Re: ORBit-WARNING **: No POA found for operation hcmDebug [-1073747064]

[Sebastian Wilhelmi <wilhelmi ira uka de>]

Hi Rajesh,
 
> CLI_PROC -> binds to stale IOR -> SERV_PROC has come back up again
> with a new IOR -> CLI_PROC makes an ORB call to SERV_PROCs object
> using the stale IOR -> CLI_PROC gets back CORBA_SYSTEM_EXCEPTION ->
> CLI_PROC binds to the new IOR written by SERV_PROC -> next call
> succeeds.
> 
> The above explained scenario does not go through. when
> CLI_PROC makes the ORB call for the first time (with the stale IOR),
> though it is supposed to get system exception, it does not. It just
> HANGS after spewing out the message
> ORBit-WARNING **: No POA found for operation hcmDebug [-1073747064]
> on the console. which i believe is coming from the SERV_PROC.
>
> I have tried to debug this, and found that the when the CLI_PROC
> makes the initial request, writev goes through w/o any SIGPIPE which
> should have been logically the case. Because writev goes thru,
> SERV_PROC receives this request, tries to process this request and
> fails at the function :

When the new started SERV_PROC is called for the stale IOR, it means,
that both the dead process (of the stale IOR) and the new one listen to
the same socket. Because using the same unix socket is very improbable
(a random number is used) they surly use the same IP port. That way the
client's call goes through to the server, which in turn doesn't know
about the POA (because it hasn't got it). That's where the error comes
from, I would say. 
 
I agree, that there should be an exception to the server. But there
isn't one in ORBit-stable and I don't want to fix things in ORBit-mt,
which are also faulty in ORBit-stable.

As a solution I would propose to either disable listing on IP ports or
to be more careful with stale IORs.

Bye,
Sebastian
-- 
Sebastian Wilhelmi
mailto:wilhelmi@ira.uka.de
http://goethe.ira.uka.de/~wilhelmi