[Openmcl-devel] Debugging a class precedence problem

Gary Byers gb at clozure.com
Wed Mar 17 22:16:47 PST 2004

On Mon, 15 Mar 2004, Gary King wrote:

> This is from code that runs under MCL 5.0 but not under OpenMCL. I
> think that the code because of changes in the MOP. I'm not, however,
> quite sure how to interpret the error message or what I should do to
> fix my code.
>  > Error in process listener(1): Class precedence list of
>  > While executing: #<CCL::STANDARD-KERNEL-METHOD
>  > Type :POP to abort.
> Type :? for other options.
> 1 >
> Thanks,
> --
> Gary Warren King, Lab Manager
> EKSL East, University of Massachusetts * 413 577 0176

Sorry for not responding earlier.


(defclass a (b) ())

[assuming that neither A nor B was defined beforehand.]

the class B is just a placeholder for an as-yet-to-be-defined class.
In the MOP, the class B is an instance of the class FORWARD-REFERENCED-CLASS;
if and when it's defined, the class's class will be changed to something
appropriate (e.g., STANDARD-CLASS) via CHANGE-CLASS.

The class A in the example above isn't yet in a fully usable state:
since we don't know much of anything about one of its superclasses
(B), we don't anything about it class precedence list and therefore
don't know what slots it'll inherit or what methods will be applicable
for its instances.  If we try to do anything that requires that A's
inheritance be "finalized" we'll get an error like the one above.
Making an instance of A would require that it not have any
forward-referenced superclasses; defining a (similarly constrained)
subclass of A shouldn't require that

I'm a little surprised that whatever provoked this error worked in MCL;
its CLOS implementation doesn't really formalize the notion of a
FORWARD-REFERENCED-CLASS, but it should recognize that it doesn't
know much about B in this situation and refused to do anything that'd
require this knowledge.

What provoked the error you got ?

More information about the Openmcl-devel mailing list