Hello, Roman,

>> Another issue. I suppose that server exceptions need to be avoided
>> only for unnamed savepoints (because users are not expected to see
>> driver-internal savepoint names). To avoid them actually, you need
>> to invalidate unnamed savepoints at the end of transaction.

> You mean that I should not invalidate named savepoint at all in favor
> of server exception?

Yep.

> But if I have to track unnamed savepoints, I can also track named
> savepoints too and invalidate them all on tx commit... What do you
> think?

I think that if server can produce a nice and correct message why
prevent it from happening ? This exception will have a correct GDSCode,
for example. But this issue is of really minor importance.


--
Nickolay Samofatov mailto:skidder@...