Re: Refcounted object semantics

lör 2006-09-02 klockan 17:59 +0800 skrev Adrian Chadd:

> The big thing that brought along this bug was the semantic change
> when Duane moved the fwdstate code over to use the refcounted code.
> My fledgling C++ knowledge tells me that the destructor isn't the
> place to put this - there's no real explicit way to say "I really,
> honestly want you to kick off the forwarding completion code *now*"

Agreed. refcounting is nice, but things like FwdState needs to have
explicit invalidation in addition to the destructor. Actually true for
most objects actually doing something actively and not only storing some
value.

Regards
Henrik