Change Details

Calls to __atomic_load now get converted to use a thunk call, which is instrumented with the relevant shadow/origin code. This is because, when compiled with exceptions, a call to __libatomic_load might come at the end of the function, with no succeeding BB. So the previous approach of instrumenting right after the call didn't work in that case.

Have the front-end use the `nounwind` attribute on atomic libcalls. This prevents us from seeing `invoke __atomic_load` in MSAN, which is problematic as it has no successor for instrumentation to be added.

Calls to __atomic_load now get converted toHave the front-end use a thunk call, which is instrumented with the relevantthe `nounwind` attribute on atomic libcalls. shadow/origin code.This prevents us from seeing `invoke __atomic_load` in MSAN, which This is because, when compiled with exceptions, a call to __libatomic_load might come at the end of the function, with no succeeding BB. So the previous approach of instrumenting right after the call didn't work in that caseis problematic as it has no successor for instrumentation to be added.