CPM's ability to fetch dependencies from Github is cool, and something I will eventually get to. I also like there's no need for 61 million helper files like conan or vpkg.

However, in the present case I have a large mono repository ~25 million lines of C/C+ code. This repository has ~250 libraries all inside the repo. As such there is no need to fetch from Github. The libraries are organized with a canonical layout that makes it easy to scan and discover.

Here's what I want to do:

• run a script (that I will provide) to scan the libraries
• from each library create CPM cmake code for that library and its dependencies. Dependencies, say lib_network depends on lib_network_async will come from meta data I provide
• use the output of the previous bullet to autogenerate a Cmakefile with CPM CPMAddPackage commands
• logically extend dependencies in CMakeLists.txt for all the tasks, unit tests, integration tests
• cmake to build

Desired impact:

• Other than manually specifying the dependencies between libraries in a meta-file, the full CMakeLists.txt file is auto-built
• Later I can move individual libraries to Github while keeping the overall process and structure unchanged

Now this will require a CPMAddPackage to understand a dependency is not in github, but rather it's in a directory tree.

Once I have that I can move individual libraries into their own Git repo and revert to CPMAddPackage with Githib urls.

Feasible? My guess: this is not a good use-case since CMake does this out of the box.
Rather once libraries are migrated to standalone Git repos then CPM is well suited.

Meanwhile I'll read your https://gitlab.com/external-packages/cpm.cmake carefully.

Regards!