As seen in #6267, when running the tests on Windows there are slightly different things in scope than when the tests are run normally. per @nikomatsakis:
The problem is that the checkfast procedure does not generate a use std at the top of the combined program that it generates. This is not really a concern of this PR in particular, but it'd be simple enough to fix. I am inclined to think, as a separate issue, that use std ought to be in the prelude. In any case, rather than JUST xfail-fasting them, can we put a comment as well explaining why? Potentially we could open an issue for this problem as well.
As seen in #6267, when running the tests on Windows there are slightly different things in scope than when the tests are run normally. per @nikomatsakis: