On Thu, Sep 20, 2001 at 01:35:36PM -0400, Steven Vetzal wrote:
> Hi Folks,
>
> I've run across an issue in our environment regarding concurrent
> read/write access of a LOB record. I have duplicated it on Oracle 9i /
> 8i (8.1.5 and 8.1.7), DBI 1.15/1.18, DBD::Oracle 1.06/1.07... a variety
> of environments.
>
> I believe it has to do with what Tim Bunce documents as a two-phase LOB
> update process in the DBD::Oracle driver, but it's just a hunch. I'm
> wondering if it's possible that in-between the LOB locator update and
> the actual LOB data update, a read on the same row would produce an
> empty LOB value (which is what I'm seeing).
Probably. Not sure why - I can't remember the transaction issues off-hand.
If you can't work it out from the code you could always do explicit locking.
Tim.
