Inconsistent style-warning for addition of null-adjacent values

[Patrick Poitras]

In trying to reproduce a bug (backstory in comment below) I came across this bug which yields inconsistent behavior when typechecking.

CL-USER> (defun blarg1 (value)
           (+ 2 (car value)))
BLARG1

CL-USER> (defun blarg2 (value)
           (+ 2 (when value
                  (car value))))
; in: DEFUN BLARG2
; (WHEN VALUE (CAR VALUE))
; ==>
; NIL
;
; caught STYLE-WARNING:
; This is not a NUMBER:
; NIL
; See also:
; The SBCL Manual, Node "Handling of Types"
;
; compilation unit finished
; caught 1 STYLE-WARNING condition
BLARG2

The functions compile and execute correctly:

(blarg1 (list 5)) and (blarg2 (list 5)) both return 7.
(blarg1 nil) and (blarg2 nil) return a type-error.

The assembly emitted for these functions differs.

Reproduced on SBCL 2.3.3, SBCL 2.4.3 and SBCL 2.4.4.46-c6713be25

uname -a
Linux ppoitras-desktop-blarg 5.15.0-76-generic #83-Ubuntu SMP Thu Jun 15 19:16:32 UTC 2023 x86_64 x86_64 x86_64 GNU/Linux

Reproduced with *features*

(:CLPM-CLIENT :ASDF3.3 :ASDF3.2 :ASDF3.1 :ASDF3 :ASDF2 :ASDF :OS-UNIX
 :NON-BASE-CHARS-EXIST-P :ASDF-UNICODE :ARENA-ALLOCATOR :X86-64 :GENCGC :64-BIT
 :ANSI-CL :COMMON-LISP :ELF :IEEE-FLOATING-POINT :LINUX :LITTLE-ENDIAN
 :PACKAGE-LOCAL-NICKNAMES :SB-LDB :SB-PACKAGE-LOCKS :SB-THREAD :SB-UNICODE
 :SBCL :UNIX)

and also

(:SWANK ...same features as above...)

== Note ==

This is not actively causing us issues. I'd rank it as low-priority, but I reported it anyway in the event it's a useful case to figure out weird behavior in the type checking.

[Patrick Poitras (pfpoitras)]

== Unimportant context ==
The context for this bug was that

(defun blarg0 (plist)
    (destructuring-bind (&key a b c)
        plist
      (+ a b c))

used to return a style-warning under 2.3.3 and now no longer does. This difference of behavior caused me to investigate, but the cause of this particular difference in behavior is that the destructuring-bind macro now uses getf rather than sb-c::ds-getf. The latter has a structure that ressembles blarg2, and which throws the style-warning.

[Stas Boukarev (stassats)]

There's no bug, it does what is intended.