From chohag at jtan.com  Sun Sep  1 09:45:31 2024
From: chohag at jtan.com (chohag at jtan.com)
Date: Sun, 01 Sep 2024 17:45:31 +0100
Subject: [CM] S7 C function arguments corrupted in loop
Message-ID: <202409011645.481GjXFR418130@zeus.jtan.com>

After a few rounds of golf I have arrived at this minimal demonstration
of a problem where the arguments that a C function receives are not
the arguments it is given.

Originally the *wtf* let also declared an integer to look up but
this was removed by the golf. If it were still there then only the
*wtf* let and the final (map) would be necessary to see the problem
occur. Everything else is here to demonstrate the problem not
occurring.

I have tested with today's s7 tarball.

(require cload.scm)

; This is broken:
(define *wtf* (let ()
        (c-define `(
                        (C-function ("anything" _anything "..." 2))
                        (in-C "static s7_pointer
                        _anything (s7_scheme *sc, s7_pointer args)
                        {
                                printf(\"%s\", s7_object_to_c_string(sc, args));
                                return s7_nil(sc);
                        }"))
                "" () "" "" "libs7crash")
        (curlet)))

; This is for demonstration:
(define *other*
        (let ((one 1))
                (define (anything . AB)
                        (format #t "~A" AB)
                        ())
                (curlet)))


; Works fine if looping with for-each
(for-each (lambda (N)
        ((*wtf* 'anything)
                "any string"
                (logior 42 (*other* 'one))))
        '(ok fail))
(newline)
; -> ("any string" 43)("any string" 43)


; Works fine if not calling a C function
(map (lambda (N)
        ((*other* 'anything)
                "any string"
                (logior 42 (*other* 'one))))
        '(ok fail))
(newline)
; -> ("any string" 43)("any string" 43)


; This combination fails:
;   call a C function found in a let
;   have one of its arguments be a simple call to logior (or some others)
;   have one of logior's arguments be looked up in a (any) let.
;   loop at least twice
;
; The first argument to the C function will be replaced by the first
; argument to logior on the second and subsequent loop iterations
(map (lambda (N)
        ((*wtf* 'anything)
                "any string"
                (logior 42 (*other* 'one))))
        '(ok fail))
(newline)
; -> ("any string" 43)(42 43)

Matthew




From bil at ccrma.Stanford.EDU  Sun Sep  1 13:31:15 2024
From: bil at ccrma.Stanford.EDU (bil at ccrma.Stanford.EDU)
Date: Sun, 01 Sep 2024 13:31:15 -0700
Subject: [CM] S7 C function arguments corrupted in loop
In-Reply-To: <202409011645.481GjXFR418130@zeus.jtan.com>
References: <202409011645.481GjXFR418130@zeus.jtan.com>
Message-ID: <d350c8aff91b8f314dc79a3c14b438c2@ccrma.stanford.edu>

Thanks very much for the bug report!  That was a tricky one -- I managed
to overoptimize op_x_aa, using a constant list where it got stepped on.
A quick fix, around line 90626, change

   sc->value = c_function_call(f)(sc, with_list_t2(fx_call(sc, 
cdr(code)), fx_call(sc, cddr(code))));

to

   sc->value = c_function_call(f)(sc, list_2(sc, fx_call(sc, cdr(code)), 
fx_call(sc, cddr(code))));

I need to figure out what the "right thing" is here.

By the way, it had never occurred to me to use c-define in the midst of 
an expression like that --
kinda neat!  I think I'll include something like that in the docs.

Thanks again.



From bil at ccrma.Stanford.EDU  Mon Sep  9 13:47:28 2024
From: bil at ccrma.Stanford.EDU (bil at ccrma.Stanford.EDU)
Date: Mon, 09 Sep 2024 13:47:28 -0700
Subject: [CM] Snd 24.7
Message-ID: <2172e9a79f4fe85ce09c7e393ba57aed@ccrma.stanford.edu>

Snd 24.7:

s7: added complex-vectors. These are directly compatible with 
gsl_vector_complex.

   make-iterator's optional third argument can be #t which tells s7
   to choose the carrier based on the iterated sequence.

   s7_immutable has been deprecated, replaced by s7_set_immutable which
   has the necessary s7_scheme* argument.

   added s7_define_expansion.

checked: sbcl 2.4.8

Thanks!: Matthew