[r6rs] expansion bug, probably free-identifier=? related

[r6rs] expansion bug, probably free-identifier=? related

[Marco Maggi]

The following little package:

<http://github.com/downloads/marcomaggi/nausicaa/synrec6-1.0.1.tar.gz>

works with  Larceny, Mosh,  Petite Chez, Racket,  Vicare and
Ypsilon; with Guile 2.0 it fails with:

GUILE_LOAD_PATH=.: guile -l guile-r6rs-setup.scm --auto-compile -s ./test-flatt-syntactic.sps; GUILE_LOAD_PATH=.: guile -l guile-r6rs-setup.scm --auto-compile -s ./test-synrec6.sps;
;;; note: auto-compilation is enabled, set GUILE_AUTO_COMPILE=0
;;;       or pass the --no-auto-compile argument to disable.
;;; compiling /home/marco/src/devel/scheme/icaro/src/standalone/synrec6/./test-flatt-syntactic.sps
;;; WARNING: compilation of /home/marco/src/devel/scheme/icaro/src/standalone/synrec6/./test-flatt-syntactic.sps failed:
;;; key r6rs:exception, throw_args (#<r6rs:record:&raise-object-wrapper>)
Backtrace:
In module/ice-9/boot-9.scm:
 170: 13 [catch #t #<catch-closure 822a5d0> ...]
In unknown file:
   ?: 12 [catch-closure]
In module/ice-9/boot-9.scm:
  62: 11 [call-with-prompt prompt0 ...]
In module/ice-9/eval.scm:
 389: 10 [eval # #]
In module/ice-9/boot-9.scm:
1917: 9 [save-module-excursion #<procedure 80a2320 at module/ice-9/boot-9.scm:3341:3 ()>]
3348: 8 [#<procedure 80a2320 at module/ice-9/boot-9.scm:3341:3 ()>]
1189: 7 [%start-stack load-stack ...]
1194: 6 [#<procedure 8723ac8 ()>]
In unknown file:
   ?: 5 [primitive-load "/home/marco/src/devel/scheme/icaro/src/standalone/synrec6/./test-flatt-syntactic.sps"]
In module/ice-9/eval.scm:
 458: 4 [#<procedure 813b680 at module/ice-9/eval.scm:452:4 (exp)> #]
In module/ice-9/psyntax.scm:
1066: 3 [chi-top (six.define-record-type # # #) () (#) ...]
1007: 2 [syntax-type (six.define-record-type # # #) () (#) ...]
1303: 1 [chi-macro #<procedure 8704ab0 (stx)> (six.define-record-type # # #) () ...]
In unknown file:
   ?: 0 [#<procedure 8704ab0 (stx)> #(syntax-object # # #)]

ERROR: In procedure #<procedure 8704ab0 (stx)>:
ERROR: ERROR: R6RS exception:
  1. &who: define-record-type
  2. &message: "invalid clause in record type definition"
  3. &syntax:
      form: (six.define-record-type (point make-point point?) (six.fields (six.immutable x point-x) (six.mutable y point-y set-point-y!)) (six.nongenerative point-4893d957-e00b-11d9-817f-00111175eb9e))
      subform: (six.fields (six.immutable x point-x) (six.mutable y point-y set-point-y!))

Some deprecated features have been used.  Set the environment
variable GUILE_WARN_DEPRECATED to "detailed" and rerun the
program to get more information.  Set it to "no" to suppress
this message.
-- 
Marco Maggi

Re: [r6rs] expansion bug, probably free-identifier=? related

[Andy Wingo]

Indeed, it seems free-id=? should look up the identifiers in their
modules, and compare the bindings (variables), if present, for
eq?-ness.

Andy
-- 
http://wingolog.org/

Re: [r6rs] expansion bug, probably free-identifier=? related

[Andy Wingo]

Hello Marco,

I have a branch pending to fix free-identifier=? in Guile.  It will
probably go in 2.0.2.

However, even with this patch, test-flatt-syntactic.sps still fails
because of two things, one difficult and the other even more so:

  1) Bug https://savannah.gnu.org/bugs/?31472 -- the introduced
     "the-parent-rtd" identifier gets resolved relative to the (synrec6)
     module rather than the module of the expanded form.  This will be
     fixed at some point.

  2) Guile does not hygienically introduce toplevel bindings.  I would
     like to fix it, but it will not be fixed in 2.0.x, as it would
     break ABI.  You can work around this one as Guile's own srfi-9
     does, if you so wish, by generating a temporary yourself.

Regards,

Andy
-- 
http://wingolog.org/