freenode/#sbcl - IRC Chatlog
Search
11:30:20
pfdietz
Bug space is mined out, at least those parts the current random testers can sample. Except for dynamic extent bugs, but no point in finding the same problem there over and over (and ir1-copy bugs).
14:07:11
makomo
hello. does anyone know why this doesn't signal an error? (defgeneric test (a b &rest c)) (defmethod test (a b &key c)) (test 1 2 :d 100)
14:08:35
makomo
case 4 talks about an implicit :allow-other-keys t keyword argument, but i don't think case 4 applies here, since my generic function doesn't have &key in its lambda list
14:09:05
makomo
for regular functions however, this would signal an error: (defun test (a b &key c)) (test :d 100)
14:10:04
makomo
the last sentence of case 4 would explain this behavior, but as i said, it shouldn't be applicable in this case. or is that last sentence a general fact maybe?
14:18:09
Xof
makomo: I won't argue very strongly here, but if I had to defend the current behaviour I would say that "the generic function itself checks the allowable keys" and "the generic function by having &rest in its arglist accepts all keys"
14:18:50
makomo
Xof: that's what i'm thinking too, but apart from the last sentence of case 4, i can't find anything else that would suggest that behavior
14:21:38
Xof
Since the (&rest c) would otherwise do nothing -- it's not like there are real bindings there -- I am somewhat for the existing behaviour