Thanks for the answers. For sure it is possible to buy an ATmega and even a full-featured debugger would be acceptable compared to the effort of porting. But still, I am very biased towards the PIC without logical or economical arguments.
I did not yet work with the mysensors libraries and only browsed quickly through the github repo. Maybe someone can give me a few hints to estimate the effort deeper than just claiming it as "much workload"?
What I see so far is:
- make the C++ code compile (translate to C with clang/llvm and compile with the XC8 compiler)
- create a new HAL (in hal/architecture) which seems not too much effort for making it initially working
- Do I see it right, that there is a linux-port available? -> I would expect much more effort to port from AVR to linux than to port it to a different MCU
- What about the licensing? It looks like the code is GPLv2, but in the CLA it seem that contributors need to give away their rights on the contributions and that mysensors can even redistribute the code under another license - which seems completely against the principles of the GPL. Can someone explain that in more detail?