Prevent players from disconnecting during the loading screen by increasing the timeout tolerance to 60 seconds for that period, fixes #5163.

The NetClient runs in the main thread, so any part of the loading screen
consuming several seconds makes that client timeout.
This is a workaround because threading the NetClient would have prevent
these timeouts, refs #3700.
Coutnerintuitively, since enet timeout tolerance is proportional to the
latency, the better the connection of the player, the more likely it was
to drop on gamestart.

This problem became very frequent in Alpha 23, at least due to the Aura
bugfix 583b6ec625, AIInterface being particularly slow and that not
having been disabled yet in the loading screen resulting in additional
10 second freezes during the loading screen, even on empty maps, refs
#5200, 8e168f85e6.

Differential Revision: https://code.wildfiregames.com/D1513
Based on patch by: causative
This was SVN commit r21842.

binaries/data/config/default.cfg

@@ -445,6 +445,7 @@ name_id = "0ad"
 duplicateplayernames = false      ; Rename joining player to "User (2)" if "User" is already connected, otherwise prohibit join.
 lateobservers = everyone          ; Allow observers to join the game after it started. Possible values: everyone, buddies, disabled.
 observerlimit = 8                 ; Prevent further observer joins in running games if this limit is reached
+gamestarttimeout = 60000          ; Don't disconnect clients timing out in the loading screen and rejoin process before exceeding this timeout.
 
 [overlay]
 fps = "false"                     ; Show frames per second in top right corner

source/network/NetClient.cpp

@@ -672,6 +672,8 @@ bool CNetClient::OnPlayerAssignment(void* context, CFsmEvent* event)
 	return true;
 }
 
+// This is called either when the host clicks the StartGame button or
+// if this client rejoins and finishes the download of the simstate.
 bool CNetClient::OnGameStart(void* context, CFsmEvent* event)
 {
 	ENSURE(event->GetType() == (uint)NMT_GAME_START);
@@ -680,6 +682,8 @@ bool CNetClient::OnGameStart(void* context, CFsmEvent* event)
 	JSContext* cx = client->GetScriptInterface().GetContext();
 	JSAutoRequest rq(cx);
 
+	client->m_Session->SetLongTimeout(true);
+
 	// Find the player assigned to our GUID
 	int player = -1;
 	if (client->m_PlayerAssignments.find(client->m_GUID) != client->m_PlayerAssignments.end())
@@ -820,15 +824,26 @@ bool CNetClient::OnClientsLoading(void *context, CFsmEvent *event)
 
 	CClientsLoadingMessage* message = (CClientsLoadingMessage*)event->GetParamRef();
 
-	std::vector<CStr> guids;
-	guids.reserve(message->m_Clients.size());
-	for (const CClientsLoadingMessage::S_m_Clients& client : message->m_Clients)
-		guids.push_back(client.m_GUID);
-
 	CNetClient* client = (CNetClient*)context;
 	JSContext* cx = client->GetScriptInterface().GetContext();
 	JSAutoRequest rq(cx);
 
+	bool finished = true;
+	std::vector<CStr> guids;
+	guids.reserve(message->m_Clients.size());
+	for (const CClientsLoadingMessage::S_m_Clients& mClient : message->m_Clients)
+	{
+		if (client->m_GUID == mClient.m_GUID)
+			finished = false;
+
+		guids.push_back(mClient.m_GUID);
+	}
+
+	// Disable the timeout here after processing the enet message, so as to ensure that the connection isn't currently
+	// timing out (as it is when just leaving the loading screen in LoadFinished).
+	if (finished)
+		client->m_Session->SetLongTimeout(false);
+
 	JS::RootedValue msg(cx);
 	client->GetScriptInterface().Eval("({ 'type':'clients-loading' })", &msg);
 	client->GetScriptInterface().SetProperty(msg, "guids", guids);
@@ -875,6 +890,9 @@ bool CNetClient::OnLoadedGame(void* context, CFsmEvent* event)
 	if (client->m_Rejoin)
 		client->SendRejoinedMessage();
 
+	// The last client to leave the loading screen didn't receive the CClientsLoadingMessage, so disable here.
+	client->m_Session->SetLongTimeout(false);
+
 	return true;
 }
 

source/network/NetServer.cpp

@@ -1095,6 +1095,9 @@ bool CNetServerWorker::OnAuthenticate(void* context, CFsmEvent* event)
 		// Assume session 0 is most likely the local player, so they're
 		// the most efficient client to request a copy from
 		CNetServerSession* sourceSession = server.m_Sessions.at(0);
+
+		session->SetLongTimeout(true);
+
 		sourceSession->GetFileTransferer().StartTask(
 			shared_ptr<CNetFileReceiveTask>(new CNetFileReceiveTask_ServerRejoin(server, newHostID))
 		);
@@ -1263,6 +1266,8 @@ bool CNetServerWorker::OnLoadedGame(void* context, CFsmEvent* event)
 	CNetServerSession* loadedSession = (CNetServerSession*)context;
 	CNetServerWorker& server = loadedSession->GetServer();
 
