Maciek Sakrejda wrote:
> I'm not really sure if there's another clean approach, though. I'm
> attaching a patch that takes another dirty approach, but other than
> the fact that it seems to work, I don't really like it at all. The
> basic idea is to cancel whatever you're doing before closing the
> connection: if there's an active copy, have the executor send a
> CopyFail; otherwise, have the ProtocolConnectionImpl request a query
> cancellation.
This seems a bit heavyweight because cancelling a query isn't a trivial
amount of work, and it'll happen whenever close() is called if I read
your patch correctly. 99% of apps are effectively single-threaded so it
seems expensive to always do this when it's rarely needed ..
-O