Today, Gerd Moellmann <gerd.moellmann@...> wrote:
> I'd expect this to happen because I've incidentally just today read a
> comment and code in PCL again which explicitly makes the assumption
> that the method combination of a gf doesn't change. Someone (tm)
> would have to check if that's compliant with whatever is relevant.
"If function-name specifies a generic function that has a different
value for any of the following arguments, the generic function is
modified to have the new value: :argument-precedence-order, :declare,
This does not say anything about recomputing the effective method,
though. And this is how PCL currently behaves. I could not find mention
of the effects of GF redefinition and what happens when a method
combination actually changes. Maybe somebody with more Spec Reading
Ability can help?
> And, I'd almost bet a small amount that adding the DOLIST to the
> following method would make your case work (CMUCL + untested).
And you'd almost have won an equally small amount (-:
> (in-package :pcl)
with (in-package :sb-pcl), it seems to work like a charm.
IMHO the current behaviour re. method combination changes is not optimal
(or obvious). If you decide to keep it, may I suggest raising a
Andreas Fuchs, <asf@...>, asf@..., antifuchs