>>Yes, since MS introduced filter/nofilter cursor the trustworthy of _TALLY is somewhat diminished (btw I would test it with NOFILTER, if you SELECT inot cursor). I didn't meet particular problem you mention, but if I would then I should better use RECCOUNT()<>0 to have safer code.
>
>Ye-gads,
>Now we can no longer test the result of a query. RECCOUNT() will fail if the query is fully optimisable, because reccount() will return the number of records in the underlying table - unless you specify the NOFILTER clause which will slow it down.
>So some time ago I stopped using RECCOUNT() and always used _TALLY. Now it's a dud there's nothing reliable left!
>:-((

I'm still a believer in _tally :) I have yet to see a case where it is incorrect, so long as it is the VERY FIRST THING done after an SQL (otherwise, it's useless). This often means assigning it to a local, and then using the local for any comparisons, etc. If anyone can come up with a case where _tally is incorrect when used this way, please post it...