ffloor-derived-type-optimizer

gerd.moellmann@xxxxxxxxxxx (Gerd Moellmann) writes:

>   (defun foo (n d)
>     (declare (type (unsigned-byte 32) n d))
>     (ffloor n d))

CL-USER ==> (trace c::ffloor-derive-type-optimizer)
CL-USER ==> (cl "/tmp/ffloor" :byte-compile t)

; Python version 1.1, VM version Intel x86 on 25 AUG 03 07:19:47 pm.
; Compiling: /usr/tmp/ffloor.lisp 25 AUG 03 04:01:23 pm


  0: (c::ffloor-derive-type-optimizer
      #<c::combination #x482AA715
          fun= #<c::ref #x482AA6E5
                   leaf= #<c::defined-function #x482AA675
                             name= ffloor
                             inlinep= :inline>>
          args= (#<c::ref #x482AA78D
                     leaf= #<c::lambda-var #x482AA20D
                               name= n
                               type= #1=#<numeric-type #>>>
                 #<c::ref #x482AA805
                     leaf= #<c::lambda-var #x482AA25D  name= d  type= #1#>>)>)
  0: c::ffloor-derive-type-optimizer returned
       #<values-type (values (rational 0) (mod 4294967295))>

CL-USER ==> (in-package :c)
#<The C package, 3817/6255 internal, 310/390 external>
COMPILER ==> (values-type-intersection
 (values-specifier-type '(values (rational 0) (mod 4294967295)))
 (values-specifier-type '(values float real)))
#<named-type nil>

That's why Python thinks FOO's return type is NIL.  Compiling the file
natively, C::FFLOOR-DERIVE-TYPE-OPTIMIZER is not called, which
probably explains the difference in behavior.

Someone knowing that part of the code intimately enough to spot the
bug immediately?