fix thread safety due to read lock for write operation

Author: BewareMyPower

lib/ClientImpl.cc

@@ -831,15 +831,9 @@ void ClientImpl::shutdown() {
     lookupCount_ = 0;
 }
 
-uint64_t ClientImpl::newProducerId() {
-    std::shared_lock lock(mutex_);
-    return producerIdGenerator_++;
-}
+uint64_t ClientImpl::newProducerId() { return producerIdGenerator_++; }
 
-uint64_t ClientImpl::newConsumerId() {
-    std::shared_lock lock(mutex_);
-    return consumerIdGenerator_++;
-}
+uint64_t ClientImpl::newConsumerId() { return consumerIdGenerator_++; }
 
 uint64_t ClientImpl::newRequestId() { return (*requestIdGenerator_)++; }
 

lib/ClientImpl.h

@@ -207,8 +207,8 @@ class ClientImpl : public std::enable_shared_from_this<ClientImpl> {
     std::unordered_map<std::string, LookupServicePtr> redirectedClusterLookupServicePtrs_;
     ConnectionPool pool_;
 
-    uint64_t producerIdGenerator_;
-    uint64_t consumerIdGenerator_;
+    std::atomic_uint64_t producerIdGenerator_;
+    std::atomic_uint64_t consumerIdGenerator_;
     std::shared_ptr<std::atomic<uint64_t>> requestIdGenerator_{std::make_shared<std::atomic<uint64_t>>(0)};
 
     SynchronizedHashMap<ProducerImplBase*, ProducerImplBaseWeakPtr> producers_;