tooling(test): one command to test VoxelForge, with a commandlet-vs-game parity check

Tools/VoxelForgeTest.ps1 runs scenarios (canonical, owner, probe, perf, tests, parity) with the known-good flag sets, isolated user/config dirs, serialized launches, an owner-process guard, and a compact result.json + summary.txt instead of 300 MB logs that Codex has to stream. Answers the owner's question 'can the commandlet reflect in-game behaviour': -gameconfig=1 makes the commandlet use the authored settings, and the parity scenario compares per-tile geometry hashes between the commandlet export and the game's streaming path. Parity PASS: 64 commandlet tiles vs 810 game records, no differing tile.

First run of the tests scenario: 44 tests, 28 pass, 16 fail (8 composer, 6 op-stack equivalence, PassageLandsInOpenSpace, StrateParamBlendCoverage). Triage is the next round: stale expectations from the floor removal and the degrees change, versus real defects, especially the legacy-vs-op-stack equivalence ones.

Co-Authored-By: Claude Opus 5 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_01MmSoF64BbRFmk9C3p1VsBJ

Source/VoxelForge/Public/VoxelWorld.h

@@ -5,6 +5,7 @@
 
 #include "CoreMinimal.h"
 #include "GameFramework/Actor.h"
+#include "HAL/CriticalSection.h"
 #include <atomic>
 #include "VoxelTypes.h"
 #include "VoxelGenerator.h"
@@ -137,6 +138,21 @@ struct FChunkResult
     TArray<uint8> CaptureGrid;
 };
 
+/** One worker-produced mesh identity emitted only by the explicit parity harness. */
+struct FVoxelTileHashDumpRecord
+{
+    FVoxelTileKey Tile;
+    FIntVector OriginVoxels = FIntVector::ZeroValue;
+    int32 Step = 1;
+    int32 Cells = 0;
+    int32 BandChunkLo = MIN_int32;
+    int32 BandChunkHi = MAX_int32;
+    bool bSheetTile = false;
+    bool bEmpty = true;
+    int32 NumTriangles = 0;
+    FString GeometryHash;
+};
+
 UCLASS()
 class VOXELFORGE_API AVoxelWorld : public AActor
 {
@@ -512,6 +528,10 @@ private:
     void ObserveHeadlessStrateCrossingTest(const FVector& PlayerPosition, APawn* PlayerPawn);
     void MaybeFinishHeadlessStrateCrossingTest();
     void MaybeFinishHeadlessStreamingTest();
+    void RecordTileHash(const FVoxelTileKey& Tile, const FIntVector& OriginVoxels,
+                        int32 Step, int32 Cells, int32 BandChunkLo, int32 BandChunkHi,
+                        bool bSheetTile, const FVoxelMeshData& MeshData);
+    void WriteTileHashDump();
 
     // Packed shader params, recomputed each Tick. ALL meaningful data is in .xyz — a material Vector
     // Parameter only delivers float3 (RGB) into a Custom node (the alpha is dropped), so we never use .w.
@@ -977,6 +997,13 @@ public:
     double HeadlessStreamingTestBeginSeconds = 0.0;
     double HeadlessStreamingTestLastElapsedSeconds = 0.0;
 
+    // Explicit parity evidence. Workers append under the lock; EndPlay writes the compact JSON
+    // after ActiveTaskCount has reached zero, so the dump is a complete worker-side snapshot.
+    bool bTileHashDumpEnabled = false;
+    FString TileHashDumpPath;
+    FCriticalSection TileHashDumpMutex;
+    TArray<FVoxelTileHashDumpRecord> TileHashDumpRecords;
+
     // Real inter-strate crossing validation. This is enabled only by -voxel.TestStrateCrossing=1;
     // the route is copied from UVoxelStrateManager's generated passage and movement is supplied to
     // the ordinary pawn movement component. The initial placement onto the upper landing is test