Library dependency graphs in CMake
One of the main advantages of using popular programs instead of reinventing the wheel every time is the integration with other programs.
cmake -S <source dir> -B <build dir>; mkdir <build dir>/deps; cd <build dir>/deps; cmake --graphviz=deps <build dir>; mkdir <build dir>/deps-svg; mkdir <build dir>/deps-png; for i in *; do :; dot -Tsvg -o "<build dir>/deps-svg/$i.svg" "$i"; dot -Tpng -o "<build dir>/deps-png/$i.png" "$i"; done
The main advantage oh having a visual representation of the dependency graph, it that it makes it easier to reason about side-effects when working on a library.
It also makes it possible to answer easily a whole series of question, without looking at the code
Are there circular dependencies?
Are two libraries somehow linked together?
Are there libraries that are used too much?
Are there libraries that are not necessary anymore?
Do libraries have too many dependencies?
How risky is upgrading a library, what are the possible pain points.
If a library changes, what components do we need to test and adapt.
While it is also possible to answer all those questions by looking at the CMake files, in practice it is not that easy.
Build files are not always declarative, and the logic is normally split over multiple files.
Thus generally one should rebuild by hand the graph (which might depend on the platform, build parameters, and other factors), which is an error-prone process.
The visual representation on the other hand gives an easy to understand overview, albeit of a single configuration.
The textual representation (encoded in the
.svg files) can even be used for doing some automatic validation.
Do you want to share your opinion? Or is there an error, same parts that are not clear enough?
You can contact me here.