[racket] Running code on exit....

Does it make sense to have the resource managed by a custodian? That
is, operate the ffi/C level where you can extend what happens when a
custodian is shutdown to close whatever needs closing or finalizing or
whatever before it gets gc'd but when the custodian is shutdown?

Robby

On Thu, Sep 1, 2011 at 7:24 PM, John Clements <clements at brinckerhoff.org> wrote:
> I'd love to find a way to call a sound library's "terminate" function when Racket shuts down.  My recollection of the lay of the land here is that it's very hard to make this happen correctly without mangling Racket itself. At a high level, I believe the explanation is that it violates the boundaries of the user "sandbox"--not a literal sandbox here.
>
> My experiments have involved will executors, and it seems that they get called fine when I explicitly remove all references to a value and then call collect-garbage, but not when I just quit drracket or re-run the buffer.
>
> For instance:
>
> #lang racket
>
> (require racket/date)
>
> (define my-executor (make-will-executor))
>
> (define token-value 13)
> (set! token-value (list 3 4 5))
>
> (will-register my-executor token-value
>               (lambda (dc) (with-output-to-file "/tmp/foo"
>                              (lambda ()
>                                (format "~a\n"
>                                        (date->string
>                                         (seconds->date (current-seconds)))))
>                              #:exists 'append)))
>
> (thread (lambda ()
>          (will-execute my-executor)))
>
>
> I'm guessing that to make this work, I'd have to move the thread outside of the current custodian, and even that might not work if you quit the whole application.
>
> Am I missing something obvious, or is this genuinely hard to do without "cheating" ?
>
> John
>
>
> _________________________________________________
>  For list-related administrative tasks:
>  http://lists.racket-lang.org/listinfo/users
>