A Box Map Under a New Map Name (game/wolfsdk/box)

Round-9 pass: can a TNC custom map use a brand-new map name? Static analysis of the runtime snapshot found no step that limits map names to the retail list — the name only has to be listed in dlc_0's packagemapspec.json. A box map is built as a third map in dlc_0's own chunk_22, next to c2v1 and c2v1_custom, and all build checks pass. Not yet seen in the game; the route also depends on the dev-menu command branch, which is itself still unconfirmed in game.

Date: 2026-10-02. Nothing was installed and the game was not started. Nothing was written under <Steam library> or Saved Games; the install was only read. Builds went only to D:/wolf-decomp/tmp/boxmap/. New files: tools/build_boxmap.py and this report.

Answer

Yes, a new map name works. Static analysis of the runtime snapshot (re_probes/runtime_87400/module.bin, same EXE SHA256 as installed) found no step that limits map names to the retail list. You only have to list the name in dlc/dlc_0/base/packagemapspec.json. The box map is therefore built as a third map in dlc_0's own chunk_22, next to c2v1 and c2v1_custom. Retail does the same thing: dlc_2's chunk_22 holds c2v1, c2v2 and c2v3. The c2v1 slot (kit level H) stays byte-identical, so the box costs no map slot. A second folder such as dlc_00 is worse because:

• it would need its own dlcid (SearchDLC drops a duplicate id, 0xabe2c6);
• it would need its own 886 MB texdb link, sound packs and pages.

Not yet seen in the game. Static analysis covers every step up to the spawn. The whole route depends on the dev menu command branch (mods/tnc_devmenu_fix), and that branch is itself still unconfirmed in game.

Research, step by step

"Read" means the result comes from disassembly or decompiled code of the snapshot. "Measured" means it was observed in data or in the game.

Gaps: 0xabf3d0 and 0xabf6a0 were read from disassembly, not run. SearchDLC (0xabdcf0) was not re-traced; it is taken from dlc4_cfile.md and r2_mapnodlc.md. There it is measured by log order and a screenshot-confirmed run.

What was built

python tools/build_boxmap.py takes about 2 minutes and builds twice. --once builds once. Output in D:/wolf-decomp/tmp/boxmap/box/:

The 29 entries:

• 17 map-level c2v1 entries copied under the new name:
  • maps/game/wolfsdk/box.entities
  • .../box/_combo/box.proc
  • .../box/box.prockd
  • generated/buildgame/game/wolfsdk/box.mapresources
  • _world.tome (0 B, as in H) and -metadata.json
  • box.ambientsh, box.hknavmesh, devmapinfo.json
  • mega/megatrans/megafoliagetrans.decl
  • _world.bmodel and sky.bmodel (40 B each in H)
  • _world.pvs
  • world.hkshape (c06p2's, as in H)
  • generated/staticstreamtree/game/wolfsdk/box.bsstree
• 12 new entries: a model and a cm for each of the 6 slabs.

The rename rule (maplevel()/rename()) takes the map's own entries:

• maps/<m>.entities, generated/buildgame|staticstreamtree/<m>.*;
• under maps/<m>/: _combo/*, <base>.*, devmapinfo.json, mega*.decl.

Per-entity resources stay under c2v1 (trigger cms, lightprobes, kiscules). The box has none of those entities.

The box itself is reused, not newly written code:

• Room: inside 8 × 8 × 4 m, slabs 0.25 m thick, floor top at (5, -183, -4.5). That is the kit hall floor inside c2v1's proc areas. c06p2's world collision is ≥ 295 m away.
• Slabs: build_gltfmap.pieces() builds them: r4 model writer, box cm with contents | 0x18 (variant D), megatexturedModel = false, material textures/common/blockout_gray, one .proc texel each (variant H recipe).
• .entities: H's worldspawn verbatim; player_start_1, cloned from c2v1_ps_challenge_01 with no layer, 2 m south of the centre and 0.1 m above the floor; wolfsdk_box_light, a clone of H's point light kitmap_light_hall_fill at 3.2 m, radius 9, warm white.
• .mapresources: H's 20,838 rows with the map-level rows renamed, the map list set to maps/game/wolfsdk/box, and 12 always-load rows.

Checks (all 31 OK, 0 FAIL; D:/wolf-decomp/tmp/boxmap/build_log.txt)

• Container:
  • Before finalize, probe_idclhash rejects the file (negative control).
  • Finalize moves the data hash of exactly the 29 added entries.
  • Entry count is 7,130 + 29.
  • All 7,130 base entries are unchanged (payload, version, flags, dependencies), so c2v1 is still kit level H.
  • The 29 entries are each present once, as written.
  • probe_datahash +0x50 7158/7158 (the 1 mode-4 entry is skipped, as in every build). probe_idclhash +0x08 1/1 +0x20 1/1.
• Naming: naming_problems() returns [] on the built container. It checks four things:
  • every map-level c2v1 entry has a copy under the new name;
  • no mapresources row of the box still names a map-level c2v1 resource;
  • every such row exists renamed, and the map list is the new name;
  • the box's text payloads do not contain game/dlc/c02/c2v1. The only allowed exception is the mega*.decl page reference (#9).

The controls each fail as they should: - A: box.ambientsh left under the old name only; - B: its mapresources row left on the old name; - C: the old map path inside the box .entities.

