Re: [jamboost] Re: <user-interface> and dependencies

From: Vladimir Prus (ghost_at_[hidden])
Date: 2003-11-04 03:50:49

Hi Brock,

> > I think that propagaration is still the simplest solution, though
>
> it's a
>
> > question if it's right or not. But it's a question for you, and
>
> other users.
>
> I don't think I understand the system well enough to say that this
> solution is more correct. It 'feels' better off-hand.

Understood. In fact, the difference between the two solutions is that if
you make <user-interface> relevant to exe but not to any other target, and you 
make <optimization> relevant to everything but bison and flex, you get your 
targets placed in many different directories.

When you keep the current scheme, all targets for a single property 
combination most likely will be placed in one directory. That's what Jurgen 
wanted.

> > What's you're asking is
> >
> > http://zigzag.cs.msu.su:7814/scarab/issues/id/BB19
>
> I can't get to this issue. Could you forward me the issue as text?

It's at 

http://zigzag.cs.msu.su:7813/BB19.html

(though with a horrible formatting). Jurgen's post is 

http://article.gmane.org/gmane.comp.lib.boost.build/4030

> > I'd be happy to accept any resolution. If you belive this feature
>
> is needed
>
> > and is willing to invest some time in it, both in design and
>
> inplementation,
>
> > you're welcome!
>
> I will read the exchange between you and Jurgen and start looking at
> the code. Is it all in 'C'? That would certainly dampen my
> enthusiasm about contributing. Am I missing something?

No, it's not at 'C', at least now ;-) In fact, we need to first decide on 
logic used to decide/specify that a feature is relevant to a target.

> Also, I have made the change to builtin.jam and verified that it
> works as expected. How do I submit the fix?

Take a look at 'hacking.txt'.

- Volodya