If you change it to 4 or 5, do you have the same problem?

>All the tables are row buffered (3).
>
>>Are you using record level buffering or table buffering (4 or 5) ?
>>
>>>>>>>That's a good question. The command is
>>>>>>>

>>>>>>>insert into tiplog (tip_id,employee,changedate,changetable,changefield,oldvalue,newvalue) values ;
>>>>>>>(lnTIP_ID,employee,time(),tcAlias,lcFieldName,alltrim(transform(oldval(lcFieldName))),alltrim(transform(evaluate(lcFieldName))))
>>>>>>>

>>>>>>>tiplog is a plain vanilla dbf living in the dbc. There are no triggers in it or other relations, or relational integrity. So how could there be another table involved? Does multilocks on mandate buffering be enabled?
>>>>>>
>>>>>>I think it's the other way around, can't have table buffering (only record) with multilocks off. Also checked that oldval() doesn't error out when buffering is off - and it doesn't.
>>>>>>
>>>>>>If this is the audit log table... is then maybe the table which is audited to blame for the error?
>>>>>
>>>>>You are correct, this is the audit log. The table being audited has buffering set to 3, as do all of the buffered tables.
>>>>
>>>>I think I've exhausted my telepathic capabilities with this :). IOW, I've run out of clues. Event and coverage logging maybe?
>>>
>>>Thanks for your efforts to here.
>>>Maybe you can help with a related issue. Several of the tables in the dbc have update triggers set to call this logging routine. Something is firing the update trigger whenever the record pointer is moved even when no edits have been made. In my controlled tests this should not happen. I have searched the entire project for explicit tableupdate code and can find none that should be executed with a simple move of the record pointer. Can you help me figure out what is causing the trigger to fire? Debug is no help.