symbols incorrectly GC'ed

With 32-bit revision 1719, I get something different:

[d@eep:~/bug-p]-> ikarus --r6rs-script problem.sps
Unhandled exception:
 Condition components:
   1. &who: let-string-start+end
   2. &message: "unbound identifier"
   3. &undefined
   4. &trace: #<syntax let-string-start+end>
[d@eep:~/bug-p]-> cd ../
[d@eep:~]-> export IKARUS_LIBRARY_PATH=`pwd`/bug-p
[d@eep:~]-> ikarus --r6rs-script bug-p/problem.sps
Unhandled exception:
 Condition components:
   1. &who: let-string-start+end
   2. &message: "unbound identifier"
   3. &undefined
   4. &trace: #<syntax let-string-start+end>
[d@eep:~]->

I haven't investigated the SRFIs yet.

On Dec 15, 2008, at 9:29 PM, Derick Eddington wrote:

> I haven't investigated the SRFIs yet.

I think this is coming from SRFI-14. I wonder if resolve/include is
doing something funny or if I somehow broke it.

Aziz,,,

I don't think it has to do with your aliased libraries, because the same
problem happens when the (xitomatl srfi ---) are used directly:

[d@eep:~/bug-p]-> echo $IKARUS_LIBRARY_PATH
/home/d/bug-p
[d@eep:~/bug-p]-> ikarus
Ikarus Scheme version 0.0.3+ (revision 1719, build 2008-12-15)
Copyright (c) 2006-2008 Abdulaziz Ghuloum

> (import (spells define-values) (xitomatl srfi char-set) (xitomatl srfi strings))
Unhandled exception
 Condition components:
   1. &who: let-string-start+end
   2. &message: "unbound identifier"
   3. &undefined
   4. &trace: #<syntax let-string-start+end>
>

It is strange because both let-string-start+end and check-substring-spec
are from (xitomatl srfi strings), but if I comment-out one of the other
imports, it works:

[d@eep:~/bug-p]-> ikarus
Ikarus Scheme version 0.0.3+ (revision 1719, build 2008-12-15)
Copyright (c) 2006-2008 Abdulaziz Ghuloum

> (import #;(spells define-values) (xitomatl srfi char-set) (xitomatl srfi strings))
> (string-join '("a" "b" "c") ":")
"a:b:c"
>
[d@eep:~/bug-p]-> ikarus
Ikarus Scheme version 0.0.3+ (revision 1719, build 2008-12-15)
Copyright (c) 2006-2008 Abdulaziz Ghuloum

> (import (spells define-values) #;(xitomatl srfi char-set) (xitomatl srfi strings))
> (string-join '("a" "b" "c") ":")
"a:b:c"
>
[d@eep:~/bug-p]->

Seems it has something to do with the datum comment in the export clause
of xitomatl/srfi/strings.sls, because deleting it makes the problem go
away:

[d@eep:~/bug-p]-> ikarus
Ikarus Scheme version 0.0.3+ (revision 1719, build 2008-12-15)
Copyright (c) 2006-2008 Abdulaziz Ghuloum

> (import (spells define-values) (xitomatl srfi char-set) (xitomatl srfi strings))
Unhandled exception
 Condition components:
   1. &who: let-string-start+end
   2. &message: "unbound identifier"
   3. &undefined
   4. &trace: #<syntax let-string-start+end>
>
[d@eep:~/bug-p]-> cp xitomatl/srfi/strings.sls xitomatl/srfi/strings.sls--prev
[... edited file ...]
[d@eep:~/bug-p]-> diff -u xitomatl/srfi/strings.sls--prev xitomatl/srfi/strings.sls
--- xitomatl/srfi/strings.sls--prev 2008-12-15 19:52:46.000000000 -0800
+++ xitomatl/srfi/strings.sls 2008-12-15 19:52:57.000000000 -0800
@@ -64,13 +64,6 @@
     ; R5RS re-exports:
     string? make-string string-length string-ref string-set!
     string string-append list->string
- ; Low-level routines:
- #;(make-kmp-restart-vector string-kmp-partial-search kmp-step
- string-parse-start+end
- string-parse-final-start+end
- let-string-start+end
- check-substring-spec
- substring-spec-ok?)
     )
   (import
     (except (rnrs) error string-copy string-for-each string->list
[d@eep:~/bug-p]-> ikarus
Ikarus Scheme version 0.0.3+ (revision 1719, build 2008-12-15)
Copyright (c) 2006-2008 Abdulaziz Ghuloum

> (import (spells define-values) (xitomatl srfi char-set) (xitomatl srfi strings))
> (string-join '("a" "b" "c") ":")
"a:b:c"
> (string->char-set "abc")
#[:char-set ...]
> (define-values (x y) (values 1 2))
> (list x y)
(1 2)
>

I'm suspecting that the problem lies in the new symbol table. Investigating ...

Definitely a bug in the symbol table. This test exhibits it.

(let ()
  (define random-string
    (lambda (n)
      (list->string
       (map (lambda (n)
              (integer->char (+ (char->integer #\a) (random 26))))
            (make-list n)))))
  (let ([str1 (random-string 70)]
        [str2 (random-string 70)])
    (printf "sym1=~s\n" (string->symbol str1))
    (do ((i 0 (+ i 1))) ((= i 1024)) (collect))
    (let ([sym1 (string->symbol str1)])
      (printf "sym1=~s\n" (string->symbol str1))
      (printf "sym2=~s\n" (string->symbol str2))
      (let ([sym3 (string->symbol str1)])
        (printf "sym3=~s\n" (string->symbol str1))
        (assert (eq? sym1 sym3))))))

Added a fix in revision 1720. Can you please confirm the fix?

Yes, the problem went away with 1720. Thanks for the quick fix!