[omniORB] problem using ThreadPool in OmniORB

[Malge Nishant]

hi Duncan,
	If I restart my server and if client issues directly GIOP::Request
then the Server throws an OBJECT_NOT_EXIST and returns without setting the
connection to be watch-able so when worker thread calls notifyWkDone, the
Peek call returns false. This is true only in case of ThreadPool because in
ThreadPerConnection a thread is dedicated to serve the connection till it is
open.

-----Original Message-----
From: Duncan Grisby [mailto:duncan at grisby.org]
Sent: Tuesday, May 20, 2003 5:32 PM
To: Malge Nishant
Cc: omniorb-list at omniorb-support.com
Subject: Re: [omniORB] problem using ThreadPool in OmniORB 


On Friday 16 May, Malge Nishant wrote:

> 	I am using ThreadPool in omniORB. I have enabled it by disabling
> threadPerConnectionPolicy. (-ORBthreadPerConnectionPolicy 0). Now the
> problem --
> I am running eg2_impl sample given in examples/echo. I have modified the
> eg2_clt in examples/echo so that after each call to server (e->echoString)
> it will sleep for one second. (loops infinite)

OK.

> During run time, If I stop the server (eg2_impl) and restart it, for the
> next call -  eg2_clt gets an exception - OBJECT_NOT_EXIST. eg2_clt ignores
> the exception and loops again and makes an echoString call to the server
and
> hangs forever. No exception is received. 

eg2_impl does not use a persistent POA, so it starts with new objects
each time. You wouldn't expect the client to continue to work.

[...]
> I think there is no need to attach the traceLevel because there are no
> messages on server screen (even with trace level 40).

I don't understand what the symptoms are. Why do you get no trace?
Please figure out how to get a trace, with -ORBtraceLevel 25
-ORBtraceInvocations 1 -ORBtraceThreadId 1.

Cheers,

Duncan.

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