Breaking out of RecordLock...

[Ken Ray]

> The way I understand what you describe will indeed leave a cursor 
> open until the program owning that cursor quits. But this should 
> happen only in a local mode. In server mode, the server should close 
> cursors when client disconnects AFAIK. I believe that this is what 
> Ruslan means "cannot be". If this does happen, Ruslan has to go bug 
> hunting.

No, Ruslan's right... it *does* close when the client disconnects, but
since we don't know which client left the cursor open, everyone on the
network needs to exit our program and disconnect from the server. This
is something I was hoping to avoid if possible, because it is a terrible
inconvenience to our customers.

> The way I deal with that is that as a rule, I create and remove 
> cursors within the same handler. It makes it easier to see whether 
> all calls are there and I can use a scavenger script (which lists all 
> Valentina calls per handler in each object) and quickly inspect if 
> anything is amiss. Further, use SQLSelectRecords if you are only 
> retrieving data.

That's exactly what we've been doing, which is why we find it odd that a
cursor might be left "open"... we've been very careful to create and
remove cursors in the same handler. Our program is quite large, though,
with dozens (if not hundreds) of SQL calls, so tracking this down will
be hard.

Ken Ray
Sons of Thunder Software
Email: kray at sonsothunder.com
Web Site: http://www.sonsothunder.com/