18.11.4.3/4: class-name and namespace-name for closure and global scope

[hubert-reinterpretcast]

The response to CH 057 cuts out the interpretation that closure types and the global scope would be considered unnamed. This leaves another problem: get_name_v is now not supposed to be the empty string for closure types and for the global scope, but there is no clear name to produce.

[camio]

What would you suggest we do here?

[hubert-reinterpretcast]

get_name_v should probably be the empty string for a closure type (yes, it means that is_unnamed for a meta-object type reflecting a closure type true, but the closure type is not an unnamed type). For the global scope, I guess :: is okay?

[Axel-Naumann]

To me the current wording requires implementations to give out the impl-defined name of the closure type, hitting the case

for T reflecting a cv-unqualified type that is .. a class type, its class-name;

For the global namespace we hit

for T reflecting ... a namespace, its namespace-name;

That's broken - the global scope has no namespace-name. We can leave the first issue (leaking implementation-defined closure name) until after publication. The second issue is a wording defect; propose we modify the bullet (4.5.1) as follows:

(4.5.1) a namespace other than the global scope, its namespace-name;

As this is non-editorial we should fix this as an issue on the TS.