Re: [jamboost] Re: Boost.Build V2, Milestone 3 preview

From: Vladimir Prus (ghost_at_[hidden])
Date: 2003-04-28 04:50:09

Hi Kirill,

first of all, thanks for trying the preview version! Some of the problems you 
report certainly should be closed before releasing M3 for a more wide 
audience.

> Few problems. First of all path-constant seems to
> misbehave. I define some constant, say WORK, then I
> try to refer file $(WORK)/foo.cpp, and build complains
> saying that don't know how to create /C:/Work/foo.cpp,
> even though the file is present at that location and
> accesible just fine via full path.

I've just comitted a test and fix for this problem to the CVS.

> Secondly, when subproject has requirement include
> certain absolute path, build systems fails to
> recognize that path is absolute, and appends it to
> subproject path.

I think I've fixed this problem as well.

> Another minor bug is commented in attached testcase.

You mean "duplicated virtual target" error? Good catch! The second
target should compile hello.cpp into a separate directory. I'll see why it's 
not done.

> Also I have a quick question. Suppose I am using some
> third party prebuilt library, e.g.
>
> lib bar : : <file>somelocation/bar.lib ;
>
> How do I specify the include dir for that library?
> Currently I am trying to put it in project
> requirements.

I think usage requirements are up to this job:

lib bar : : <file>somelocation/bar.lib : : <include>somelocation/include ;

> Also suppose it is a shared library, and names of the
> dll and import lib are different, would it be
> sufficient to refer to lib only? 

I believe so.

> Should I add
> <link>shared in this case? 

I think no. As far as build system is concerted, you'd linking to a static 
library. 

> When I am trying to add
> <link>shared I am seeing odd behavior -- no error
> messages (apart from numerous warnings of unused
> sources) and no "...patience..." message, no attempt
> to build anything, it just prints warnings and quits.
> I did not create a test case for that, since I am
> little bit fuzzy on how to specify such an external
> library, and it most likely fails because of my fault.

Actually, I think the "unused sources" warning should be 
made more reasonable. After that we'll have to check that the
use case you describe produces either some reasonable results, or a good error 
message.

Thanks,
Volodya