bug#18583: possibly buggy top level behavior in master

[Andy Wingo]

On Mon 29 Sep 2014 15:24, Daniel Llorens <address@hidden> writes:

> The following code:
>
> (define-syntax define-lues
>   (syntax-rules ()
>     ((_ () exp)
>      (call-with-values (lambda () exp) (lambda () 'unspecified)))
>     ((_ (var . vars) exp)
>      (begin
>        (define var (call-with-values (lambda () exp) list))
>        (define-lues vars (apply values (cdr var)))
>        (define var (car var))))
>     ((_ var exp)
>      (define var (call-with-values (lambda () exp) list)))))
>
> (define-lues (a) (values (make-vector 3 9)))
> (vector-length a)
>
> In 2.0, either pasting this in the REPL or (load "...") a file
> containing this produces the expected result, 3.
>
> In the current master (f7582), pasting in the REPL produces the
> same result, but loading from a file this produces the error:
>
>> In procedure vector-ref: Wrong type argument in position 1 (expecting 
>> vector): (#(9 9 9))
>
> I'm aware of the workarounds, I'm reporting the behavior itself.

Thank you thank you thank you for this report!

Incidentally it has a nice reduction, at the REPL or whereever:

  (begin
    (define x (list 1))
    (define x (car x))
    x)

Should be 1, Guile master was giving (1).  The reasons are a bug in the
effects analysis for the `define!' primitive.  I wonder how I didn't
catch this before!  Fixed in master, though sadly with a bytecode
version bump.

Andy