scrutinizer makes too strong assumptions about mutable data

[felix]

Like this:

(define foo (make-vector 1 #f))
(vector-set! foo 0 'sym)
(print (get (vector-ref foo 0) 'a))

When compiled in block mode, we get:

Warning: at toplevel:
  (x.scm:5) in procedure call to `get', expected argument #1 of type `symbol', but was given an argument of type `boolean'

[felix]

The entries for make-vector and cons, list and all other constructors probably have to be weakened in types.db - the returned data may be mutated and so can change type.

[felix]

Replying to felix:

The entries for make-vector and cons, list and all other constructors probably have to be weakened in types.db - the returned data may be mutated and so can change type.

That's nonsense, actually. The real problem is that smashing happens only for env + blist entries, not for globals.

[felix]

Pushed to master in ae6af2c1894b8ab8178f21f8169a98c3b5e0327a.

[felix]

Milestone 4.8.0 deleted