Deepa Remesh (JIRA) wrote:
I'm reworking finalize method to continue sending CLSQRY as Kathey suggested
and will upload new version of patch 4.
What about the concern that Dan expressed a few weeks back:
1) Doing anything complex in a finalizer can be trouble, as the order
of finalizer execution is not guaranteed and so any synchronization
can lead to deadlocks. The embedded engine works around this by
having the finalizer just mark some object as not being used any
more and some later action on the connection will also cleanup
the finailzed object.
Do we perhaps need to construct some more sophisticated mechanism in the
client-side infrastructure, so that we can mark a statement as "reclaimable"
in the finalizer, and then have a normal thread perform the cleanup
operations subsequently?
thanks,
bryan
