Re: [Sbcl-devel] sbcl bug: behaviour of THE VALUES

[Christophe R. <cs...@ca...>]

Bruno Haible <br...@cl...> writes:

> According to the ANSI CL description of the VALUES type specifier - which is
> flaky regarding &KEY and &OPTIONAL, but unambiguous regarding &REST -
>
>   (the (values integer &rest list) (values 3 4 5))
>
> should succeed and
>
>   (the (values integer &rest integer) (values 3 4 5))
>
> should fail (because the list (4 5) is not an integer). In SBCL 0.8.16 it's
> just the other way around:
>
>   (the (values integer &rest list) (values 3 4 5)) => ERROR
>   (the (values integer &rest integer) (values 3 4 5)) => 3; 4; 5

Take a look at the Compound Type Specifier Description for the
FUNCTION type, and then tell me it's not ambiguous ;-)

Specifically, although I think your reading is arguable from the
VALUES page in isolation, I don't think it holds up given the language
about &rest on the FUNCTION page -- even though the FUNCTION page
talks about argument types not return types -- because of the
description of the VALUES type in terms of MULTIPLE-VALUE-CALL.

Cheers,

Christophe