Bug 609

Summary:	performance-tests/Latency multi-threaded test fails on some platforms
Product:	TAO	Reporter:	levine
Component:	Performance Test	Assignee:	Carlos O'Ryan <coryan>
Status:	RESOLVED DUPLICATE
Severity:	normal
Priority:	P3
Version:	1.1.4
Hardware:	All
OS:	All

Description levine 2000-07-11 07:55:01 CDT

./run_test.pl 
================ Multi-threaded test
server (14525|1024): user is not superuser, test runs in time-shared class
Activated as <IOR:010000000d00000049444c3a546573743a312e300000000001000000000000
007c00000001010200140000006a61726162652e63732e777573746c2e656475001e0e00001b0000
0014010f0052535481f46939b1c50000000000000100000001000000000300000000000000080000
0001000000004f4154010000001400000001d09e4001000100000000000901010000000000004f41
540400000001d00000>
server (14531|1024): user is not superuser, test runs in time-shared class
(14531|2051) EXCEPTION, Latency: exception raised
system exception, ID 'IDL:omg.org/CORBA/TRANSIENT:1.0'
TAO exception, minor code = 54410085 (invocation connect failed; ECONNREFUSED), 
completed = NO

This is on Linux, w/exceptions.  Does it succeed for you on Linux?
They fail the same way if I start up the processes manually.

It succeeds on Tru64 and Solaris.

The single-threaded and deferred synchronous tests do run successfully
on Linux.

And it fails on Solaris, but differently:

================ Multi-threaded test
server (2393|1): user is not superuser, test runs in time-shared class
Activated as <IOR:00000000000000010000000000000003000000000000007c00010200000000
1464616e7a6f6e2e63732e777573746c2e65647500ca4800000000001b14010f005253543969f3b8
000efbfa000000000000000100000001000000000300000000000000080000000054414f00000000
0100000014000000000001000100000000000101090000000054414f000000000400000000000000
000000007c000102000000001464616e7a6f6e2e63732e777573746c2e65647500ca480000000000
1b14010f005253543969f3b8000efbfa000000000000000100000001000000000300000000000000
080000000054414f000000000100000014000000000001000100000000000101090000000054414f
000000000400000000000000000000007c000102000000001464616e7a6f6e2e63732e777573746c
2e65647500ca4800000000001b14010f005253543969f3b8000efbfa000000000000000100000001
000000000300000000000000080000000054414f0000000001000000140000000000010001000000
00000101090000000054414f000000000400000000>
server (2394|1): user is not superuser, test runs in time-shared class
TAO_Perfect_Hash_Table:find failed
ERROR: client timedout
ERROR: server timedout

Race condition?  Problem with egcs?  (Both my Linux and Solaris builds still
use egcs, not later g++.)

Comment 1 Irfan Pyarali 2000-07-20 21:34:48 CDT

The "connect failed" problem is probably the one noted in bug 189.  Not sure 
about the other one.

Comment 2 Carlos O'Ryan 2000-07-29 12:53:33 CDT

The perfect hashing problem was due to a bad compile, a realclean solved the 
problem.  The TRANSIENT failure is probably due to unaccepted connections,
something we have known about for a long time (bug 189 as Irfan pointed out).
Now we have other problems at shutdown, have to check with Irfan...

Comment 3 Carlos O'Ryan 2000-08-08 13:06:03 CDT

The only errors I'm seeing in the builds are TRANSIENT problems when trying
to connect, that problem is already described in bug 189.


*** This bug has been marked as a duplicate of 189 ***