cjappl wrote: Ok, it appears that on Apple systems, we would catch this, as we intercept OSSpinLockLock used by compiler-rt atomics:
https://github.com/llvm/llvm-project/blob/c55e24b8507d47a8cc04b5d9570e8e3d02be1ca3/compiler-rt/lib/builtins/atomic.c#L95-L102 But, in the non-darwin case this is an atomic CAS spinlock, which we do not intercept (but we are trying to figure out how to catch potentially unbound loops like this): https://github.com/llvm/llvm-project/blob/c55e24b8507d47a8cc04b5d9570e8e3d02be1ca3/compiler-rt/lib/builtins/atomic.c#L112-L119 Would you please confirm that you end up in this area of atomic.c when you run this test? If so, I think it is adequate to #ifdef out this test under the circumstances that it is 1. Compiling with the builtin atomics 2. Not on darwin https://github.com/llvm/llvm-project/blob/d58d105cdaf366d7db3f60d356b21bc8e64666fb/compiler-rt/lib/rtsan/tests/rtsan_test_functional.cpp#L164-L171 Or if you're feeling ambitious, you could pull this little bit into a new test and ifdef out that, might be cleaner. Let me know if I can be of any assistance with any of this. https://github.com/llvm/llvm-project/pull/106603 _______________________________________________ cfe-commits mailing list [email protected] https://lists.llvm.org/cgi-bin/mailman/listinfo/cfe-commits
