Your quandry is that you want to generate a build-file set, with CMake, that is displaced from
the source tree and then in the generated build-file set run an IDE project
that is not displaced from the source tree. You can't have it both ways.
The CMake build directory, whether in-source or out-of-source, is the
directory in which CMake will generate build files for your project,
for some target build system. These artifacts are of service to the end-user who
wants to build your project from packaged source. You need to distinguish the target
build system from your development environment. Your development environment
is a personal productivity choice that has no essential link with the target build system.
It is just co-incidence that your target build system is Visual Studio and so
is your development environment. Imagine instead that your development environment is, say, Code::Blocks,
and your target build system is GNU Make, with CMake employed to generate the GNU Make build files.
In that case you have a Code::Blocks .cbp project file that defines
your development project, and a CMakeLists.txt file that defines a distributable
GNU Make build.
The CMake project and the Code::Blocks project have no connection with each other
except the connections you choose to make in terms of common tools, files and targets.
You might have more files and/or targets in your development project than you
ever choose to reflect in the CMakeLists.txt. The target toolchain could be different
from the development toolchain. As far as the CMake project is concerned it doesn't matter
at all where the Code::Blocks .cbp file resides or if such a thing exists, although you will
want the .cbp file as well as the CMakeLists.txt to be under source control in your development branch.
When you run CMake and create build files, you come under no compulsion to stop using
your CodeBlocks project, in situ, as your development environment and somehow
migrate the development environment into the build-files folder. Even
if you instructed CMake to generate a Code::Blocks .cbp file in the build-files,
so that end-users of the package can build it with Code::Blocks, you wouldn't then be tempted to
start using that project file as for your development environment: it's a throw-away artifact of running CMake.
If you want to develop the project with Visual Studio then make a VS project for your development purposes in a normal relationship to
the source tree. Make a CMake project for distribution of the project on your target build system. Forget the incidental
fact that the target build-system is also Visual Studio. Keep the files and targets of your CMake project aligned with
those of your development project to the extent that the CMake project is up to date
whenever you build it. And ensure, of course, that the gating tests you run, for commit, release or whatever, are all run on the CMake build.
If you are not interested in distributing your project and are using CMake merely for
the purpose of generating a Visual Studio project in which to develop it - well that
would be a gratuitously laborious alternative to just developing it in Visual Studio -
but if it's what you want to do you will have to be clear that CMake and Visual Studio cannot both control
the composition of the project. CMake controls it and the VS project is just a CMake artifact. If you want to
add files to the project or change it in any other way then you must do it CMakeLists.txt, then rerun CMake and accept the
VS project that drops out - even if you don't quite like the look of it.
