Re: Typecheck hole in safety = 3, speed = 3 code

Martin Cracauer <cracauer@xxxxxxxx> writes:

> Under certain conditions, a setf on a struct accessor does not check
> for the type of the assignment even under maximum safety.
> 
> (declaim (optimize (safety 3) (debug 3) (speed 0)))
> 
> (defstruct foo
>   (bar 0 :type fixnum))
> 
> (defun test (baz)
>   (setf (foo-bar baz) "hello, world")
>   baz)
> 
> (print (foo-bar (test (make-foo))))
> 
> ;; you get a compilation warning, but not a runtime error:

With CVS CMUCL, I get:

; In: defun test

;   (setf (foo-bar baz) "hello, world")
; --> let* multiple-value-bind let funcall c::%funcall 
; ==>
;   #:g0
; Warning: This is not a (values &optional fixnum &rest t):
;   "hello, world"
; 
;   (defun test (baz) (setf # "hello, world") baz)
; --> block 
; ==>
;   baz
; Note: Deleting unreachable code.
; 

; Compilation unit finished.
;   1 warning
;   1 note


; /tmp/defstruct.x86f written.
; Compilation finished in 0:00:01.

; Loading #p"/usr/tmp/defstruct.x86f".

Type-error in kernel::object-not-type-error-handler:
   "hello, world" is not of type fixnum
   [Condition of type type-error]

Seems this is caught by one of the changes I made that were triggered
by Paul Dietz' tests.