Re: [Chicken-janitors] #1041: -d0 causes different application behavior

[Chicken Trac]

#1041: -d0 causes different application behavior
-----------------------+----------------------
  Reporter:  mario     |      Owner:
      Type:  defect    |     Status:  reopened
  Priority:  critical  |  Milestone:  4.11.0
 Component:  unknown   |    Version:  4.8.x
Resolution:            |   Keywords:
-----------------------+----------------------

Comment (by sjamaan):

 OK, here we go:

 {{{
 #;1> (use awful-path-matchers srfi-69)
 #;2> (##sys#peek-fixnum (path-match "foo" <int>) 0)
 140289424854051
 #;3> (##sys#size (path-match "foo" <int>))
 3
 #;4> (##sys#slot (path-match "foo" <int>) 0)
 70144712427025
 #;5> (##sys#slot (path-match "foo" <int>) 1)
 ("foo" #<procedure (awful-path-matchers#<int> thing17)>)
 #;6> (##sys#slot (path-match "foo" <int>) 2)
 #<lambda info (f_525 path90)>
 #;7> (define equiv? (hash-table-equivalence-function (make-hash-table)))
 #;8> (equiv? (path-match "foo" <int> <int>) (path-match "foo" <int>))
 #f
 }}}

 Without lambda info:

 {{{
 #;1> (use awful-path-matchers srfi-69)
 #;2> (##sys#peek-fixnum (path-match "foo" <int>) 0)
 140155045890064
 #;3> (##sys#size (path-match "foo" <int>))
 2
 #;4> (##sys#slot (path-match "foo" <int>) 0)
 70178053519952
 #;5> (##sys#slot (path-match "foo" <int>) 1)
 ("foo" #<procedure>)
 #;6> (##sys#size (cadr #5))
 1
 #;7> (define equiv? (hash-table-equivalence-function (make-hash-table)))
 #;8> (equiv? (path-match "foo" <int> <int>) (path-match "foo" <int>))
 #t
 }}}

 I think the problem here is the fact that, for example, (path-match "foo"
 <int>) creates a closure containing a rest list with the string "foo" and
 a procedure which is not inspectable.  If then you do (path-match "foo"
 <string>), it also creates a closure containing a rest list with the
 string "foo" and a procedure which is not inspectable.

 If you look at the way the srfi-69 hashing works, it will hash "special"
 objects (closures are such) by looking at all the slots except the first
 (because it doesn't contain a Scheme object).  Without lambda info, the
 closure objects have one less slot.

 See {{{%%special-vector-hash}}}.  The very strange thing is that it uses
 {{{##sys#peek-fixnum}}} to extract the procedure's memory location from
 the closure and convert it to a fixnum, which it then uses as a random
 seed!

 And all of that aside, like I said the lists are NOT equivalent so that
 should not make a difference, I think... Tweaking {{{recursive-hash-max-
 length}}} or {{{recursive-hash-max-depth}}} doesn't seem to help.

--
Ticket URL: <http://bugs.call-cc.org/ticket/1041#comment:17>
CHICKEN Scheme <http://www.call-cc.org/>
CHICKEN Scheme is a compiler for the Scheme programming language.