Alvaro Herrera Munoz wrote:
> I think it would be wise to create a function to discover what savepoints
> are available, and from that know when to release an automatically-
> established savepoint.
>
> Of course, the function would only work when the backend is not in abort
> state, but I think that's a reasonable restriction.
Ok, that would need a set returning function in the backend, right? Could
you help me write it when time comes? (I don't feel backend hacking savvy.)
What about the resources consumed by savepoints
a) that can be freed by a RELEASE (trx state stack, what else?)
versus
b) that cannot be freed (xid locks, anything else?)
Is it worth the effort of extra work (not programmer's but runtime ;-)?
Bruce,
in Revision 1.1355 of the TODO you removed a line from the todo list that
said:
-Use nested transactions to prevent syntax errors from aborting a
transaction
I tought this is what I made a patch for, so shouldn't it be back for 8.1?
Or at least something similar?
Best Regards,
Michael Paesold