+	loadedSession->SetLongTimeout(false);
+
 	// We're in the loading state, so wait until every client has loaded
 	// before starting the game
 	ENSURE(server.m_State == SERVER_STATE_LOADING);
@@ -1360,6 +1365,8 @@ bool CNetServerWorker::OnRejoined(void* context, CFsmEvent* event)
 		session->SendMessage(&pausedMessage);
 	}
 
+	session->SetLongTimeout(false);
+
 	return true;
 }
 
@@ -1450,8 +1457,11 @@ void CNetServerWorker::StartGame()
 {
 	m_ServerTurnManager = new CNetServerTurnManager(*this);
 
-	for (const CNetServerSession* session : m_Sessions)
+	for (CNetServerSession* session : m_Sessions)
+	{
 		m_ServerTurnManager->InitialiseClient(session->GetHostID(), 0); // TODO: only for non-observers
+		session->SetLongTimeout(true);
+	}
 
 	m_State = SERVER_STATE_LOADING;
 

source/network/NetSession.cpp

@@ -23,6 +23,7 @@
 #include "NetStats.h"
 #include "lib/external_libraries/enet.h"
 #include "ps/CLogger.h"
+#include "ps/ConfigDB.h"
 #include "ps/Profile.h"
 #include "scriptinterface/ScriptInterface.h"
 
@@ -32,6 +33,25 @@ const u32 MAXIMUM_HOST_TIMEOUT = std::numeric_limits<u32>::max();
 
 static const int CHANNEL_COUNT = 1;
 
+// Only disable long timeouts after a packet from the remote enet peer has been processed.
+// Otherwise a long timeout can still be in progress when disabling it here.
+void SetEnetLongTimeout(ENetPeer* peer, bool isLocalClient, bool enabled)
+{
+#if (ENET_VERSION >= ENET_VERSION_CREATE(1, 3, 4))
+	if (!peer || isLocalClient)
+		return;
+
+	if (enabled)
+	{
+		u32 timeout;
+		CFG_GET_VAL("network.gamestarttimeout", timeout);
+		enet_peer_timeout(peer, 0, timeout, timeout);
+	}
+	else
+		enet_peer_timeout(peer, 0, 0, 0);
+#endif
+}
+
 CNetClientSession::CNetClientSession(CNetClient& client) :
 	m_Client(client), m_FileTransferer(this), m_Host(nullptr), m_Server(nullptr), m_Stats(nullptr), m_IsLocalClient(false)
 {
@@ -201,6 +221,11 @@ u32 CNetClientSession::GetMeanRTT() const
 	return m_Server->roundTripTime;
 }
 
+void CNetClientSession::SetLongTimeout(bool enabled)
+{
+	SetEnetLongTimeout(m_Server, m_IsLocalClient, enabled);
+}
+
 CNetServerSession::CNetServerSession(CNetServerWorker& server, ENetPeer* peer) :
 	m_Server(server), m_FileTransferer(this), m_Peer(peer), m_IsLocalClient(false), m_HostID(0), m_GUID(), m_UserName()
 {
@@ -261,3 +286,8 @@ void CNetServerSession::SetLocalClient(bool isLocalClient)
 	enet_peer_timeout(m_Peer, 0, MAXIMUM_HOST_TIMEOUT, MAXIMUM_HOST_TIMEOUT);
 #endif
 }
+
+void CNetServerSession::SetLongTimeout(bool enabled)
+{
+	SetEnetLongTimeout(m_Peer, m_IsLocalClient, enabled);
+}

source/network/NetSession.h

@@ -105,6 +105,12 @@ public:
 	 */
 	u32 GetMeanRTT() const;
 
+	/**
+	 * Allows increasing the timeout to prevent drops during an expensive operation,
+	 * and decreasing it back to normal afterwards.
+	 */
+	void SetLongTimeout(bool longTimeout);
+
 	CNetFileTransferer& GetFileTransferer() { return m_FileTransferer; }
 
 private:
@@ -183,6 +189,12 @@ public:
 	 */
 	void SetLocalClient(bool isLocalClient);
 
+	/**
+	 * Allows increasing the timeout to prevent drops during an expensive operation,
+	 * and decreasing it back to normal afterwards.
+	 */
+	void SetLongTimeout(bool longTimeout);
+
 	/**
 	 * Send a message to the client.
 	 */