From:Steve Adams
Date:08-Sep-2000 00:31
Subject:   Strange internal error

I've had a look at the traces and source code. The PL/SQL code is working with a RECORD containing an INDEX BY table, which is internally exactly the same as a transient object containing a varray. The failure appears to happen when trying to establish the array size. Given that this is a transient object, I am inclined to suspect that the MONITORING issue is a red herring. Before you push Oracle too hard on this, I suggest that you build a test case that reproduces the problem. This example is too messy. For example, the SQL in the trace files does not match the source code!

No we don't varrays. At least that's what the developer tells me.

If you have time, here are some trace dumps and the package to look at. Perhaps you can spot something that Oracle Support can't.

KO is the objects layer. Do you have a VARRAY in this table?

Env: 8.1.6 on Solaris 2.7

Have you noticed anything strange with turning on monitoring for a table?

We've got a problem whereby when a user-defined package is executed the first time, it all works well. But subsequent executions cause ORA-7445. ie. reconnecting and re-executing is fine. Just multiple executions from the one session fails. The SQL*Plus session gets ORA-3114 and ORA-3113. The alert.log gets ORA-7445.

Exception signal: 10 (SIGBUS), code: 1 (Invalid address alignment), addr: 0x49442c8b, PC: kolcsiz()+24
ksedmp: internal or fatal error
ORA-07445: exception encountered: core dump [kolcsiz()+24] [SIGBUS] [Invalid address alignment] [1229204619] [] []
Current SQL statement for this session:
BEGIN pkg_jim.call_accumulation; END;

I've done the validate structure cascade for ALL tables in the schema and there are no errors. I've also dropped and recreated the package. But error persists.

But when the table was dropped and recreated without monitoring, multiple executions from the one session are fine. The only thing I can see that's different is that I've turned off monitoring. Will have to test further, to prove the theory. But would like to know if anybody has already encountered this problem.