Question, possible feature request

For data members of rule contexts, is there any particular reason that specifying some rule context class name as the type for a parameter for that some other rule tries to accept the argument only by value instead of by address? Would changing this so that if a type name used for a parameter refers to a rule context name in the grammar, then that parameter is actually just a pointer and not the concrete value cause more issues?

No file, test, or entry point is named. Start by tracing how the C++ target represents rule-context types used as grammar parameters, then determine the compatibility and code-generation impact of passing them by address; done requires an agreed behavior and corresponding tests.

Written by the indexing model from the issue text.