Re: state of the Lua nation on resource cleanup

[Asko Kauppi <askok@...>]

Roberto Ierusalimschy kirjoitti 17.2.2009 kello 20:14:

> > > I do not see why the need for the <scoped_statements>. It would be
> > > simpler to add only a new form of locals. Something like
> > >
> > > local finalized f, g = ...
> > >
> > > that calls ":close()" (or __gc??) when it goes out of scope.  Of
> > > course,
> > > you can (should?) use a conventional do-end around it:
> > And all of that can easily be done using token modifiers.
> >
> > Case solved?  :)
> I do not see how this can be done with token modifiers, if we want to
> handle exceptional exits too.
>
> -- Roberto

Yes, you are right.  Token modifiers would cope with the normal exits  
(break/return/end of scope) but they couldn't "see" an error happening  
in some called function, for example. Not without a pcall.
As to the coroutine issue, would it be valid to just say a coroutine  
switch (yield/resume) is not allowed to take place within a  
finalizable scope. Would that be too restricted in practice?  If not,  
then the coroutine problem vanishes.
I do have one working solution to this, which is the Lanes approach.  
But I know it's not the generic answer.
-asko