Re: Bad define placement?

[Mikael Djurfeldt <mdj@mdj-pc.nada.kth.se>]

"<Brad Knotwell" <knotwell@f5.com> writes:

> Along the same lines, code like this contrived example worked in
> guile1.3 but doesn't work now:
> 
> 	(define (test-fun x)
> 	  (define tt? (eq? x 3))
> 	  (define return-val
>              (if tt? #t #f))
> 	   return-val)

[...]

> Is this a bug related to the currently discussed one, a brand new
> bug or an incorrect method that took advantage of a previous bug in
> guile?

Well, let's see...  we'll first rewrite it accroding to R5RS:

(define (test-fun x)
  (letrec ((tt? (eq? x 3))
	   (return-val (if tt? #t #f)))
    return-val))

The bindings introduced by a `letrec' as a scope which includes the
definitions, but since all bindings are bound "in parallell", it is
not allowed to *use* any of the introduced bindings during the
definition.

So, the access of `tt?' in the `if' is a bug, and, yes, it was
undetected previously.