Bug 1092 - Intermittent COMM_FAILURE exception calling method with out parameter
Summary: Intermittent COMM_FAILURE exception calling method with out parameter
Status: ASSIGNED
Alias: None
Product: TAO
Classification: Unclassified
Component: Test (show other bugs)
Version: 1.2
Hardware: SPARC Solaris
: P3 critical
Assignee: Edward Scott
URL:
Depends on:
Blocks:
 
Reported: 2001-11-23 02:21 CST by Stephan Bettermann
Modified: 2004-10-05 21:37 CDT (History)
0 users

See Also:

Attachments
Archive of files to reproduce bug 1092 on Solaris (10.01 KB, application/octet-stream)
2001-11-23 02:32 CST, Stephan Bettermann 		Details
MT_ClientServerException.tar.gz contains the files that may be used to reproduce bug 1092. (10.01 KB, application/gzip)
2001-11-23 02:41 CST, Stephan Bettermann 		Details
Add an attachment (proposed patch, testcase, etc.)

Note You need to log in before you can comment on or make changes to this bug.
Description Stephan Bettermann 2001-11-23 02:21:10 CST 
I have a multi-threaded client calling on a server running a thread pool.  One 
of the methods has an argument list consisting of two out and two in arguments; 
one of each being an unbound sequence of octets.  On Solaris, the client dies 
with the following:

   system exception, ID 'IDL:omg.org/CORBA/COMM_FAILURE:1.0'
   TAO exception, minor code = 6 (failed to recv request response; ENOENT),
      completed = MAYBE

I have created a test-scenario, based on MT_Server, MT_Client and 
Native_Exceptions that -- intermittently  -- reproduces this fault, on our 
machine it occurs after about 1000 iterations.  I
Comment 1 Stephan Bettermann 2001-11-23 02:32:13 CST 
Created attachment 98 [details]
Archive of files to reproduce bug 1092 on Solaris
Comment 2 Stephan Bettermann 2001-11-23 02:41:58 CST 
Created attachment 99 [details]
MT_ClientServerException.tar.gz contains the files that may be used to reproduce bug 1092.
Comment 3 Nanbor Wang 2001-12-08 12:18:02 CST 
We need someone to investigate this. Accepting for tao-support
Comment 4 Edward Scott 2002-07-12 04:46:21 CDT 
Investigating problem.
Comment 5 Edward Scott 2002-07-23 11:43:32 CDT 
accepted bug
Comment 6 Eider Oliveira 2004-10-05 21:37:38 CDT 
I had the same problem running on Linux.
Using tcpdump, I am able to see the server accepts the connection and sends a
RESET packet to client, closing the connection.
Changing the server to Thread per connection, the problem disapears.