14:46:17makomoXof: the issue regarding invalid keyword arguments within methods and SBCL not signaling an error might be a non-conformance issue after all
14:46:54makomothe example was: (defgeneric test (&rest a)) (defmethod test (&key b)) (test :c 100)
14:47:18makomo(test :c 100) doesn't signal an error
14:48:08makomothe relevant section is 7.6.5 http://www.lispworks.com/documentation/HyperSpec/Body/07_fe.htm
14:48:48makomothe set of accepted keyword arguments should be formed by collecting all of the keyword arguments from all of the applicable methods
14:48:59makomoeven when the GF itself has no &key in its lambda list
14:49:48makomoin comparison, CLISP does signal an error
14:50:26makomoalso, as said before, the last sentence from case 4 from 7.6.4 http://www.lispworks.com/documentation/HyperSpec/Body/07_fd.htm shouldn't apply in this case since the GF doesn't have a &key in its lambda list
16:45:36Xofyeah, but. If any applicable &key method also had &rest, :c would be acceptable
16:45:49Xofthe generic function lambda list has &rest.
16:46:06XofThis might be a question of inferred intent vs letter of the law
16:46:18Xof(and maybe we should signal an error, just because someone else does)
16:48:31jackdanielclisp does; abcl, ccl, ecl and sbcl doesn't (I don't have other implementations locally)
16:55:34makomoXof: would it though? why would the existence of &rest along with &key disable the checking of the validity of the keyword arguments?
16:55:44makomoCLISP signals an error in that case too
16:56:05makomo^ the above is within methods of course, the generic function still only has &rest