[orbitcpp-list] running orbitcpp client and server programs from separate machines

[Bill L. <bl...@mi...>]

Hi,

I installed ORBit-0.5.0-1, ORBit-devel-0.5.0-1,  orbitcpp-0.22-1, and
orbitcpp-0.22-1 on three machines, all running RedHat 6.1.  As far as I
can tell, all three machines have similar network configurations.

I built and ran the simple helloworld sample on various combinations of
these machines (call them A, B, and C).

I perverted the sample slightly so that the programs could read and
write IOR strings from an LDAP server on A.

Running server on A and client on A works.
Running server on B and client on B works.
Running server on C and client on C works.

Running server on A and client on B works.
Running server on B and client on A works.

Running server on A and client on C causes the client to crash and dump
core.
Running server on C and client on A causes the client to crash and dump
core.

I added code to trap exceptions in the client program, and it appears to
be a system exception, minor code of 5, thrown when the client program
calls the _narrow() function.

I can't figure out why the sample works for A and B but not for A and
C.  I thought maybe C might have too low a limit on open file
descriptors, but that was not the case (1024 for all machines).

Is there a particular setting that causes client crashes (untrapped
exception, minor code 5) when the orbitcpp helloworld samples are run on
two separate machines?

The ORBit source code looks like a steep learning curve, so I thought
I would ask around before I go any further into the source, in case
there is something very simple wrong here.   I am not very knowledgeable
about CORBA in general, which is why I used the ldap server instead of
the orbit name server.

If anybody has any suggestions, I am (like Ross Perot) all ears.

Thanks.

--Bill Lasley