Merge pull request #18281 from shamser/issue31229

HPCC-31229 RelaxedAtomic fetch_sub implementation wrong

Reviewed-By: Richard Chapman <[email protected]>
Reviewed-by: Gavin Halliday <[email protected]>
Merged-by: Gavin Halliday <[email protected]>

system/jlib/jatomic.hpp

@@ -74,7 +74,7 @@ class RelaxedAtomic : public std::atomic<T>
     inline T load(std::memory_order order = std::memory_order_relaxed) const noexcept { return BASE::load(order); }
     inline T exchange(T _value, std::memory_order order = std::memory_order_relaxed) noexcept { return BASE::exchange(_value, order); }
     inline T fetch_add(T _value, std::memory_order order = std::memory_order_relaxed) noexcept { return BASE::fetch_add(_value, order); }
-    inline T fetch_sub(T _value, std::memory_order order = std::memory_order_relaxed) noexcept { return BASE::fetch_add(_value, order); }
+    inline T fetch_sub(T _value, std::memory_order order = std::memory_order_relaxed) noexcept { return BASE::fetch_sub(_value, order); }
     inline T add_fetch(T _value, std::memory_order order = std::memory_order_relaxed) noexcept { return ::add_fetch(*this, _value, order); }
     inline T sub_fetch(T _value, std::memory_order order = std::memory_order_relaxed) noexcept { return ::sub_fetch(*this, _value, order); }
     inline void store_max(T _value) noexcept { while (_value > load()) _value = BASE::exchange(_value, std::memory_order_acq_rel); }