9:55:12Xofwell the logic about "can't change behaviour with optimization settings" is true if the code in question is conforming
9:55:56Xofif I had to guess, I'd suspect constant modification or something
9:56:10Xofbut I also wouldn't want to rule an SBCL bug out completely
9:56:12jackdanielACTION mumbles something unintelligibly about mcclim hacks
9:56:47XofI mean, if McCLIM currently has code that looks a bit like a pfdietz test case, that might also be a problem :-)
9:57:26jackdanielhttps://files.mastodon.social/media_attachments/files/009/933/331/original/ee68293b48ba16c6.png (and it is not a bug! :)
10:35:13jdzJust run into a problem: compiling a method that calls the GF it belongs to does not warn about invalid number of arguments. Example: (defmethod test (a b) (test 42)).
12:51:28jdzTry with (defmethod whatever (x) (test 42)).
13:06:08scymtymi tried (defun foo (x y) (class-name x y)) with sbcl-1.4.15 and it doesn't signal a warning
13:10:04jdzSignals for me. Maybe because I have safety set to 3.
13:13:25jdzOr rather, sb-ext:*derive-function-types* is true.
13:17:12jdzSo yes, it would be nice if this warning would be emitted when the callee is a generic function without setting sb-ext:*derive-function-types*.
14:00:18pfdietzACTION reads "calls the GF" as "calls the girlfriend".