nrf_rpc: fix build error for C++

[Damian-Nordic]

In C++, conversion from a function pointer to void* is either undefined (prior to C++11) or not implicit.

Move the static inline functions to the C file for better portability. The flash impact imposed by this is negligible.

[alxelax]

Why is C code built by C++ compiler?

[Damian-Nordic]

Why is C code built by C++ compiler?

Because Matter firmware needs to include nrf_rpc.h (to call nrf_rpc_init), which includes the c++-prohibited casts.

[alxelax]

Why is C code built by C++ compiler?

Because Matter firmware needs to include nrf_rpc.h (to call nrf_rpc_init), which includes the c++-prohibited casts.

I do not think that inline functions are a correct implementation for headers, which are used in both C and C++.
I see problem not in void * usage in general, but in inlining of C code within C++ without explicit letting know about that to compiler.

I do not like forced typecasting since it might dig in deep the type mismatching. Would it be more meaningful to move inline functions into C file and remove inline modificator?

[Damian-Nordic]

I do not think that inline functions are a correct implementation for headers, which are used in both C and C++. I see problem not in void * usage in general, but in inlining of C code within C++ without explicit letting know about that to compiler.

I do not like forced typecasting since it might dig in deep the type mismatching. Would it be more meaningful to move inline functions into C file and remove inline modificator?

As long as the inline functions don't use any C-only features like compound literals, there's no issue with that, no?

Moving the functions to C file will not remove the casts. To make it 100% type-safe I would probably need to duplicate some code, which contradicts the authors' intention to optimize the code size of this fragment.

[alxelax]

Discussed IRL. Let it be but for future to be accurate with potential miscasting in such functions.

[Damian-Nordic]

@doki-nordic Could you please review?