Revise semantic checking of serial port connections

[bocchino]

• Exclude the cases where a serial port is connected to a typed port, and the typed port has a return type.
• Update the spec, implementation, unit tests, and user's guide.
• Make some incidental revisions to the user's guide.

Closes #183.

[tiffany1618]

Should we also be excluding the case where a typed output port with a return type is connected to a serial input port?

[bocchino]

Yes, that sounds right. What happens in this case in the current implementation? Does the code even compile?

[tiffany1618]

I think the expected behavior is to have an assertion fail on a nullptr, but currently it seg faults when FW_PORT_SERIALIZATION is defined. Here is the code that's generated:

https://github.com/fprime-community/fpp/blob/d55f67d48c8fe01994bd1abddd47974e98ab6667/compiler/tools/fpp-to-cpp/test/port/ReturnTypePortAc.ref.cpp#L110-L115

Maybe we should remove lines 110-112 for ports with return types.

[bocchino]

Maybe we should remove lines 110-112 for ports with return types.

Yes, that sounds right. Also, we should systematically replace FW_ASSERT(ptr) with FW_ASSERT(ptr != nullptr). C++ does not guarantee that nullptr == 0.

[bocchino]

OK, I revised the PR to exclude both serial -> typed and typed -> serial when the port type has a return type.