>Hogwash, Harry.

ooh. I love it when you're Masterful! ;-)

> You just have to break the problem into manageable parts.
etc

Seriously. I take the point. But is more difficult than you suggest.

> In other words, if this is a VFP bug, then you should be able to reduce it to a simple set of reproducible circumstances.

not easily. The conditions for the bug are that the whole app is up and running with the appropriate files open and relations set.

>Take the offending code, and put it in a base class form, and test.

we've already tried running it in a separate prg.
And suspending it at the critical point and manually typing in the command which triggers the bug (ANY index command will do it - even the simplest single variable index)

>Just manipulate the error's environment,

this is the crucial bit. I'm sure its something weird in the memory, probably resulting from code run a dozen lines earlier, rather than anything happening at the apparent point of trouble. But how to isolate prior events without making it impossible to give the instruction in the first place is a knotty problem in its own right.

>You can always solve this sort of problem with brute force and the scientific method.

The brute force appeals right now. Think I'll take a hammer to it!

Harry