burke@pollux.usc.edu (Sean Burke) (08/13/89)
Here's the code which was the culprit, condensed to essentials, from
walk.lisp:
;;;
;;; VARIABLE-GLOBALLY-SPECIAL-P is used to ask if a variable has been
;;; declared globally special. Any particular CommonLisp implementation
;;; should customize this function accordingly and send their customization
;;; back.
;;;
(defun variable-globally-special-p (symbol)
(or (not (null (member symbol *globally-special-variables* :test #'eq)))
(when (eval `(flet ((ref () ,symbol))
(let ((,symbol '#,(list nil)))
(and (boundp ',symbol) (eq ,symbol (ref))))))
(push symbol *globally-special-variables*)
t)))
The above code would choke if symbol were a constant, such as nil, t or a
keyword, since the let form attempts to bind symbol. I am trying to figure
out what this code is trying to do. The flet defines ref so that it will
return the global binding of symbol, then the let binds symbol to (list nil),
presumably a fresh cons which could not be the value of any previous binding.
Within the let then, symbol is bound and could never be eq the global
binding of symbol, even if symbol had been declared special. So this code
must be trying to determine whether symbol were PROCLAIMed special, in which
case even the let form binds the special variable, and thus ref would return
the new binding created by let. Is this the implied meaning of the term
"declared globally special"? (CLTL uses the term pervasive", p158.)
Anyway, assuming the above is true, then there is no need to worry that
some constant might be "globally special", since you can't rebind one
in any case. So my fix was to insert (not (constantp symbol)) at the
appropriate spot.
The only thing missing from this tidy explanation is why this hasn't
happened to everyone else who's ever tried to load this system into common
lisp. Kind of disquieting. IMHO, the difference between scheme and
commonlisp is that scheme is only one language.
Sean Burke
"The nice thing about true hopelessness
is that you don't have to try again" - Jules Shear