Hmm, I'm modeling continuation marks as a slot in each continuation frame
holding key-value pairs. But that requires mutation of continuation frames
when a new mark is pushed. It seemed a bad idea now, but is there a spec
that prohibits such a model?
The reason I put the example is that the continuation frames CONT captured
and the continuation frames of the continuation of (with-continuation-marks
...) are the same (shared), so if the marks are a part of a continuation
frame, I thought attaching new marks to one would affect the other. It is
more like that a list of marks is a part of dynamic environment and call/cc
captures it with a list of continuation frames, but there's no direct link
between the activation records and continuation marks?
On Sat, Oct 15, 2022 at 12:22 PM Marc Nieper-Wißkirchen <
marc.nie...@gmail.com> wrote:
> Am Sa., 15. Okt. 2022 um 13:25 Uhr schrieb Shiro Kawai <
> shiro.ka...@gmail.com>:
> >
> > Just to be sure. In the following code, the returned mark set contains
> the key-value pair {key, mark1} but not {key, mark2}, correct?
>
> Continuation marks are attached without mutation like when you cons to a
> list.
>
> In the example below, CONT captured by call/cc does not include any of
> the marks set by the code so the returned mark set neither contains
> any key-value pair of the two.
>
> >
> > ===
> > (let ((marks #f))
> > (call/cc
> > (lambda (cont)
> > (with-continuation-marks 'key 'mark1
> > (begin
> > (set! marks (continuation-marks cont))
> > (with-continuation-marks 'key 'mark2 #f)))))
> > marks)
> > ===
> >
> >
>
