catch std::future_error after transport on_connect (#149)

CHANGELOG.md

@@ -4,6 +4,10 @@ All notable changes to this project are documented below.
 
 The format is based on [keep a changelog](http://keepachangelog.com/) and this project uses [semantic versioning](http://semver.org/).
 
+### [2.8.2] - [2023-10-31]
+### Fixed
+- Fixed a race condition that could happen during socket connection which would cause `std::future_error` to be thrown.
+
 ### [2.8.1] - [2023-05-25]
 ### Fixed
 - Fixed a memory issue with retrieving HTTP response codes.

core/core-rt/NRtClient.cpp

@@ -114,10 +114,10 @@ std::future<void> NRtClient::connectAsync(NSessionPtr session, bool createStatus
         // if old promise not ready, just complete it for the user.
         _connectPromise->set_value();
     }
-    catch(const std::future_error& e)
+    catch(const std::future_error&)
     {
         // if we get an exception here, it means the connect promise has completed already from a previous connect.
-        // this is very expected.
+        // this is very expected. there is no way to check from the promise itself if it has completed already.
     }
 
     // stomp the old promise
@@ -196,7 +196,16 @@ void NRtClient::onTransportConnected()
         _listener->onConnect();
     }
 
-    _connectPromise->set_value();
+    try
+    {
+        // signal to the user's future that the connection has completed.
+        _connectPromise->set_value();
+    }
+    catch (const std::future_error&)
+    {
+        // if we get an exception here, it means the connect promise has completed already from a previous connect.
+        // this can happen if the transport double fires or some other unexpected cases, like the user disconnecting while a connection is being made.
+    }
 }
 
 void NRtClient::onTransportDisconnected(const NRtClientDisconnectInfo& info)

test/src/realtime/test_lifecycle.cpp

@@ -76,7 +76,7 @@ namespace Nakama {
             bool threadedTick = true;
             NTest test(__func__, threadedTick);
             test.setTestTimeoutMs(15000);
-            bool hadConnectCallback = false;
+
             test.runTest();
 
             NSessionPtr session = test.client->authenticateCustomAsync(TestGuid::newGuid(), std::string(), true).get();
@@ -92,5 +92,29 @@ namespace Nakama {
 
             test.stopTest(true);
         }
+
+        void test_rt_connect_callback()
+        {
+            bool threadedTick = true;
+
+            NTest test(__func__, true);
+            test.runTest();
+
+            NSessionPtr session = test.client->authenticateCustomAsync(TestGuid::newGuid(), std::string(), true).get();
+
+            bool connected = false;
+
+            test.listener.setConnectCallback([&connected](){
+                connected = true;
+            });
+
+            test.rtClient->connect(session, true);
+
+            // try to trigger any issues with the underlying promise.
+            std::this_thread::sleep_for(std::chrono::milliseconds(5000));
+
+            test.stopTest(connected);
+
+        }
     }
 }

test/src/realtime/test_realtime.cpp

@@ -34,7 +34,7 @@ void test_rt_joinGroupChat();
 void test_rt_quickdestroy();
 void test_rt_rapiddisconnect();
 void test_rt_reconnect();
-
+void test_rt_connect_callback();
 
 void run_realtime_tests()
 {
@@ -52,6 +52,7 @@ void test_realtime()
 {
     // These tests are not protocol specific
     test_rt_rapiddisconnect();
+    test_rt_connect_callback();
     /// change to 10 iterations to trigger https://github.com/microsoft/libHttpClient/issues/698 bug
     for (int i = 0; i < 1; i++) {
         test_rt_reconnect();

version.cmake

@@ -1,4 +1,4 @@
 # Easy to update by CI scripts file containing our version as well as
 # dependent git repos to achieve reproducible builds
 
-set(LIBNAKAMA_VERSION 2.8.0)
+set(LIBNAKAMA_VERSION 2.8.2)