Separately, all 14 engine-derived names from the format strings exist under game/wolfsdk/box. _world.bshadows is missing on purpose, as it is in H, which loaded in game. - .entities: - byte-exact round trip; - build_custom_map.check finds nothing (996 entityDefs mounted through the new spec); - exactly 1 worldspawn, 1 start, 1 light and 6 static instances; - every slab model resolves, and every cm resolves "exact" to our own cm; - the start and the light are inside the room; - the 6 collision boxes equal the designed slabs (closed room). - .mapresources and .proc: base rows kept in order with only the map-level ones renamed, plus 12 always-load rows; the only unresolved map-level row is _world.bshadows, exactly as in H; the .proc keeps all base records and gains 1 record per slab. - Determinism: two builds give the same sha256.

Install (game closed; Freedom Chronicles still required)

wolfsdk/mapinstall.py and the devmenu builder belong to others, so here is what they would need:

1. mapinstall.py:
   • let install() write the box entry as a third spec map (or take the spec from the build);
   • add one hardlink: dlc_0/base/maps/game/wolfsdk/box/textures.cache → dlc_2/base/maps/game/dlc/c02/c2v1/textures.cache;
   • record both in wolfsdk_karte.json, so uninstall recognises them.

After that: python -m wolfsdk map-install D:\wolf-decomp\tmp\boxmap\box\dlc_0\base\chunk_22.resources. 2. tools/build_devmenu_fix.py: add the item below to devmenuoption/main and raise num from 3 to 4. Field order and quoting are the same as the existing campaign command entries.

item[3] = {
    devMenuDisplayName = "WolfSDK: Box";
    devMenuCommandStrings = {
        num = 7;
        item[0] = "si_spawnspot \"player_start_1\"";
        item[1] = "si_nextlayeractive \"layer/dlc/c2v1\"";
        item[2] = "si_filllayeractive \"\"";
        item[3] = "si_devprogression \"\"";
        item[4] = "si_checkpointName \"\"";
        item[5] = "si_gamemode 0";
        item[6] = "devmap game/wolfsdk/box";
    }
}

layer/dlc/c2v1 loads the same layered preload rows as in the kit-level tests. The box's own entities have no layer.

Manual route until then (PowerShell)

Revert the mods in the loader first: mapinstall forbids creating files in dlc_0 while tncpatch journals it.

$g = "<Steam library>\steamapps\common\Wolfenstein.II.The.New.Colossus\dlc"
$b = "D:\wolf-decomp\tmp\boxmap\box\dlc_0\base"
Copy-Item "$g\dlc_0\base\packagemapspec.json" "D:\wolf-decomp\tmp\boxmap\spec_before.json"
Copy-Item "$b\chunk_22.resources" "$g\dlc_0\base\chunk_22.resources.new"
Move-Item -Force "$g\dlc_0\base\chunk_22.resources.new" "$g\dlc_0\base\chunk_22.resources"
Remove-Item "$g\dlc_0\base\packagemapspec.json"; Copy-Item "$b\packagemapspec.json" "$g\dlc_0\base\packagemapspec.json"
New-Item -ItemType Directory -Force "$g\dlc_0\base\maps\game\wolfsdk\box" | Out-Null
New-Item -ItemType HardLink -Path "$g\dlc_0\base\maps\game\wolfsdk\box\textures.cache" -Target "$g\dlc_2\base\maps\game\dlc\c02\c2v1\textures.cache"

Never open the hardlinked files for writing. To replace one, delete it first and then copy. Undo:

• copy re_probes\kitmap\variants\H\chunk_22.resources back the same way;
• put spec_before.json back;
• Remove-Item -Recurse "$g\dlc_0\base\maps\game". This only unlinks the link; dlc_2's file stays.

Until it is restored, the loader's uninstall will name the changed chunk and spec as unknown.

In-game test

1. Apply the mods (with the box item in tnc_devmenu_fix) and start the game. Open the dev menu → "WolfSDK: console test". The console must show g_debugLayerChanges = 1. If it does not, the command branch does not work and the box test means nothing.
2. Dev menu → "WolfSDK: Box".
   • Pass: you spawn in a closed grey room (8 × 8 × 4 m) lit by one warm light, can walk, do not fall, and cannot leave through the walls.
3. Afterwards type condump boxmap1.txt in the console and send the file from Saved Games/.../base/. Lines that explain a failure:

Falling through the floor means a collision problem. A black room means a lighting problem.

Variant B if campaign mode misbehaves: use si_gamemode 5 plus an extra line si_mainmenustart 1, which is what the c2v1 entries use.

Not proven (worst first)

1. The dev menu command branch, and devmap itself, have never run in game. All of r8's analysis is static.
2. A non-retail map name has never been loaded: steps 1–12 are static only.
3. New model/cm entries have never been loaded in game. This is DESIGN.md risk 1, the same as the build_gltfmap build.
4. Campaign mode (si_gamemode 0) with no chapter part on a DLC-chunk map:
   • the player becomes the generic player entityDef, with that decl's default inventory;
   • whether the game makes an initial autosave for a chapterless map is unknown.
5. Lighting: borrowed lightmap texels plus one dynamic light. The room may look blotchy or dark (variant G looked blotchy).
6. No map sound banks (the warning from step 10). Ambience will be silent.
7. Static reading: the heap-stack leak when textures.cache is missing is inferred from decompiled code. The link avoids that case, so it is not tested.

No fallback into the c2v1 slot was built, because the new name is not ruled out. If the box fails at steps 2–6 (one of the "not present", "not installed" or "cache file" lines above), the same pieces can go into c2v1.entities the way the kit level does, launched with the existing "Death is Coming (C2V1)" dev menu entry.

Related: Research Report: Chunklist, Research Report: Dlc4 Cfile, Research Report: R2 Mapnodlc, Research Report: R8 Devmenu Fix Verify, Research Report: R10 Modtest Map, TNC Custom Map Status, DLC Overrides.