Re: Oracle oledb bug?

BicycleRepairman (if that is your real name),

      Thanks for your reply. I did check the docs and you're correct that it doesn't indicate that ownname defaults to the current user on that proc. Strange since it does for GATHER_SCHEMA_STATS, maybe the documentation has a bug. Anyway, I did try hardcoding the user in and it still had the same effect. Are you able to replicate this in your environment? In our real scenario the call is inside a package procedure. Pseudo code is something like this:

--===============

Procedure MyProc

Insert into parent_table select distinct parent_vals from bulk_load_table; --could be anywhere from 100 to 100,000+

gather stats-- needed because otherwise the next call(s) will end up using undesired explain plans (full table scans, cartesian joins, etc)

insert into child select distinct child vals joined to distinct parent vals joined to parent table to get surrogate key generated during parent insert

...bunch more dependent inserts/updates

--===============

used to work but now fails with the EOF error because the parent table has a composite index. Note that it doesn't fail when called from anywhere else (sqlplus, oracle job).

I think I have managed to run into ever single obscure Oracle bug on this project, including jobs ceasing to run after 497 days of uptime on a Sun Box (seriously, bug #3427424). I'm going to submit this as a TAR and see what Oracle says and I'll post back with the results.

Thanks Received on Thu Feb 23 2006 - 12:52:29 CST