"Adriano dos Santos Fernandes" <adrianosf@...> wrote:

>
> > 2) A domain change forces all dependent BLR to load and parse. Those

failed

> > to parse are marked as invalid in the schema until its BLR is

regenerated

> > and/or parsed successfully. It's the same as above, but not all objects

are

> > invalidated but only those affected by an incompatible change.
>
> I'd vote for 2.

And how/where should we store the "valid/invalid" flag? Note that I
suggested to make it persistent. RDB$FLAGS in RDB$PROCEDURES?

> And if at least one BLR has invalidated, the engine returns a warning
> when altering the domain.

I'm not sure it's really useful. Sometimes it could cause confusion.

> We may also create a ISQL command to show invalidated SPs/triggers.

Nice idea.


Dmitry