Vincent Povirk madewokherd@gmail.com writes:
While we could modify our test to detect this seemingly impossible situation, I think it's better to just prevent registry virtualization. (Then again, if we ever want to write tests for registry virtualization, they should probably go in advapi32, so maybe editing the test is better.)
We will want to add tests for virtualization, so yes, I think that fixing the test would be preferable.