Re: CURSOR_SHARING bugs fixed in 8.1.7.4?

andreyNSPAM_at_bookexchange.net (NetComrade) wrote in message news:<3e835cde.2690891387_at_nyc.news.speakeasy.net>...
> unfortunately i don't this this is the case:
> every trace file contains a query going over the link
> as soon as i disabled cursor_sharing=force, it went away
>
> Thu Mar 27 15:16:06 2003
> Errors in file /u31/app/oracle/admin/MLS1/bdump/mls1_s002_226.trc:
> ORA-00600: internal error code, arguments: [kkslgop1], [], [], [], [],
> [], [], []
> Thu Mar 27 15:18:20 2003
> Errors in file /u31/app/oracle/admin/MLS1/bdump/mls1_s002_226.trc:
> ORA-00600: internal error code, arguments: [kkslgop1], [], [], [], [],
> [], [], []
> Thu Mar 27 15:19:41 2003
> Errors in file /u31/app/oracle/admin/MLS1/bdump/mls1_s002_226.trc:
> ORA-00600: internal error code, arguments: [kkslgop1], [], [], [], [],
> [], [], [
>
> This was also a problem in a different db when i enabled
> cursor_sharing=force:
>
> SQL here was of a different nature (database has different functions):
> select count(*) into :b0 from member_resource_privileges mrp
> ,resources r where ((((mrp.member_id=:b1 and mrp.resource_id=r.id) and
> (:b2 like concat(r.ac
> cess_dir,:"SYS_B_0") or (:b2||:"SYS_B_1") like
> concat(r.access_dir,:"SYS_B_2"))) and
> sysdate>=trunc(mrp.access_begin_ts)) and trunc(mrp.access_end_ts)>=sys
> date)
>
> Wed Mar 26 03:18:22 2003
> Errors in file /u11/app/oracle/admin/VCRS/udump/vcrs_ora_29526.trc:
> ORA-07445: exception encountered: core dump [kkslhib()+464] [SIGSEGV]
> [Address not mapped to object] [16] [] []
> Wed Mar 26 03:18:42 2003
> Errors in file /u11/app/oracle/admin/VCRS/udump/vcrs_ora_29473.trc:
> ORA-07445: exception encountered: core dump [kkslhib()+464] [SIGSEGV]
> [Address not mapped to object] [16] [] []
> Wed Mar 26 03:19:40 2003
> Errors in file /u11/app/oracle/admin/VCRS/udump/vcrs_ora_29496.trc:
> ORA-07445: exception encountered: core dump [kkslhib()+464] [SIGSEGV]
> [Address not mapped to object] [16] [] []
> Wed Mar 26 03:19:45 2003

Hi.

I didn't notice that your original post included in its scope db_links.
Yes, I had also seen such behavior for queries executed over db_links, so those databases that have db_links do not have cursor_sharing set to FORCE.
You could set this at the instance level, and disable the use of cursor_sharing at the session level (e.g. in a logon trigger).

I have not tested such behavior yet in 9.2.0.3.

From the readme in the 8.1.7.4.1 patchset:

2221407 - ORA-3106 / OERI:15212 using BINDS over DBLINK with CURSOR_SHARING=FORCE
1551563 - ORA-7445 re-executing SQL with BINDS against SYNONYM which was REMOTE but is now LOCAL

From the readme in the 8.1.7.4.6 patchset: 2644000 2514869 RDBMS SELECT STATEMENT END UP IN CORE DUMP WHEN

                            CURSOR_SHARING=FORCE IS SET

Paul

> On 19 Mar 2003 21:48:02 -0800, drak0nian_at_yahoo.com (Paul Drake) wrote:
>
> >andreyNSPAM_at_bookexchange.net (NetComrade) wrote in message news:<3e78980d.1985145708_at_nyc.news.speakeasy.net>...
> >> Does anyone know if oracle stopped dumping cores in 8.1.7.4 when
> >> cursor_sharing is set to force?
> >>
> >> Thanks,
> >> .......
> >> We use Oracle 8.1.7.4 on Solaris 2.7 boxes
> >> remove NSPAM to email
> >
> >yes.
> >I believe that it had stopped for us with 8.1.7.2.6.
> >I didn't see any as of 8.1.7.3.0.
> >8.1.7.1.x dumped in a big way.
> >
> >Paul
>
> .......
> We use Oracle 8.1.7.4 on Solaris 2.7 boxes
> remove NSPAM to email
Received on Sat Mar 29 2003 - 10:16:38 CST