Introduced in #103488, they are used only in super_relate_tys.
I believe super_relate_tys should always relate types structurally. Any smarter inference rules should be handled by the caller instead, similar to how we do lazy normalization in the new solver.
Introduced in #103488, they are used only in
super_relate_tys.I believe
super_relate_tysshould always relate types structurally. Any smarter inference rules should be handled by the caller instead, similar to how we do lazy normalization in the new solver.