Xref: utzoo comp.object:579 comp.lang.c++:5799
Path: utzoo!utgpu!jarvis.csri.toronto.edu!cs.utexas.edu!sun-barr!newstop!sun!brahmand!grover
From: grover%brahmand@Sun.COM (Vinod Grover)
Newsgroups: comp.object,comp.lang.c++
Subject: Re: Continuations
Message-ID: <128758@sun.Eng.Sun.COM>
Date: 4 Dec 89 22:34:49 GMT
References: <1989Nov30.002204.4899@mentor.com>
Sender: news@sun.Eng.Sun.COM
Reply-To: grover@sun.UUCP (Vinod Grover)
Organization: Sun Microsystems, Mountain View
Lines: 18

In article <1989Nov30.002204.4899@mentor.com> plogan@mentor.com (Patrick Logan) writes:
>
> >>From: grover%brahmand@Sun.COM (Vinod Grover)
> >                                       Admittedly, this is not as elegant as
> >Scheme's notion of continuation but workable in concept.
> >
>To be more accurate, longjump is not workable in concept because a
>longjump can only be executed once per setjmp. A continuation in
>Scheme can be captured once and executed zero or more times. Things
>become more complicated when you start considering nested setjmps
>and longjumps.
Yes, you are quite right. In fact the problem is even deeper than that, as I
just learned, if the function calling setjmp returns before the call to
longjmp is made then the longjmp is made to a non-existent environment. 
Thanks for correcting me.

Vinod Grover
Sun Microsystems.