The problem was definitely not the intended behavior -- the sandbox
trusts anything in the collection tree and installed planet packages,
so the fix that Matthew did is the right thing.
Quick question about it -- in Ray's original question there was this
problem:
channel-put: expected argument of type <channel>; given: #f
Did this come from DrRacket or from the sandbox?
More than a week ago, Matthew Flatt wrote:
> I'm pretty sure the sandbox should treat linked collections the same
> as collections accessible via `(current-library-collection-paths)',
> so I'll push that change.
>
> At Mon, 14 May 2012 08:00:36 -0500, Robby Findler wrote:
> > Oh, yse. I'm seeing that too. Thanks! DrRacket now protects itself
> > against this specific problem by catching errors during creation
> > of the sanbox and just giving up on looking at the language.
> >
> > I made an example at the end of this message show what appears to
> > be a problem with the sandbox.
> >
> > I'm not sure, but based on past experience, I think Eli will say
> > it is the intended behavior. (It seems a bit unfortunate, tho,
> > that creating an environment to do a protected form of evaluation
> > can fail outside of the environment instead of inside of it. After
> > all, isn't that kind of the point of the sandbox, to handle such
> > things for me in the library?)
--
((lambda (x) (x x)) (lambda (x) (x x))) Eli Barzilay:
http://barzilay.org/ Maze is Life!
____________________
Racket Users list:
http://lists.racket-lang.org/users
