[ASAN] Access node through a weak_ptr on distributed_work dtor

nano/node/distributed_work.cpp

@@ -22,6 +22,7 @@ std::shared_ptr<request_type> nano::distributed_work::peer_request::get_prepared
 
 nano::distributed_work::distributed_work (nano::node & node_a, nano::work_request const & request_a, std::chrono::seconds const & backoff_a) :
 node (node_a),
+node_w (node_a.shared ()),
 request (request_a),
 backoff (backoff_a),
 strand (node_a.io_ctx.get_executor ()),
@@ -35,23 +36,26 @@ elapsed (nano::timer_state::started, "distributed work generation timer")
 nano::distributed_work::~distributed_work ()
 {
 	assert (status != work_generation_status::ongoing);
-	if (!node.stopped && node.websocket_server && node.websocket_server->any_subscriber (nano::websocket::topic::work))
+	if (auto node_l = node_w.lock ())
 	{
-		nano::websocket::message_builder builder;
-		if (status == work_generation_status::success)
+		if (!node_l->stopped && node_l->websocket_server && node_l->websocket_server->any_subscriber (nano::websocket::topic::work))
 		{
-			node.websocket_server->broadcast (builder.work_generation (request.root, work_result, request.difficulty, node.network_params.network.publish_threshold, elapsed.value (), winner, bad_peers));
-		}
-		else if (status == work_generation_status::cancelled)
-		{
-			node.websocket_server->broadcast (builder.work_cancelled (request.root, request.difficulty, node.network_params.network.publish_threshold, elapsed.value (), bad_peers));
-		}
-		else if (status == work_generation_status::failure_local || status == work_generation_status::failure_peers)
-		{
-			node.websocket_server->broadcast (builder.work_failed (request.root, request.difficulty, node.network_params.network.publish_threshold, elapsed.value (), bad_peers));
+			nano::websocket::message_builder builder;
+			if (status == work_generation_status::success)
+			{
+				node_l->websocket_server->broadcast (builder.work_generation (request.root, work_result, request.difficulty, node_l->network_params.network.publish_threshold, elapsed.value (), winner, bad_peers));
+			}
+			else if (status == work_generation_status::cancelled)
+			{
+				node_l->websocket_server->broadcast (builder.work_cancelled (request.root, request.difficulty, node_l->network_params.network.publish_threshold, elapsed.value (), bad_peers));
+			}
+			else if (status == work_generation_status::failure_local || status == work_generation_status::failure_peers)
+			{
+				node_l->websocket_server->broadcast (builder.work_failed (request.root, request.difficulty, node_l->network_params.network.publish_threshold, elapsed.value (), bad_peers));
+			}
 		}
+		stop_once (true);
 	}
-	stop_once (true);
 }
 
 void nano::distributed_work::start ()

nano/node/distributed_work.hpp

@@ -86,6 +86,8 @@ class distributed_work final : public std::enable_shared_from_this<nano::distrib
 	void add_bad_peer (nano::tcp_endpoint const &);
 
 	nano::node & node;
+	// Only used in destructor, as the node reference can become invalid before distributed_work objects go out of scope
+	std::weak_ptr<nano::node> node_w;
 	nano::work_request request;
 
 	std::chrono::seconds backoff;

nano/rpc_test/rpc.cpp

@@ -3118,15 +3118,14 @@ TEST (rpc, work_peer_many)
 	node1.config.work_peers.push_back (std::make_pair (node3.network.endpoint ().address ().to_string (), rpc3.config.port));
 	node1.config.work_peers.push_back (std::make_pair (node4.network.endpoint ().address ().to_string (), rpc4.config.port));
 
-	for (auto i (0); i < 10; ++i)
+	std::array<std::atomic<uint64_t>, 10> works;
+	for (auto i (0); i < works.size (); ++i)
 	{
 		nano::keypair key1;
-		std::atomic<uint64_t> work (0);
-		node1.work_generate (key1.pub, [&work](boost::optional<uint64_t> work_a) {
-			ASSERT_TRUE (work_a.is_initialized ());
+		node1.work_generate (key1.pub, [& work = works[i]](boost::optional<uint64_t> work_a) {
 			work = *work_a;
 		});
-		while (nano::work_validate (key1.pub, work))
+		while (nano::work_validate (key1.pub, works[i]))
 		{
 			system1.poll ();
 			system2.poll ();