Scripts dont work after loading save

[Freeeaky]

Reported by Baaa, IRC

[zh0ul]

I've been asked about this innumerable times. Can ye provide some insight into this issue, or where to start digging?

It feels like GTALua simply loses its tick/hook, which likely means the game mem has shifted and somehow GTALua did not pick up on this fact and re-register with scripthook?

[Freeeaky]

What exactly happens when you load a save?

[zh0ul]

Repro steps:

• When using GTALua, vanilla or my own modified version...
• When using any mod (I have tried many configurations, but the cleanest was enabling 1 of the example mods and only that mod).
• Load GTA 5
• ESC -> Game -> Replay Mission or Replay Strangers and Freaks
• Pick/Load any mission.
• A 'Save Game' is created before loading mission (by GTA's own engine)
• At this point, we still have hook and menu is working fine.
• You can fail the mission and 'Retry', menu still works.
• Complete/Exit the mission (either successfully, or failed)
• At this point, the Save Game that was ^^ saved by GTA is reloaded. "Loading Story Mode"
• The menu itself stops working. (hotkeys, anything that was drawn on screen, etc)
• However, the console itself is working.
• When using console.cmd_show_all_addons() , main_thread is reported to be Active: true and Running: true
• However, The tick functions are not running (i.e. my own mod also shows as Active/Running, but no tick calls are being made).

[zh0ul]

• I note that the GTALua::UpdateLoop() (from Main.cpp) contains the following:

void GTALua::UpdateLoop()
{
  while (m_bActive)
  {
    Update();
  }
}

• However, comparing this to scripthookvdotnet, it appears what is lacking is 'What to do after m_bActive is no longer active'.

void ScriptMain()
{
        // Set up fibers
	sGameReloaded = true;
	sMainFib = GetCurrentFiber();

	if (sScriptFib == nullptr)
	{
		const auto callback = [](LPVOID)
		{
			while (ManagedInit())
			{
				sGameReloaded = false;

				// Run main loop
				while (!sGameReloaded && ManagedTick())
				{
					// Switch back to main script fiber used by Script Hook
					SwitchToFiber(sMainFib);
				}
			}
		};

		// Create our own fiber for the common language runtime once
		sScriptFib = CreateFiber(0, callback, nullptr);
	}

	while (true)
	{
		// Yield execution
		scriptWait(0);

		// Switch to our own fiber and wait for it to switch back
		SwitchToFiber(sScriptFib);
	}
}

• Lastly, I'm not a real C++ coder, I just play one on T.V.

[zh0ul]

• The last major diff between scripthookdotnet and GTALua appears to be in scriptRegister/Unregister and GTALua's lack of Unregister framework.
• Below is an excerpt from scripthookdotnet's DllMain.

BOOL WINAPI DllMain(HMODULE hModule, DWORD fdwReason, LPVOID lpvReserved)
{
	switch (fdwReason)
	{
		case DLL_PROCESS_ATTACH:
			DisableThreadLibraryCalls(hModule);
			scriptRegister(hModule, &ScriptMain);
			keyboardHandlerRegister(&ScriptKeyboardMessage);
			break;
		case DLL_PROCESS_DETACH:
			DeleteFiber(sScriptFib);
			scriptUnregister(hModule);
			keyboardHandlerUnregister(&ScriptKeyboardMessage);
			break;
	}

• And for GTALua, there appears to be scriptRegister for GTALua.asi, (found in LB_ScriptHookV.cpp) , but no steps defined for unregister/re-register

[Freeeaky]

I agree, it's probably related to not using scriptUnregister properly. Try finding out when m_bActive is false.

void GTALua::UpdateLoop() { while (m_bActive) { Update(); } MessageBox(0, "m_bActive is false", 0, 0); }

for example

[zh0ul]

• Since the console still works after such events, I went ahead and added a printf after the while loop.

void GTALua::UpdateLoop()
{
  while(m_bActive)
  {
    Update();
  }
  printf("m_bActive == false\n");
}

• Interestingly; after loading save, no message is displayed.
• Actually, after finally realizing that UpdateLoop() simply calls Update()..
• And all Update() does, is ProcessConsoleInput();
• And we know that the console still works after loading save.
• Then its safe to assume that m_bActive and this update loop are never losing their thread.
• In fact, m_bActive appears to only be set = false in the destructor class ~GTALua
• Thus, this bit of code appears to work as expected.
• At this point I'm looking into LB_MainScriptThread.cpp , and DLLMain.cpp

[Freeeaky]

That's interesting, i'll try looking into it too.