I've only seen these failures happen in CI for any version of MacOS and with all tested LLVM versions.
Initial notes from @pgoodman :
we'd have to look at what leaf of cpuid that is testing, so we'd have to find in the manual or something what the meaning of those selectors for cpuid are then figure out if it is sensical for them to every produce different values. It seems unusual to get different values in a given run, though it could also be a problem somewhere in how data is copied around, e.g. maybe lifted_state isn't correctly saving rdx, and so it's always zero. Would need to look at __remill_sync_hyper_call in tests/X86/Run.cpp
I've only seen these failures happen in CI for any version of MacOS and with all tested LLVM versions.
Initial notes from @pgoodman :
The following is the output from this test run:
It isn't always the same test (but there are many similar error messages), so here is different output from this test run: