Currently all of them live in the hir-def crate, making use of nameresolution, but I believe we should have some tests within hir-expand that strictly only test the expansion parts. It feels wrong to me that they all live within hir-def (this also means we cant easily test a subset of r-a swapped to the new salsa for macro expansion if we wanted ;)

Kind of a funny stance as that goes directly against #10494 😬 I don't fully disagree, but I t think the issue at hand is more that we just have to little tests for macros. Some would make more sense within hir-def, some more in hir-expand.