Re: [Sbcl-devel] defmacro and destructuring-bind Improvements

[Paul K. <pv...@pv...>]

Jan Moringen wrote:
> a few days ago, I started looking into bug 327537 [1] under the
> (much appreciated) guidance of pkhuong. With the implementation
> technique suggested by pkhuong, a sketch for compile time
> checks/optimizations of PROPER-LIST-OF-LENGTH-P as suggested in the
> bug report was relatively straight forward.
[...]
> The carefully designed part is that #:WHOLE13940 is only used once:
> in the check-proper-list-of-length call. This fact can be exploited
> by a deftransform for check-proper-list-of-length to conclude that
> WHOLE13940 is not mutated and that its derived type can be trusted.

I recently had a sad insight into this: what does it mean when a 
function like
   (lambda () (let ((x (list 1 2 3))) (values x x)))
is of type
  (function * (values (cons integer (cons integer (cons integer null))) 
cons &optional))?

Our current machinery, with lvar-conservative-type (used in CAR and CDR 
derive-type and to determine array dimensions), already seems wrong here.

Is such aliasing disallowed by the type system?

Paul Khuong