> Now it is a mystery to me why I got different results from Chibi and
> Sagittarius when I tried the example yesterday.

Sagittarius has 2 implementation of syntax-rules, one is for (rnrs base)
and the other one is for (scheme base). If you start REPL without -r7
command line option, then it imports (rnrs) to the REPL environment.
And if you do pass the -r7 option, then it imports (scheme base).

OK, now suppose we have the following 2 files:

;; syn-match-r6.scm
(import (rnrs))

(define-syntax foo
  (syntax-rules ()
    ((foo bar x)
     (define-syntax bar
       (syntax-rules (x)
     ((bar c) 'matched)
     ((bar z) 'unmatched))))))

(foo bar c)
(display (bar q)) (newline)

;; syn-match-r7.scm
(import (scheme base) (scheme write))

(define-syntax foo
  (syntax-rules ()
    ((foo bar x)
     (define-syntax bar
       (syntax-rules (x)
     ((bar c) 'matched)
     ((bar z) 'unmatched))))))

(foo bar c)
(display (bar q)) (newline)

Both are identical except the import clauses. And the following is the
result of execution

$ sash -r6 syn-match-r6.scm
matched

$ sash -r7 syn-match-r7.scm
unmatched

The R6RS version of syntax-rules returns matched the same as other R6RS
implementations (Tested on Chez and Racket's plt-r6rs).

The R7RS version returns unmatched and Chibi also returned the same
on my environment. (NB: The R7RS version is based on Chibi's syntax-rules,
so it's not surprising to me)

Interestingly, Gauche with -r7 option returned matched. So not all R7RS
implementations agreed to return unmatched.

Cheers,

--
_/_/
Takashi Kato
E-mail: xxxxxx@gmail.com