https://github.com/labath commented:
I think there's a race in the cancellation handling: 1. request handling thread sets active_seq (but doesn't start executing the request yet) 2. cancellation request comes in, queue_reader sets the interrupt flag 3. request handling thread begins handling the request, starts by clearing the flag 4. request is not cancelled I think this would be easier to guarantee correctness (and review) if `active_seq` was *not* an atomic and its modification, as well as the acts of setting and clearing the interrupt flag were done under a mutex https://github.com/llvm/llvm-project/pull/130169 _______________________________________________ lldb-commits mailing list lldb-commits@lists.llvm.org https://lists.llvm.org/cgi-bin/mailman/listinfo/lldb-commits
