[bug #29817] `define*' procedures get invalid source info [1.9.10]

[bug #29817] `define*' procedures get invalid source info [1.9.10]

[Ludovic Courtès]

URL:
  <http://savannah.gnu.org/bugs/?29817>

                 Summary: `define*' procedures get invalid source info
[1.9.10]
                 Project: Guile
            Submitted by: civodul
            Submitted on: Fri 07 May 2010 09:48:31 AM GMT
                Category: None
                Severity: 3 - Normal
              Item Group: None
                  Status: None
                 Privacy: Public
             Assigned to: None
             Open/Closed: Open
         Discussion Lock: Any

    _______________________________________________________

Details:

Consider this procedure of one optional argument:

#v+
scheme@(guile-user)> (define* (foo #:optional y) (baz))
scheme@(guile-user)> (program-sources foo)
$25 = ((25 #f 64 . 28))
scheme@(guile-user)> ,x foo
Disassembly of #<procedure foo (#:optional y)>:

   0    (assert-nargs-ge 0 0)           
   3    (bind-optionals 0 1)            
   6    (assert-nargs-ee 0 1)           
   9    (reserve-locals 0 1)            
  12    (local-bound? 0)                
  14    (br-if :L730)                   ;; -> 21
  18    (make-false)                    
  19    (local-set 0)                   ;; `y'
  21    (toplevel-ref 1)                ;; `baz'
  23    (tail-call 0)                                         at (unknown
file):64:28
#v-

Its program source info points at IP 25 whereas its last IP is 23.  In
addition, it has no info associated with IP 0, which single-arity procedures
always have AFAICS.

The latter is solved when using `lambda*', but not the former:

#v+
scheme@(guile-user)> (define foo (lambda* (#:optional y) (baz)))
scheme@(guile-user)> (program-sources foo)
$26 = ((0 #f 67 . 12) (25 #f 67 . 36))
scheme@(guile-user)> ,x foo
Disassembly of #<procedure foo (#:optional y)>:

   0    (assert-nargs-ge 0 0)           
   3    (bind-optionals 0 1)            
   6    (assert-nargs-ee 0 1)           
   9    (reserve-locals 0 1)            
  12    (local-bound? 0)                
  14    (br-if :L757)                   ;; -> 21
  18    (make-false)                    
  19    (local-set 0)                   ;; `y'
  21    (toplevel-ref 1)                ;; `baz'
  23    (tail-call 0)                                         at (unknown
file):67:36
#v-

Finally, using `case-lambda' leads to different code, with more source info,
but again with one IP off (19 instead of 17):

#v+
scheme@(guile-user)> (define foo (case-lambda (() (baz)) ((y) (baz))))
scheme@(guile-user)> (program-sources foo)
$27 = ((0 #f 70 . 12) (13 #f 70 . 29) (15 #f 70 . 12) (19 #f 70 . 41))
scheme@(guile-user)> ,x foo
Disassembly of #<procedure foo () | (y)>:

   0    (br-if-nargs-ne 0 0 :L785)      ;; -> 13
   6    (reserve-locals 0 0)            
   9    (toplevel-ref 1)                ;; `baz'
  11    (tail-call 0)                                         at (unknown
file):70:29
  13    (assert-nargs-ee/locals 1)                            at (unknown
file):70:12
  15    (toplevel-ref 1)                ;; `baz'
  17    (tail-call 0)                                         at (unknown
file):70:41
#v-

Ludo'.




    _______________________________________________________

Reply to this item at:

  <http://savannah.gnu.org/bugs/?29817>

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

Re: [bug #29817] `define*' procedures get invalid source info [1.9.10]

[Andy Wingo]

On Fri 07 May 2010 11:48, Ludovic Courtès <INVALID.NOREPLY@gnu.org> writes:

>                  Summary: `define*' procedures get invalid source info

Interesting bug! Added to my list. Should not be difficult to fix.

Andy
-- 
http://wingolog.org/

[bug #29817] `define*' procedures get invalid source info [1.9.10]

[Andy Wingo]

Update of bug #29817 (project guile):

                  Status:                    None => Fixed                  
             Open/Closed:                    Open => Closed                 

    _______________________________________________________

Follow-up Comment #1:

Source location with IP 25 is not actually invalid, as IP 23 is actually the
IP just before the tail-call; 25 is the IP afterwards, where source location
info is associated. That part is correct, giving the right location for the
procedure call.

The fact that there is no source location associated with the definition was
a bug in which output introduced by a macro would have no source location
information. Logically it actually has at least two source locations: the
location of the macro definition, and that of the use.

Usually the use location is more useful, which is fortunate as we don't
actually have a way to record locations in macro definitions (due to lack of
needed fields in syntax objects).

So, the solution was to associate the source location of use with output
introduced by the macro (i.e. define*). This has been checked into master.

    _______________________________________________________

Reply to this item at:

  <http://savannah.gnu.org/bugs/?29817>

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