>Hi all,
>
>After a server change-over two weeks ago my users begin coming to me with "Invalid seek offset" (Error 1103) messages. VFP help is no use and there is nothing useful on a KB search using 'Invalid seek offset' as string.
>
>1) What does this message mean?

It typically means that a pointer to a file location is grunged - either an index is corrupted, or the pointer to a memo or general field in the FPT file is invalid. In the case of an index file, reconstructing the index from scratch (not using REINDEX, since it is possible that the index file header is corrupted) should fix it; in the case of a grunged pointer into a memo file, a DBF repair tool will be needed, and it is likely that some data in the memo or general fields will be lost as a result.

>
>2) Is it application driven or data-driven?
>

It's an artifact of file corruption on the VFP side; the problem could well be caused by operator error (turning off the computer without exiting the program) or an OS error. It has also been reported under specific circumstances with some versions of VFP in specific scenarios; check the MSKB for specific instances.

>3) Ideas on how to resolve?
>

In a shared data environment, disabling some caching options, increasing the number of simultaneously supported share sessions, and disallowing server-initiated disconnects of a session due to idling behavior may all help.

>We are a VFP6 SP5 shop running WIN2000 (network and workstations). The application uses a DBC and a couple of small free tables. When and where the error occors is variable.
>

I'd start by disabling deferred writes on the server, opportunistic locking, increasing the session count and disabling NT-style file handling; these are all accomplished via registry patches on the server.