[bug #34102] Incorrect procedure arities with Guile 2.x if no compilation

[bug #34102] Incorrect procedure arities with Guile 2.x if no compilation

[Patrick Bernaud]

URL:
  <http://savannah.gnu.org/bugs/?34102>

                 Summary: Incorrect procedure arities with Guile 2.x if no
compilation 
                 Project: Guile
            Submitted by: patb
            Submitted on: Tue 23 Aug 2011 12:33:23 PM CEST
                Category: None
                Severity: 3 - Normal
              Item Group: None
                  Status: None
                 Privacy: Public
             Assigned to: None
             Open/Closed: Open
         Discussion Lock: Any

    _______________________________________________________

Details:

If Guile is run without auto-compilation, the arities of the procedures having
a rest or an optional parameter are wrong.

With the test script attached:
$ guile -v | head -1
guile (GNU Guile) 2.0.2.50-b8287
$ guile --fresh-auto-compile --no-auto-compile -s test.scm
(2 0 #f)
(0 0 #t)
(0 0 #t)
$ guile --auto-compile -s test.scm
(2 0 #f)
(2 0 #t)
(2 1 #f)




    _______________________________________________________

File Attachments:


-------------------------------------------------------
Date: Tue 23 Aug 2011 12:33:23 PM CEST  Name: test.scm  Size: 267B   By: patb

<http://savannah.gnu.org/bugs/download.php?file_id=23852>

    _______________________________________________________

Reply to this item at:

  <http://savannah.gnu.org/bugs/?34102>

_______________________________________________
  Message sent via/by Savannah
  http://savannah.gnu.org/

[bug #34102] Incorrect procedure arities with Guile 2.x if no compilation

[Stefan Israelsson Tampe]

Follow-up Comment #1, bug #34102 (project guile):

Note,

scheme@(guile-user)> (compile '(define (f x . l) x) #:env (current-module))

scheme@(guile-user)> (procedure-minimum-arity f)
$1 = (1 0 #t)

scheme@(guile-user)> (eval '(define (f x . l) x) (current-module))
$2 = #<variable 27320a0 value: #<procedure 24f37e0 at ice-9/eval.scm:238:6
%args>>

scheme@(guile-user)> (procedure-minimum-arity f)
$3 = (0 0 #t)

So there must be a bug in the eval code!

Diving into eval.scm one can see that the more advanced
lambdas will be translated into (lambda %args ...) e.g.

(define (make-general-closure env body nreq rest? nopt kw inits alt)
      (define alt-proc
        (and alt
             (let* ((body (car alt))
                    (nreq (cadr alt))
                    (rest (if (null? (cddr alt)) #f (caddr alt)))
                    (tail (and (pair? (cddr alt)) (pair? (cdddr alt)) (cdddr
alt)))
                    (nopt (if tail (car tail) 0))
                    (kw (and tail (cadr tail)))
                    (inits (if tail (caddr tail) '()))
                    (alt (and tail (cadddr tail))))
               (make-general-closure env body nreq rest nopt kw inits alt))))
      (lambda %args
          (let lp ((env env)
                 (nreq* nreq)
                 (args %args)) ...


So in alles e.g. (procedure-minimum-arity test2) will see
(lambda x ...) signature and issue the resulting faulty signature
information.

A solution would be to keep a weak hashtable from lambdas to signature
information. The the above function procedure-mi... could first ask that map
for a signature information and if that fails use the old method assuming that
make-general-closure will
populate the weak map correctly.

/Stefan

    _______________________________________________________

Reply to this item at:

  <http://savannah.gnu.org/bugs/?34102>

_______________________________________________
  Message sent via/by Savannah
  http://savannah.gnu.org/

[bug #34102] improve debuggability of interpreted procedures (procedure-minimum-arity)

[Andy Wingo]

Update of bug #34102 (project guile):

                 Summary: Incorrect procedure arities with Guile 2.x if no
compilation  => improve debuggability of interpreted procedures
(procedure-minimum-arity)

    _______________________________________________________

Follow-up Comment #2:

This is a known issue, but the arities are not incorrect, strictly speaking. 
The procedure is called "minimum-arity", after all -- it does not claim to be
the most exact arity.

So, this is not a bug.  Strictly speaking it is an enhancement request, to
improve the debuggability of interpreted procedures.  I am changing the title
to reflect that.

    _______________________________________________________

Reply to this item at:

  <http://savannah.gnu.org/bugs/?34102>

_______________________________________________
  Message sent via/by Savannah
  http://savannah.gnu.org/

[bug #34102] improve debuggability of interpreted procedures (procedure-minimum-arity)

[Patrick Bernaud]

Follow-up Comment #3, bug #34102 (project guile):

A procedure of n arguments and a rest has to be passed at least n arguments.
Yet this lowest limit can not be expressed by the rest boolean returned by
'procedure-minimum-arity'.


    _______________________________________________________

Reply to this item at:

  <http://savannah.gnu.org/bugs/?34102>

_______________________________________________
  Message sent via/by Savannah
  http://savannah.gnu.org/

[bug #34102] improve debuggability of interpreted procedures (procedure-minimum-arity)

[Stefan Israelsson Tampe]

Follow-up Comment #4, bug #34102 (project guile):

With the following applied git-diff (arity.diff) we get:

~/stis/src/guile/meta/guile --fresh-auto-compile --no-auto-compile -s
arity.scm
(2 0 #f)
(2 0 #t)
(2 1 #f)

And the bug/feature request is fixed!

/Stefan

(file #23913)
    _______________________________________________________

Additional Item Attachment:

File name: arity.diff                     Size:3 KB


    _______________________________________________________

Reply to this item at:

  <http://savannah.gnu.org/bugs/?34102>

_______________________________________________
  Message sent via/by Savannah
  http://savannah.gnu.org/

bug#10059: [bug #34102] improve debuggability of interpreted procedures (procedure-minimum-arity)

[Andy Wingo]

On Wed 31 Aug 2011 14:29, Stefan Israelsson Tampe <INVALID.NOREPLY@gnu.org> writes:

> Follow-up Comment #4, bug #34102 (project guile):
>
> With the following applied git-diff (arity.diff) we get:
>
> ~/stis/src/guile/meta/guile --fresh-auto-compile --no-auto-compile -s
> arity.scm
> (2 0 #f)
> (2 0 #t)
> (2 1 #f)
>
> And the bug/feature request is fixed!
>
> /Stefan

Thanks for the patch, Stefan!  I reworked it a bit and applied it.  I
also made it so that procedures defined at the toplevel in `eval' get
their names set appropriately.

I am having trouble with savannah right now, so I am not yet able to
close the bug.

Andy
-- 
http://wingolog.org/

bug#10059: fixed

[Andy Wingo]

Fixed, whenever I can push this to sv (having connectivity issues now).

Andy
-- 
http://wingolog.org/