Re: Brainstorming

[CRPence <CRPbottle@xxxxxxxxx>]

On 05 Jun 2013 12:28, Thomas Garvey wrote:
> We have a program that is apparently failing at a user site, but not
> on our test and development systems.  The nature of the failure is
> that selected data does not appear as expected in a subfile program.
>
> To test it we had the object library and data library from the user
> system saved and sent to our test system.  We restored the objects
> and tested the process here.  It works just fine.
>
> So, what could be different at the user site and our site that would
> change whether records are displayed in a subfile?
>
> <<SNIP>>
>
> Any other ideas that would make something work on one system and not
> another?  Remember, both the program object and data files libraries
> from the user site were restored to the test system.
>
> Thanks for any advice.

  Check the history log from after the restore of the data library, for 
"access path rebuilt" messages.  Also check for increased Access Path 
sharing on the restored copy as compared to the test and dev systems.

  An assumption by the code that a specific [un]shared keyed AccPth is 
being used, could cause the code to fail when the assumption fails due 
to the side effects of save\restore for database file access paths. 
Note: having properly requested the saving of access paths would 
minimize [but not prevent] rebuilds, but would do nothing to prevent 
sharing, because only an incompatible definitional difference between 
sufficiently-like AccPths will prevent the DB from effecting the sharing.