bug#43601: 27.1; Macroexpansion bug in `push'

[Phil Sainty]

On 2020-09-25 10:54, Sean Devlin wrote:
> Hi folks,
>
> I believe I've found a macroexpansion bug in the `push' macro. Open a
> new instance of Emacs and evaluate the following form in the scratch
> buffer:
>
>     (macroexpand '(push (list 'x)
>                         (cdr my-list)))
>
> The result (with some reformatting) is this:
>
>     (let* ((v (list 'x))
>            (v my-list))
>       (setcdr v
>               (cons v
>                    (cdr v))))
>
> Both values are bound to `v', so the former is shadowed by the latter.

It's actually fine.

What you're seeing is the *printed representation* of the lisp objects,
and two completely independent symbols, each with the name "v".

(setq foo (macroexpand '(push (list 'x) (cdr my-list))))
=> (let* ((v (list 'x)) (v my-list)) (setcdr v (cons v (cdr v))))

(setq vfirst (caaadr foo))
=> v

(setq vsecond (car (cadadr foo)))
=> v

(eq vfirst vsecond)
=> nil

(cdaddr (caddr foo))
=> (v (cdr v))

(eq vfirst (car (cdaddr (caddr foo))))
=> t

(eq vsecond (cadadr (cdaddr (caddr foo))))
=> t


This is like:

(eq (make-symbol "v") (make-symbol "v"))
=> nil