Hi Ashley

Thanks for your comments. As a matter of fact, the data used to recreate the error is a local dbf copy of the remote data. The test prg was designed to select specific records just for the purpose you mention, finding corrupted data. None was found and the error seems to be related to a minimum number of records but even this wasn't clear. Just to be sure, I limited the tables to just the single fields necessary to set the relation. The error still happened. Finally, I physically created new tables with similarly structured data and still got the error.

The developers of the odbc dll tested the test exe. The strange thing is that the prg in development mode displays 1 or 2 error messages depending on whether a disconnect was issued, the exe just shuts down. No errors are displayed or trapped, nor does the event manager report a problem. I just know the error occurs because there is a read events and no clear events is issued in that part of the code.

Anyway the developers claim there is a division by zero in vfp8.dll although they can't say where, so even though this doesn't happen in another ODBC connection we are using, I guess I will report it to MS.

Charles Richard

>Throwing in my CAD 0.02...
>I have had working apps terminate abnormally when it finds control characters in the data itself. Eg EOF markers.
>This was usually the result of corrupted data. I have also expereinced this when there was corruption in the header itself.
>What has worked for me is to try similar data from another source to determine if the data is the problem, and if it is, inspect and clean it.