[bug #31445] Incorrect multiple-value truncation [1.9.13]

[bug #31445] Incorrect multiple-value truncation [1.9.13]

[Ludovic Courtès]

URL:
  <http://savannah.gnu.org/bugs/?31445>

                 Summary: Incorrect multiple-value truncation [1.9.13]
                 Project: Guile
            Submitted by: civodul
            Submitted on: Mon 25 Oct 2010 09:57:33 AM GMT
                Category: None
                Severity: 3 - Normal
              Item Group: None
                  Status: None
                 Privacy: Public
             Assigned to: None
             Open/Closed: Open
         Discussion Lock: Any

    _______________________________________________________

Details:

This is with `map' implemented in C:

#v+
scheme@(guile-user)> (map (lambda (x) (values x x)) (iota 2))
$5 = (#<values (0 0)> #<values (1 1)>)
scheme@(guile-user)> ,x map
Disassembly of #<procedure map (_ _ . _)>:

   0    (assert-nargs-ge 0 2)           
   3    (push-rest 0 2)                 
   6    (object-ref 0)                  ;; #<pointer 0x7f4ee0395d20>
   8    (subr-call 3)                   
#v-

Instead, the result should be '(0 1).

Conversely:

#v+
scheme@(guile-user)> (map (lambda (x) (identity (values x x))) (iota 2))
$6 = (0 1)
scheme@(guile-user)> ,x identity
Disassembly of #<procedure identity (x)>:

   0    (assert-nargs-ee/locals 1)      
   2    (local-ref 0)                   ;; `x'
   4    (return)                        
#v-





    _______________________________________________________

Reply to this item at:

  <http://savannah.gnu.org/bugs/?31445>

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

[bug #31445] Incorrect multiple-value truncation [1.9.13]

[Andy Wingo]

Follow-up Comment #1, bug #31445 (project guile):

This is not a regression, as the results would have been the same before 1.9.
But clearly it's not ideal either.

I propose we add new API, scm_call_0_v, scm_apply_3_v, etc. Such code would
truncate or raise an error for 0 values. Then we change all internal users to
use the new code, if appropriate.

WDYT?

    _______________________________________________________

Reply to this item at:

  <http://savannah.gnu.org/bugs/?31445>

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

[bug #31445] Incorrect multiple-value truncation [1.9.13]

[Ludovic Courtès]

Follow-up Comment #2, bug #31445 (project guile):

Hmm, maybe that's a complex solution for a situation that'll be less and less
common, especially when `map' & co. are rewritten in Scheme.

WDYT?

    _______________________________________________________

Reply to this item at:

  <http://savannah.gnu.org/bugs/?31445>

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

[bug #31445] Incorrect multiple-value truncation [1.9.13]

[Andy Wingo]

Follow-up Comment #3, bug #31445 (project guile):

I could go either way. I think using a single-valued, truncating API is the
right thing to do though. Grep your tree for scm_call_ and scm_apply to see
the size of the set of current use cases.

    _______________________________________________________

Reply to this item at:

  <http://savannah.gnu.org/bugs/?31445>

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