would presumably have to use GIOP as well for this
> sort of communication.

If the ORBit designers have misinterpreted (or ignored) the GIOP spec,
they could have come up with a non-standard way of marshalling certain
things that works fine between instances of ORBit, but doesn't cope
with the correctly marshalled form.

Cheers,

Duncan.

-- 
 -- Duncan Grisby         --
  -- duncan at grisby.org     --
   -- http://www.grisby.org --