need opinions on the digraph in libbpf-tools/deadlock

I have a working CO-RE deadlock that is based on deadlock.py.
Deadlock tool requires different data structures and algorithms, such as digraph, hash map, etc.
Now that all libbpf-tools are built in C, my deadlock code is also written in C.
However, I couldn't find any C-based digraph open source that satisfies all the requirements of the deadlock.
So for now I've been using a simple C digraph with auxiliary code to support dynamic graph sizes, arbitrary data types and values. And additionally, I used an open source hash map.
However, I think the current code is a bit complicated and not good for code maintenance.
So I'm looking for your opinions on the use of various algorithms like digraph and hashmap in libbpf-tools.
There are several candidates.

1. use simple C digraph + auxiliary code to support dynamic graph sizes, arbitrary data types and values. (my current version)
graph: simple C graph like https://kalkicode.com/detect-cycle-in-directed-graph-using-bfs-in-c + auxiliary code
hash: uthash (https://github.com/troydhanson/uthash)

2. use stdlibc++ and opensource digraph based on c++
graph: https://github.com/grame-cncm/digraph
hash: stdlibc++ (stl)
I roughly checked and the C++ graph above meets all the deadlock requirements like networkx used in deadlock.py.
and I checked deadlock.cpp can be compiled with some modifications of libbpf-tools/Makefile.

3. use a C-based digraph that supports all deadlock needs. (I don't have it now.)

Which one do you prefer more?
After more checking, is it ok to choose the second if I want?
Any comments are welcome.
Please let me know if you need more information.
Thank you.