[Sbcl-devel] bug in gf treatment of &rest arguments

[<fa...@gm...>]

In sbcl 0.9.12.26, our application had a problem that could be reduce
to the following:

(defmethod foo ((x string) &key bar baz)
  (list '1 x bar baz))

(defmethod foo ((x cons) &rest keys)
  (list '2 x keys))

(foo "abc" :bar 3) ;ok

(foo '(a) :bar 4) ;error
invalid keyword argument: :BAR (valid keys are ).
  3: (SB-PCL::CHECK-APPLICABLE-KEYWORDS (:BAR 4) 0 NIL)

Works well with CMUCL, clisp.

Also works well if I first do a
(defgeneric foo (x &key bar baz))

I believe it's a bug in SBCL. (But maybe you'll tell me that it's
perfectly conformant for the implementation to throw an error in such
a case.)

[ Fran=E7ois-Ren=E9 =D0VB Rideau | Reflection&Cybernethics | http://fare.tu=
nes.org ]
Every program has at least one bug and can be shortened by at least one
instruction -- from which, by induction, one can deduce that every
program can be reduced to one instruction which doesn't work.