Re: [Gcl-devel] ANSI compliance errors with CONCATENATE

[Camm Maguire]

Greetings!  I'm checking in a fix for these as well.  Please verify.

Regarding the last, I may need some advice.  The compiler uses
(make-sequence 'vector ...), which gets promoted to (make-sequence
'(array * (*)) ...) by normalize type.  So this has got to work unless
we want to make signiicant changes.  This is interpreted as a simple
one-dimensional array of objects.  The spec, though, says that
(make-sequence '(vector * (3)) ...) should throw an error, as makes
sense.  So basically I just triggered an error if the size of an array
type was specified:

               (unless (and (or (eq (car type) 'array)
                                (eq (car type) 'simple-array))
                            (not (si::fixnump (car (caddr type)))))
                       (error "~S is not a sequence type." type))


Take care,


"Paul F. Dietz" <address@hidden> writes:

> The following forms are evaluated improperly in gcl
> compiled with --enable-ansi:
> 
> (concatenate 'cons '(a b c d e))
>    ;; should yield (a b c d e), but instead throws an error.
>    ;; (CONS is a subtype of LIST, so this should be legal.)
> 
> (concatenate 'null () ())
>    ;; should yield NIL (since NIL is in type NULL and NULL
>    ;; is a subtype of LIST) but instead throws an error.
> 
> (concatenate '(vector * 3) '(a b c d e))
>    ;; should throw an error (since the number of elements is
>    ;; incompatible) but instead returns (a b c d e).
> 
>       Paul
> 
> 
> _______________________________________________
> Gcl-devel mailing list
> address@hidden
> http://mail.gnu.org/mailman/listinfo/gcl-devel
> 
> 

-- 
Camm Maguire                                            address@hidden
==========================================================================
"The earth is but one country, and mankind its citizens."  --  Baha'u'llah