That project is part of the showcase platform for C++ projects.

That project makes use of the random generation ('randgen') component,
and there are two distinct implementations for that integration,
materialised in both the 'trunk' and 'trunk_w_randgen' Git branches:
 - One implementation, materialised by the 'trunk' Git branch, does not embed
   the 'randgen' component. The build (CMake) configuration has thus to import
   the 'randgen' component, as an external library. It is a standard CMake
   CMake configuration, which can therefore remain simple.
   The build of all the components, 'combase, 'randgen' and 'main', is handled
   by a simple (Ruby-based) Rake file, maintained by the 'meta' project
   (http://github.com/cpp-projects-showcase/meta).
 - The other implementation, materialised by the 'trunk_w_randgen' branch,
   embeds the 'randgen' component, which is therefore an internal component.
   Here, the build (CMake) configuration has to create an internal/inter-module
   dependency. While that operation is not difficult per se, the complexity
   arises from the fact that the 'randgen' component must still be integrable
   as an external component by other projects. The build configuration must
   therefore support the integration of 'randgen', both as an internal and as
   an external component. That implementation (Git branch) therefore shows
   an example of a complex build (CMake) configuration.
   The build of that branch does not need the (Ruby-based) Rake file of the
   'meta' project.

The purpose of the 'trunk' branch and of the 'meta' project is to show how
it is possible to co-develop several inter-dependant components, how to build,
to test and to deliver all those projects in a consistent manner, while keeping
the configuration pretty simple, though efficient.
In contrast, the 'trunk_w_randgen' branch does not need the (very light) 'meta'
project. It implements a complex CMake configuration, targeting the same goal
as the Rake file of the 'meta' project, but in a much more complex way,
involving expertise both with Git (imposing to master sub-trees or sub-modules)
and CMake (for the inter-dependencies among components).