From: Szymon
Subject: problem with (DECLARE (IGNORE ...)) in CMUCL.
Date: Wed, 27 Oct 2004 14:18:06 +0000
Message-ID: <87d5z4nt29.fsf@eva.rplacd.net>

Hi.

In some cases (example below) (DECLARE (IGNORE ...))
seems to be 'inactive'.

(funcall
  (compile
   nil
   '(lambda ()
      (destructuring-bind (x x x x a x b . x) (list 1 2 3 5 6 7 8 9 10)
        (declare (ignore x))
        (list a b)))))

result:

[.....]

;   (DESTRUCTURING-BIND
;       (X X X X A ...)
;       (LIST 1 2 3 5 ...)
;     (DECLARE #)
;     (LIST A B))
; --> LET LET* 
; ==>
;   (LET* (# # # # # ...)
;     (DECLARE #)
;     (LIST A B))
; Note: Variable X defined but never used.
; ; [Last message occurs 5 times]

[.....]

(6 8)


Any ideas how to fix this ? (CLISP, for example, can handle this case).

Regards, Szymon.

• Re: problem with (DECLARE (IGNORE ...)) in CMUCL. Tim Bradshaw
  • Re: problem with (DECLARE (IGNORE ...)) in CMUCL. Szymon

From: Tim Bradshaw
Subject: Re: problem with (DECLARE (IGNORE ...)) in CMUCL.
Date: Wed, 27 Oct 2004 22:24:44 +0000
Message-ID: <ey3brenx0ib.fsf@cley.com>

* Szymon  wrote:
> Hi.

> In some cases (example below) (DECLARE (IGNORE ...))
> seems to be 'inactive'.

> [...]

> Any ideas how to fix this ? (CLISP, for example, can handle this case).

Well, I think this sort of thing is probably not very well defined.
What you're doing is more-or-less equivalent to something like

(let ((x 1)
      (x 2))
  (declare (ignore x))
  ...)

or perhaps

(let* ((x 1)
       (x 2))
  (declare (ignore x))
  ...)

And I'm not at all clear that the IGNORE declaration *should* affect
all the bindings of X.  I'd *want* a warning in these cases!

--tim

From: Szymon
Subject: Re: problem with (DECLARE (IGNORE ...)) in CMUCL.
Date: Fri, 29 Oct 2004 22:34:51 +0000
Message-ID: <87ekjhdugk.fsf@eva.rplacd.net>

Tim Bradshaw <···@cley.com> writes:

> * Szymon  wrote:
> > Hi.
> 
> > In some cases (example below) (DECLARE (IGNORE ...))
> > seems to be 'inactive'.
> 
> > [...]
> 
> > Any ideas how to fix this ? (CLISP, for example, can handle this case).
> 
> Well, I think this sort of thing is probably not very well defined.
> What you're doing is more-or-less equivalent to something like
> 
> (let ((x 1)
>       (x 2))
>   (declare (ignore x))
>   ...)

This form in not valid (?) (more-or-less eql to (LAMBDA (X X))).

> or perhaps
> 
> (let* ((x 1)
>        (x 2))
>   (declare (ignore x))
>   ...)
> 
> And I'm not at all clear that the IGNORE declaration *should* affect
> all the bindings of X.  I'd *want* a warning in these cases!

Ok. If you want warning -- I want it too :)

Regards, Szymon.