Bob Rogers <rogers-perl6@xxxxxxxxxxxxxxxx> wrote:
> From: Leopold Toetsch <lt@xxxxxxxxxx>
> But it isn't known, if the last reference to that PMC is reused or not.
> If it is the last reference then the finalizer has to run, else not.
> As we don't have reference counts, an assign to an object with
> finalizers ougth to trigger a GC run, IMHO.
> Is that really necessary? After a PMC has been morphed, it seems to me
> that there is no possible way for anyone to reference whatever it was in
> its previous life. So the process of morphing ought to be able to do
> any finalization of the old PMC that the GC would do, since the old PMC
> is really and truly going away forever. In fact, morph *must* finalize;
> since we still have a pointer to the PMC being morphed, the GC won't
> even know that it needs it. True?
Yep, you are right. I've mixed up referencing a PMC and reusing the PMC
storage. Finalize/destroy has to be done immediately in pmc_reuse(), if
the PMC needs it.o
Which OTOH means that it's possible that arbitrary user code could run
inmidst an assign statement, *if* such a PMC would allow to be reused.
Conclusio:
- morph for scalars is ok: e.g. an Integer transforms itself to a Float
- morph for arbitrary objects must never be done automatically
> -- Bob Rogers
leo
|
