fix(pathfinder): Fix CRC computation of PathFinder::m_wallPieces

[Caball009]

Original EA code links

https://github.com/electronicarts/CnC_Generals_Zero_Hour/blob/main/Generals/Code/GameEngine/Include/GameLogic/AIPathfind.h#L846-L847

https://github.com/electronicarts/CnC_Generals_Zero_Hour/blob/main/Generals/Code/GameEngine/Source/GameLogic/AI/AIPathfind.cpp#L10395-L10398

https://github.com/electronicarts/CnC_Generals_Zero_Hour/blob/main/GeneralsMD/Code/GameEngine/Include/GameLogic/AIPathfind.h#L858-L859

https://github.com/electronicarts/CnC_Generals_Zero_Hour/blob/main/GeneralsMD/Code/GameEngine/Source/GameLogic/AI/AIPathfind.cpp#L11095-L11098

GeneralsGameCode/Core/GameEngine/Include/GameLogic/AIPathfind.h

Lines 908 to 909 in 64b5eaa

	ObjectID m_wallPieces[MAX_WALL_PIECES];
	Int m_numWallPieces;

GeneralsGameCode/Core/GameEngine/Source/GameLogic/AI/AIPathfind.cpp

Lines 11341 to 11344 in 64b5eaa

	for (Int i=0; i<MAX_WALL_PIECES; ++i)
	{
	xfer->xferObjectID(&m_wallPieces[MAX_WALL_PIECES]);
	}

This loop uses the address of the first element after the m_numWallPieces array, which is effectively the same as:

 for (Int i=0; i<MAX_WALL_PIECES; ++i) 
 { 
 	xfer->xferObjectID(&m_numWallPieces); 
 }

This should be a change that's not user facing, but I need to check some replays to verify that everything works as expected.

TODO:

• Check replays.

[greptile-apps]

Greptile Summary

This PR fixes a bug in Pathfinder::crc() where &m_wallPieces[MAX_WALL_PIECES] (an out-of-bounds address pointing to m_numWallPieces) was being passed to xferObjectID in a loop, effectively serializing m_numWallPieces 128 times instead of each wall piece. The fix gates the behavior behind RETAIL_COMPATIBLE_CRC: the legacy path replicates the original (buggy) behavior for save/replay compatibility, while the corrected path serializes the actual m_wallPieces array via xferUser.

The verification that xferObjectID and xferInt both delegate to xferImplementation(data, sizeof(T)) confirms the static_assert(sizeof(Int) == sizeof(ObjectID)) is sufficient to guarantee identical CRC output on the compatibility path.

Confidence Score: 5/5

Safe to merge; the fix is logically sound, the compatibility path is correctly guarded, and the static_assert prevents silent breakage if type sizes ever diverge.

No P0/P1 findings. The RETAIL_COMPATIBLE_CRC path faithfully replicates the original behavior (confirmed by reading Xfer.cpp — both xferObjectID and xferInt delegate to xferImplementation with the same size), and the corrected path properly serializes the actual array. The static_assert is a well-placed defensive guard.

No files require special attention.

Important Files Changed

Filename	Overview
Core/GameEngine/Source/GameLogic/AI/AIPathfind.cpp	Fixes out-of-bounds UB in the CRC loop by gating the legacy (buggy) behavior behind RETAIL_COMPATIBLE_CRC and introducing a correct xferUser path; static_assert guards type-size equivalence.

Flowchart

%%{init: {'theme': 'neutral'}}%%
flowchart TD
    A["Pathfinder::crc(Xfer *xfer)"] --> B["xferInt(&m_numWallPieces)"]
    B --> C{RETAIL_COMPATIBLE_CRC?}
    C -- Yes --> D["Loop 128x: xferInt(&m_numWallPieces)\n(replicates original buggy behavior)\nstatic_assert sizeof(Int)==sizeof(ObjectID)"]
    C -- No --> E["xferUser(m_wallPieces, sizeof(m_wallPieces))\n(correct: serializes all ObjectIDs)"]
    D --> F["CRCDEBUG_LOG m_wallPieces"]
    E --> F
    F --> G["xferReal(&m_wallHeight)"]

Loading

_{Reviews (2): Last reviewed commit: "Tweaked code and comment." | Re-trigger Greptile}

[Caball009]

Updated PR.