Re: [jamboost] More on features

From: David Abrahams (dave_at_[hidden])
Date: 2002-08-08 09:06:32

----- Original Message -----
From: "Vladimir Prus" <ghost_at_[hidden]>

> > Hmm, "exectuted" properties anyone?
>
> Yea, except with a somewhat different semantics and the fact that you can
> have both executed features (i.e. a rule handles every value of feature)
and
> executed properties (i.e. rule handles a particular feature, value pair).
>
> > Didn't we just decide we didn't need those?
>
> We haven't yet, IIRC.
>
> > > Likewise, if there's feature
> > >
> > > python-arity
> > >
> > > with any value, another rule would translate it into appropriate
defines.
> > > Opinions?
> >
> > I worry that this approach won't afford us enough control... but
suppose
> > I'm wrong? What would we be giving up by taking this approach?
Anything?
>
> I don't think we'll give up anything. After all, properties added by
executed
> features can affect generators selection, if this is desperately needed
in
> some (very few, I think) cases.
>
> > What was the motivation for shoving all of this functionality into
> > generators in the first place?
>
> You've proposed it as a way to handle stlport. I don't remember any other
> motivation.

It certainly sounds like a massive simplification to generators. I never
liked the idea of introducing the orthogonal "generator categories". It
always felt like a hack to me.

-Dave
-----------------------------------------------------------
David Abrahams * Boost Consulting
dave_at_[hidden] * http://www.boost-consulting.com