>I store some user preferences in .MEM files on the NT server. There are various good reasons for using a .mem file in this case.
>
>I had a problem with one user where the contents of the mem file were causing an error in my program so using Citrix server I remotely deleted the .mem file while the user concerned was logged off knowing that my program user module would recreate a "default" mem file when he logged in.
>
>Well, the mem file that got created was identical to the one I deleted!
>My program does a simple

>IF NOT FILE({username}.mem)
>  SAVE TO ..... ALL LIKE ...
>ELSE
>  RESTORE FROM...
>ENDIF

>
>If the OS is undeleting the file (say from cache) I would consider this a nasty OS bug. It's completely reproduceable. What do you think?

It's unlikely that NT is undeleting the file or replacing it from cache (there's no undelete built into NT unless you do a controlled deletion through Explorer, and if it were a controlled delete, you'd actually have to perform a restore from the Recycle Bin. Deletes aren't cached or deferred.) I'd guess that it's far more likely that the delete failed in the first place, for any of a number of reasons. You can try to prove this out in a straightforward fashion, by renaming the file rather than deleting and checking the behavior. From your application's standpoint, the rename accomplishes thedeletion (the file won't exist as far as your app is concerned), and if the error occurs, you have the original physical file to examine and compare under the new name.