Re: [boost] MPC support for Boost

From: Vladimir Prus (ghost_at_[hidden])
Date: 2006-03-22 03:07:33

Kevin Heifner wrote:

> Paul A Bristow wrote:
>> I've downloaded this and thought about trying it, but like bjam,
>> something else suddenly seemed more important ;-)
> 
> Funny how that works. :)
> 
>> Isn't this rather a sledge hammer to crack a nut for this particular
>> problem (without in any suggesting that it isn't an excellent building
>> tool).
> 
> I don't think so.  The beauty of MPC is that it generates
> makefiles, MSVC++ 7.1, MSVC++ 8, etc. from one easy to maintain
> file.  

How is this different from CMake (http://www.cmake.org/HTML/Index.html) or
from QMake (http://www.trolltech.com/products/qt/qmake.html)?

> Often times no source files have to be listed.  Therefore 
> when files are added to the project all that is needed is to
> rerun MPC.  No need to modify three, four, or more "makefiles".

The same applies to Boost.Build -- there's just one file.

> It is especially nice when some developers do not have access to
> MSVC++ or Borland or whatever, they don't have to worry about
> maintaining these build files.

Again, this is the point of Boost.Build already.

> bjam provides many of these same benefits.  The difference is
> that MPC lets everyone work in their favorite environment.  

This can be more then offset by the idea to support two different build
systems. If folks (specifically MSVC users) would benefit from having
special files, then Boost.Build can be modified to generate those files.
OTOH, I don't remember such requests on Boost.Build mailing list.

> Long 
> term we hope to support bjam output as well.

In order to support bjam output, you need to have as expressive input
language as bjam's, and we already have bjam's input language.

Finally, do you have a setup that can both build boost and run regression
tests? If not, the it's a tricky question if binaries created by your tool
will pass the same regression tests that binaries created by Boost.Build.

- Volodya