Wed, 21 Jan 1998 14:56:35 -0700 (MST)
After further examination of my "problem", I think I understand what is
happening, and I don't think there is any problem with omniNames or omniOrb.
I was getting a COMM_FAILURE because before doing an unbind I was attempting
to determine if the object to be unbound was a context or object. As it
is done in nameclt, this requires the object to be resolved and then an
attempt is made to narrow it to a NamingContext. The narrow operation
can require a call to _is_a, which must sometimes go to the implementation
of the object. This causes a COMM_FAILURE if the object isn't available.
So this was purely a cockpit/learning error and not a real problem.
Maybe someone could comment on what conditions require _narrow to go
to the object implementation?
> Date: Tue, 20 Jan 1998 17:05:55 -0700 (MST)
> From: Kim Gillies <firstname.lastname@example.org>
> Subject: omniNames ?
> To: email@example.com
> MIME-Version: 1.0
> Content-MD5: 9WeE3iPfUlfMhjoqVenSnQ==
> Hello All,
> I'm working on porting some of our Neo code that uses the name service to
> omniOrb and omniNames.
> On my setup (Solaris 2.6, Ultra 30) I've found that I can't "unbind"
> an object unless it is running. If it isn't running I get a COMM_FAILURE
> exception. This also occurs from nameclt.
> Is an object really required to be running for me to unbind it? Or is
> this something odd in my environment? How does one remove a nonexistent
> Kim Gillies
> Gemini 8-m Telescopes Project