Multithreaded PRO*C application coredump

From: diwa209 <member20381_at_dbforums.com>
Date: Thu, 12 Jun 2003 01:56:33 +0000
Message-ID: <2992627.1055382993_at_dbforums.com>


Hi,
My PRO*C(8.1.7.4.0) application on SUN OS 5.7 64bit(-xarch=v9) spawns a few threads
(POSIX threads), each of which has a separate
connection to the ORACLE(8.1.7) server and each of them use their own context (sql_context) to connect as well as access the database.
Each of the threads do SELECT's heavily and almost every SELECT is via a explicit cursor.
Every thing seems to be working fine for a while, infact quite a while, after which one of the threads will have a Segment violation.
The segment violation does not happen when only one thread is spawned. The SEGV also does not happen if the database access is protected by a mutex lock. SEGV does not happen at the same place everytime. Often, two threads are accessing the same record
(from 2 diff tables, though, related with a foreign key).
Very unlikely i am runnnig out of stack space too.

Anyone who has done something similar ?? Has it anything to do with row level locking. Did not find any patch for 8.1.7 which lists such a problem. HOLD_CURSOR=YES THREADS=YES
Thanks and regards,
Diwakar

--
Posted via http://dbforums.com
Received on Thu Jun 12 2003 - 03:56:33 CEST

Original text of this message