Multiple replays that mismatch with main branch

[Skyaero42]

The following three replays mismatch in the latest main branch (13 June 10 am), but not in 1.04 original.

Replays 365357 and 366648 mismatch almost at the end.
There is a small score difference between the original (top) and main (bottom)

Replay 367398 mismatches halfway
Edit: score screens for this one are the same, so it is likely an issue with the replay. This one can be disregarded.

Replays.zip

[Caball009]

365357 already mismatches in builds from a month ago. This is the oldest I can access on Github https://github.com/TheSuperHackers/GeneralsGameCode/actions/runs/15074131414

Edit: same goes for 366648.

[Caball009]

There may be something wrong with my local VC6 setup. I'd appreciate if anyone could verify that 365357 still mismatches at this point: https://github.com/TheSuperHackers/GeneralsGameCode/tree/dc0ff50b5fd587d11f10df9090e1131d6fa6be97

EDIT1: as far as I can tell for 365357 and 367398, it's this change that causes the mismatch: 8b7b047
EDIT2: the issue with 366648 seems more complex. 8b7b047 is the first commit I can find that causes a mismatch, but unlike the other two replays, this replay still mismatches without it.

[Caball009]

There are two changes that both cause a mismatch for 366648:

1. The same change as the others: 8b7b047 from [GEN][ZH] Fix uninitialized variables in AiPathFind::findGroundPath, AiPathFind::classifyFence #632
2. [GEN][ZH] Fix crash in StateMachine when a chained StateMachine update deletes the executor StateMachine #890 I don't understand what changes in this one could cause a mismatch, but it's what I found after a somewhat lengthy process of trial and error.

[xezon]

Please check if the following change avoids mismatching:

// RefCountPtr<StateMachine> refThis(this);

[Caball009]

365357 and 367398 play without mismatch now that the cellbounds initialization has been reverted.

Please check if the following change avoids mismatching:

// RefCountPtr refThis(this);

GeneralsGameCode/GeneralsMD/Code/GameEngine/Source/Common/StateMachine.cpp

Lines 437 to 443 in 8b5ed7d

	{
	// TheSuperHackers @bugfix xezon 20/05/2025 Defer the deletion of this state machine.
	// Calling m_currentState->update() can release this state machine in certain circumstances,
	// for example if something kills the entity of this state machine as a result of this state update.
	// See https://github.com/TheSuperHackers/GeneralsGameCode/issues/212
	RefCountPtr<StateMachine> refThis = RefCountPtr<StateMachine>::Create_AddRef(this);

I used the main branch and commented out this line, but 366648 still mismatches.

[xezon]

Ok that is an interesting observation.

I did an experimental change: Can you test if this branch also mismatches?

https://github.com/xezon/CnC_Generals_Zero_Hour/tree/xezon/wip-experimental-statemachine-refcount

[Caball009]

https://github.com/xezon/CnC_Generals_Zero_Hour/tree/xezon/wip-experimental-statemachine-refcount does not mismatch.

[xezon]

Ok that is great. Then we know what to do :-)