Code actually doesn't need to be ran in the server frame thread. All the code really does is preparation work. Run it in the main thread but do join the server frame thread (FCVAR_SERVER_FRAME_THREAD) as we can't do concurrent work on the server VM.
Previously we did sq_pushroottable() and a subsequent sq_call() after compiling the text buffer, but this didn't work for code that was threaded, or using Get/SetNetVar* functions.
The second issue was that the callback for the "script" command was ran in the main thread. Server script should always run in the server frame thread, the Set/GetNetVar* functions check thread id to retrieve the correct VM context, so running server script from the main thread ended up with Set/GetNetVar* functions retrieving the client VM context rather than server's, causing undefined behavior.
Script commands are now queued to the server frame thread, ultimately fixing this bug.
Also fixed a small bug with function 'sq_compilebuffer()'; it takes an extra argument but this wasn't taken into account in the SDK.
Decoupled from net_usesocketforloopback since we actually don't want this to be tied with that of the game. Now it by default does not bind to the loopback socket unlike the game.
Can be toggled with the new cvar 'host_autoReloadRespectGameState', and used in combination with the new server script func 'SetAutoReloadState( bool state )'. This makes sure that even when the timer reaches 'host_autoReloadRate', it would wait with the reload until the game itself is finished (which is when SetAutoReloadState( true ) is being called from scripts).
Implement UserCmd command backlog limiting (the new convar 'sv_maxUserCmdProcessTicks' dictates how many ticks can be processed per second). Defaulted to 10, which is (default tick interval (0.05) * default cvar val (10) = 0.5ms window), which is equal to the default of cvar 'sv_maxunlag'.
Before this patch, you could stuff several seconds worth of usercmd's in one second and achieve speed hacking.
RCON lacked encryption, added AES-CTR encryption on RCON frames. Slightly adjusted protocol to take this into account (sending nonces, encrypted data itself, etc).
Marked convars/concommands as FCVAR_SERVER_FRAME_THREAD and removed main thread dispatching code and comments that are no longer in effect. In the RCONServer execute handler, a ThreadJoinServerJob() is placed if the commandbase is flagged FCVAR_SERVER_FRAME_THREAD since RCON dispatches the command, or sets the convar directly.
The WebSocket system is now fully split up. The global network module initialization is moved to DLL init so the websocket class could be used for anything. The LiveAPI system now also mostly runs in the server frame thread, the convar change callbacks still need to be latched to the server frame thread however.
If the code was ran in the main thread, it would never fire the actual check logic. This patch changed the code so it would always fire from the start of the frame. Also fixed a memory leak that occurred when passing in an external banned list to SV_CheckClientsForBan (previously, 'SV_CheckForBan'); the code always allocated a new banned list, but only freed it if no external banned list was provided. This patch changed the code so it only allocates it if no external one is provided.
This patch splits host logic from CServerListManager. CServerListManager is actually meant for the client to manage the server list to which the client could connect to. The hosting logic has been moved to the new CServerHostManager class.
Previously, we stored all the hosting details in CServerListManager, with connection criteria in CPylon, this data has been moved over to CServerHostManager as well.
Previously, we also needed a mutex to access the server host data, function HostState_KeepAlive() has been refactored to the point this mutex is no longer necessary as the only threaded process is the actual request, the rest is being applied in the main thread. We also now only construct a NetGameServer_t struct if we actually plan to host.
Access to CPylon::m_Language is now also protected by a mutex, as the change callback of cvar 'language' and the threaded method 'CPylon::QueryServer()' are competing for access.
Moved implementation to dedicated class 'CImguiSystem' with the following changes:
* Added mutex for ImGui_ImplWin32_WndProcHandler(), the window proc is ran from a different thread than CImguiSystem::SampleFrame() is being called from. Needs a mutex as both functions interact with the input queue list.
* Added init stages for debugging and preventing the system from running certain parts of the code if a certain stage hasn't been reached yet.
* NULL out ImGuiContext::ConfigNavWindowingKeyNext , we don't need any nav window shortcuts as users don't need then in-game, and it also interferes with certain key binds some users have.
* CImguiSystem::SampleFrame() (previously ImguiSystem_SampleFrame()), is now being called from CMaterialSystem::SwapBuffers(), as that function is called from the main thread while also getting updated during level loads/etc, so instead of switching between 2 static buffers for as long as the levels are being loaded, we imgui updates properly too and directly swap afterwards.
* Removed hack in CGame::WindowProc() where we would only run windowproc if we aren't shutting the game down. This was needed as there was a chance of a crash at some rare cases, but this was simply due to a missing mutex which has been addressed and fixed as of this commit.
* Lock mutex during init and shutdown.
ImGui drawing code now takes place in the main thread, a snapshot of the render data is created in CMaterialSystem::SwapBuffers(), and is being rendered in the render thread right before SpinPresent().
The reason why this was necessary, is because ConVar::GetString() isn't thread safe if its not marked FCVAR_MATERIAL_SYSTEM_THREAD or FCVAR_ACCESSIBLE_FROM_THREADS, and we used it for the console suggestions window, which iterates over every ConVar, accessible from threads or not.
Fully implemented ConVar class so we could statically construct all SDK convars, this avoids a level of indirection, and allows for creating ConVar's everywhere in the project.
This patch also removed the settings tab of the ImGui server browser, as it has threading issues, while it technically never caused a crash yet, it has been removed as there was no point keeping it vs the work required to make it thread save (it only managed 2 convars which are perfectly manageable through cfg's or the in-game console).
Also temporarily disabled the creation of ConVar's in the mod system due to a memory leak, we would allocate and register a convar based on details parsed out of a mod file definition, but never unregister and free it.
Properly implement the ConCommandBase and ConCommand classes so we could statically construct all ConCommand objects in the global scope of each translation unit, this way we don't need to put them in a global file and deal with preprocessor directives to compile then in/out for certain projects.
Add missing types to fully mapped out PakGlobals_s struct, which is a 13MiB+ structure! This covers the vast majority of the pakfile system which is why we had to change a bunch of stuff for this patch. This patch also comes with:
- Reversed 'JobFifoLock_s' structure
- Reversed 'PakTracker_s' structure
- Reversed 'PakAssetTracker_s' structure
Many globals have been dropped as they were covered by the large PakGlobals_s singleton.
The pak decoder logic has been changed up as well, we now use a decode mode enumerant which will make it easier to add in more decoders for the pak files in the future.
