Re: [jamboost] <file> specifier in library causes Duplicate name for actual target

From: Michael Stevens (Michael.Stevens_at_[hidden])
Date: 2004-02-13 09:38:45

Hi again Volodya,

On Wednesday 11 February 2004 08:59, Vladimir Prus wrote:
> Hi Michael,

>
> > Looking into my own problem the the matter seems rather more complicated
> >
> > :-( The problem revolves around the different way library sources are
> >
> > interpreted.
> >
> > Previously I was using the follow complex definition:
> > # System libraries - specify local link requirements for lapack
> >       lib g2c : : <name>g2c ;
> >       lib lapack : : <name>lapack ;
> >       lib lapack : : <toolset>gcc <name>lapack : : <library>g2c  ;
> >       lib lapack : : <toolset>intellinux  <name>lapack : :<library>g2c 
> > ; lib lapack : : <toolset>msvc <file>"../../CLAPACK3/clapack7.a" ; # The
> > executable for testing
> > exe rtheta : rtheta.cpp lapack ;
> >
> >
> > However this seems non functional when static linking. The library g2c
> > is never propagated to an exe as a dependency. Similar the simpler lib
> > lapack : g2c  : <name>lapack ;
> > does not propgate the g2c library dependency.
>
> It should do it... maybe there's some bug, though.
I'll look into this a bit further and report if I find anything like a but.

>     lib lapack : "../../CLAPACK3/clapack7.a" : <link>static;
>
> Boost.Build tries to find a generator which can create STATIC_LIBRARY from
> a single source of type STATIC_LIBRARY. It fails to find any, since in fact
> no such generator exist.
>
> Cool, so it thinks nothing is created and ignores source. Not nice :-(
>
> It's not really good that "lib" behaves this way, but I'm not sure I'll
> fix it right now...
I understand. I'll keep it on file as future test case.
>
> But, OTOH, it looks that "lib" above is not what you're after. You're
> really creating alias, and can write
>
>     alias lapack : "../../CLAPACK3/clapack7.a" : <link>static;
Ohhh. This is a BIG surprise to me. It is possible to mix alias and lib main 
targets with the same name. Logical I suppose, but I assumed that they would 
be mutually exclusive.

Thanks and all the best,
Michael
-- 
___________________________________
Michael Stevens Systems Engineering
Navigation Systems, Estimation and
Bayesian Filtering
http://www.sf.net/Bayes++
___________________________________