Summary

variables defined inside closure survive beyond closure!

Metadata

Attachments

Description
;debug.scm

(require-extension r7rs)

(define r #t)

(define test (lambda ()

(if r
	(begin
		(set! r #f)
		(define v 123)
		(test)
		(display v)
		(newline)
	)
	(begin
		(define v 234)
	)
)

))

(test)

; output
234
; The question is
Each time a procedure or partial block (like "let") is called, a closure should be created. Variables defined inside a closure should have a limited scope inside the closure unless "call/cc" or "call-with-values" is used. In the case above, none of them is used but the variable "v" survives beyond its closure created by "test" the 2nd time.
; Further more, I found that only when "begin" is used will this bug appear.

Changes and comments

[2016-06-09 04:28:13 UTC] ai-artisan attached debug.scm (description=csi -w -s debug.scm)

[2016-06-09 04:39:21 UTC] evhan changed priority from critical to minor

[2016-06-09 04:39:21 UTC] evhan wrote:

Internal definitions are only valid at the start of a body (see R5RS). You have definitions in invalid locations (begin does not open a body), which CHICKEN allows but treats as global assignments.

The following, where internal definitions are correctly placed, prints 123 as you expect:

 (define test
   (lambda ()
     (if r
         (begin
           (set! r #f)
           (let ()
             (define v 123)
             (test)
             (display v)
             (newline)))
         (let ()
           (define v 234)))))

[2016-06-09 10:43:48 UTC] ai-artisan wrote:

Thx. I just referred to the R7RS. You r right. I made a mistake. ^_^

[2016-06-09 11:01:45 UTC] ai-artisan changed priority from minor to not urgent at all

[2016-08-02 20:35:52 UTC] sjamaan changed status from new to closed

[2016-08-02 20:35:52 UTC] sjamaan set resolution to duplicate

[2016-08-02 20:35:52 UTC] sjamaan wrote:

This is exactly the same underlying issue as #1309