<typeinfo>: Error with import std; and typeid(T).name()

[Stephan T. Lavavej (StephanTLavavej)]

Cameron DaCamara (@cdacamar) explained that in this case, the MSVC compiler looks for the True Name of type_info in the global namespace, so we need to export it as a special exception.

I'm adding test coverage for this scenario. The using namespace std; in this test doesn't actually affect the repro, so I didn't bother to remove it.

Fixes DevCom-10349749 and internal VSO-1810086 / AB#1810086 .

[nicole mazzuca (strega-nil-ms)]

Is this conforming? I'm okay with "no, but it needs to not be conforming for our implementation", but I'm wondering if it'd be possible to make the compiler search in std:: here, rather than continuing the non-conformance.

[Stephan T. Lavavej (StephanTLavavej)]

Cameron DaCamara (@cdacamar) said on Discord:

I suppose I could change that, but I don't want to introduce a compatibility problem 😦

The Standardese that I wrote, N4944 [std.modules]/6, says:

Recommended practice: Implementations should avoid exporting any other declarations from the C++ library modules.

IMO it's slightly nonconforming to emit ::type_info because it could clash with user code, but the same is also true for classic <type_info> (it's not a reserved name, nor is it one of the C names that is in a superposition between global and std).

In practice I think it's easiest to just make this special exception and not worry about it further.

[Casey Carter (CaseyCarter)]

I suppose I can live with this as extending our pre-existing non-conformance into the Modules Universe.

[Stephan T. Lavavej (StephanTLavavej)]

I'm mirroring this to the MSVC-internal repo - please notify me if any further changes are pushed.