Re: here is another oracle 10g's bug?

On Jun 21, 6:11 am, Havel Zhang <havel.zh...@xxxxxxxxx> wrote:
On 6 21 , 12 48 , "Vladimir M. Zakharychev"



<vladimir.zakharyc...@xxxxxxxxx> wrote:
On Jun 20, 3:20 pm, Havel Zhang <havel.zh...@xxxxxxxxx> wrote:

On 6 20 , 6 10 , "Vladimir M. Zakharychev"

The explain on 10g as follows:
---------------------------------------------------------------------------­---------------------------------------------------------------------------­---------------------------
| Id | Operation |
Name | Rows | Bytes |
TempSpc| Cost (%CPU)|
---------------------------------------------------------------------------­---------------------------------------------------------------------------­---------------------------
| 0 | INSERT STATEMENT
| | 72276
| 13M| | 826K (1)|
| 1 | TABLE ACCESS BY INDEX ROWID|
STORE_PRODUCT | 1 | 74 |
| 3 (0)|
| 2 | NESTED LOOPS
| | 72276
| 13M| | 826K (1)|
| 3 | HASH JOIN
| | 72276 |
8681K| 7136K| 609K (2)|
| 4 | TABLE ACCESS FULL |
DAY_DEBIT_NE_MARGIN_GROCERY | 72276 | 6281K| | 236 (1)|
| 5 | TABLE ACCESS FULL |...

Well, the plan above says that the insert will process exactly 72276
rows, which is what you expect. The plan for 9i doesn't seem to be
right. Are you sure you didn't exchange 10g and 9i plans?

And if you just select, not insert as select, then both releases
return correct results with and without the index? If so, then it's
definitely a bug: a query should always return the same result set
given the same data and predicates, regardless if it's used as data
source for an insert operation or not.

Regards,
Vladimir M. Zakharychev
N-Networks, makers of Dynamic PSP(tm)
http://www.dynamicpsp.com- -

- -

hi Vladimir
I'm sure I haven't exchange two plans. and i tried again. I think
its Oracle's Bug.
Thank you for your help:)

Havel Zhang

So you say that the query in 10g returns correct result with and
without index, but wrong number of rows is inserted when it's used as
the data source for insert statement in presence of an index. And that
this is reproducible. Can you confirm? If so, definitely sounds like a
case for Oracle Support.

Also, I would recommend to patch your 10g to the latest patchset,
10.2.0.3, and see if the issue persists. Or, if you can't patch your
production db right away, at least load the same data into a 10.2.0.3
database as you did for 9i and test there. This may be a known defect
that's already fixed.

Regards,
Vladimir M. Zakharychev
N-Networks, makers of Dynamic PSP(tm)
http://www.dynamicpsp.com

.

Relevant Pages

  • Re: here is another oracle 10gs bug?
    ... N-Networks, makers of Dynamic PSP ... I'm sure I haven't exchange two plans. ... Also, I would recommend to patch your 10g to the latest patchset, ...
    (comp.databases.oracle.server)
  • Re: Pascal family of languages is getting crowded :-)
    ... Maxwell Smart wrote: ... > plans to release another patch, and the only thing they're currently ... I guess thinking that Borland is evaluating whether or not there should ...
    (borland.public.delphi.non-technical)
  • Re: Multiple IPs in jail
    ... > Thanks for understanding my question, Devon. ... > patch a system here and begin testing with it, ... > is no plans to do it, I just don't want to spend a lot of time fiddling ... > with a patch and then find it in the base system in 5.4 or something. ...
    (freebsd-hackers)
  • Re: Non-executable stack
    ... are there any plans of implementing this and is there a patch I can ... The patch should be for 5.x in general, I don't use it anymore since some ... enable it explicit when you are willing to compile a binary with it. ... Kind regards, ...
    (FreeBSD-Security)