On Thu, May 5, 2011 at 4:08 AM, John Cowan <cowan@mercury.ccil.org> wrote:

Andre van Tonder scripsit:

>   (define-syntax my-syntax-rules
>     (syntax-rules ()
>       ((_ blah) (syntax-rules ..........)))
>
>   (define-syntax foo (my-syntax-rules ........)) ;; PHASE ERROR
>
> This kind of thing would just work in some Scheme implementations, but
> others would require MY-SYNTAX-RULES to be imported FOR EXPAND for the
> macro definition of FOO to work.

Is it clear that it's required to work at all?  IOW, is it clear that
macro calls are expanded in the body of a DEFINE-SYNTAX?  Given the
following:

   (define-syntax yow (syntax-rules () ((yow . bow) (syntax-rules . bow))))
   (define-syntax cow (yow () ((cow) 32)))

MIT Scheme, Guile, Kawa, SISC, Chibi, Chez, SCM, Ikarus, IronScheme,
Mosh all accept both lines and (cow) => 32.  But Racket, Gauche,
Chicken, Scheme48/scsh, Larceny, Ypsilon, STklos, Scheme 9 all complain
about bad syntax or undefined variables in the second line.

Ah, yes, the spec (R7-draft) already says the right hand side define-syntax, let-syntax, letrec-syntax must be "an instance of syntax-rules", and therefore may not be an invocation of a user-defined macro.  And syntax-rules also is not "powerful-enough" to invoke a user-defined macro during its expansion (as syntax-case is).      Therefore, there are no phase issues in WG1.