Rob Quill wrote:
OK, that sounds good to me. As you say, there is no way to know until it is actually implemented how muh affect it will have. However, I won't be able to start on it until summer as I have to worry about sitting my finals, so I'll let you know when I get started.
I would say this is only worth a minimal penalty in debug info size. It is one thing to deal with cases where something is visible in terms of canonical language semantics, and invisible because of optimization. It is quite another to deal with error checks on debug requests that just make no sense at all because the entity is not visible because of language semantics. This would just be guarding against (a rather peculiar) pilot error.