Re: <link> issues solved?

From: Markus Schöpflin (gclbb-jamboost_at_[hidden])
Date: 2002-12-06 14:52:14

"Vladimir Prus" <ghost_at_[hidden]> schrieb im Newsbeitrag
news:3DF0CA23.8070000_at_cs.msu.su...

> Well, Dave and I were chatting via Jabber and seems to solved
> the <link> issues.

Great news.

> We have failed to see clear distinction between internal and external
> libraries. This is especially tricky if you use several projects,
> all of which use Boost.Build.

Ok.

> The proposed solution is: introduce a simple <link> feauture.
> It will have two values, static and shared, which will
> determine what kind of libraries are desired.
>
> So, for example <link>static will make everything static.

Ok.

> If you want a particular linkage for a certain library, you can specify
> that in two ways.
>
> 1. When referring to that library
>
> exe a : a.cpp qt/<link>shared ;

Looks good.

> 2. By using "source" rule as I've described in a previous message:
>
> source zlib : /path/to/zlib/installation/zlib : <shared>false ;
>
> As Dave said: "So you're using "source" as a way of adding a level of
> indirection with attached bulid requirements?"

Understood. Maybe we should think of a new name for the source rule then.
After all, it can do a lot more than just hold a collection of sources now.
I can't think of a really good name right now, but maybe something like
'module' or 'alias'. The latter might be a good choice, after all the rule
aliases a name with a set of sources or libraries and/or build requirements.

> We'll also introduce <link-runtime> feature, which will affect
> linking of runtime library only. It's important for Windows,
> and seems unavoidable.

Agreed.

> The last detail: <link> will be a composite feature and
> <link>static will expand to
> <link>static <link-runtime>static.
>
> Therefore, <link>static will really make everything static.
> OTOH, it's possible to override <link-runtime> explicitly.

Won't you have a problem with the recursive definition? Just asking.

> Is that solution OK for everybody?

Fine with me.

>
> - Volodya
>
> P.S. Markus, Dave told he recieve an email where you say that
> firewall did not play nice with Psi. I did not recieve that
> message yet :-( I still hope we can do something about that.
> Real-time chat proved quite effective.

I'll see if I can come up with something on Monday.

Markus