Joe DiPrimaandClaude Fable 5
dca2586aa8
Eight byte-identical field stacks from night 6, all one player, all in an
Owens: ParticleEngine::Destroy +0x11, access=0 target=0x0, from the plain
per-frame render path. Nothing in the stack touches weapons or the Owens.
Conn Man's Surface Pro 9 (Iris Xe, 128 MB shared) is simply the only GPU in
the fleet that ever actually LOSES the D3D9 device -- his two-trigger
missile+laser bursts are what provoke the timeout, not what crashes.
What crashed is our device-loss handling, which was wrong three ways at once,
in two inline copies (the scene Present and the wait-screen Present):
1. On D3DERR_DEVICELOST it called Reset() IMMEDIATELY. Reset on a
still-lost device ALWAYS fails, and V() only logs. There was no
TestCooperativeLevel gate at all.
2. It then ran ParticleEngine::Initialize against the lost device. The
creates fail there and NULL their out-params -- proven, not assumed:
the bench repro faults at target=0x0, not at a dangling address.
3. The next lost frame called ParticleEngine::Destroy again, which
Release()d those NULLs blind. Read of vtable at 0x0. Dead.
So: lost frame 1 tears down and leaves NULLs, lost frame 2 crashes. Two
frames, every time, deterministic -- which is exactly why all 8 field stacks
are byte-identical.
Reproduced before fixing. BT_DEVICELOST_TEST=<frame>,crashrepro runs the
field sequence on the bench; on the unfixed build it died at Destroy +0x11,
access=0 target=0x0, and symbolized to the same four frames as the field
logs. Same shape, same offsets-modulo-hook. That run also proved the
out-param-nulling assumption the whole diagnosis rested on.
The fix -- one shared DPLRenderer::BTResetLostDevice() replacing both inline
copies:
- Destroy() is idempotent and null-safe, and nulls after release.
- Reset() is gated on TestCooperativeLevel() != D3DERR_DEVICELOST; while
the driver still says lost, skip the frame and retry.
- The Reset HRESULT is checked; on failure, log and retry next frame
instead of driving on.
- On success, re-create via the new CreateDeviceObjects(), NOT
Initialize(): Initialize memsets the installed-effects table, so every
reset that DID succeed silently killed all particle effects for the rest
of the mission. The quieter sibling bug, fixed by the same split.
- Initialize checks its HRESULTs and defends MAXPARTICLES<=0; the draw
paths guard the NULL buffer, and ExecuteParticles keeps draining
particles while the engine is dormant so they cannot pile up.
Verified: the crashrepro shape now logs SURVIVED and play continues; three
forced full loss/reset cycles each log "[render] device reset OK"; a plain
run is assert-free.
Found while verifying, worth its own line: VIDEO\particles.png has NEVER
existed -- not in the tree, not in BTL4.RES, not anywhere in git history.
The texture load has failed on every machine since the engine was written,
and every billboard particle ever rendered was untextured quads via
SetTexture(0, NULL). RenderParticles deliberately does NOT gate on the
texture -- that would disable all particles everywhere; untextured IS the
shipped look. Filed separately; a real particle sheet is a content task.
The field verification that counts is Conn Man flying his exact crash
loadout on this build: instead of a dead process he should see at worst a
brief hitch and "[render] device reset OK" in his log. #35 stays open until
that happens.
Co-Authored-By: Claude Fable 5 <noreply@anthropic.com>
BattleTech 4.11 (bt411)
A standalone Windows port of Virtual World Entertainment's arcade BattleTech (Tesla platform, release 4.10, ~1995–96), reconstructed on the shared RP411 Windows engine. The game boots, renders, and runs a single-player drive → animate → target → fire → damage → destroy loop across all 8 maps, with two-instance multiplayer entity replication working.
This repo is a clean, self-contained extraction of the BattleTech-specific work from the larger reverse-engineering workspace — engine + game + content + build, with nothing from Red Planet or the raw archive dumps. It builds and runs out of the box.
License: the game content (
content/) and the original binary are proprietary to Virtual World / the pod owner. This repository is private; do not redistribute.
Versioning
4.10 = the 1995 arcade release. 4.11 = this win32 reconstruction. Dev builds are
4.11.<build> where <build> is the git commit count — monotonic and zero-maintenance —
plus the short commit hash (4.11.311 (980c9cd)); a trailing + on the hash means the exe was
built from an uncommitted working tree. The stamp regenerates every build
(tools/btversion.cmake → build/btversion.h) and shows in the boot banner (btl4.log line 2)
and the window title. To identify any player's build, ask for the title bar or the top of their
btl4.log.
Layout
CMakeLists.txt one build: munga_engine lib + bt410_l4 game lib + btl4.exe
engine/
MUNGA/ shared 2007 sim/render engine (149 .cpp + headers)
MUNGA_L4/ Win32/D3D9 HAL + renderer + asset loaders (44 .cpp), incl.
our BT work: bgfload / L4D3D / L4VIDEO + the image codec
shim/ minimal ATL shim (USES_CONVERSION/W2A)
lib/ OpenAL32 / libsndfile import libs + runtime DLLs
game/
reconstructed/ the reconstructed BT game logic (mech, subsystems, HUD, app; ~47 .cpp)
original/BT,BT_L4 surviving original BT source + all BT headers
fwd/ header shims forwarding <NAME.hpp> -> the engine's NAME.h
btl4main.cpp WinMain launcher / entry point
content/ runtime data: BTL4.RES, VIDEO/, GAUGE/, AUDIO/, *.EGG, BTDPL.INI
context/ progressive knowledge graph — 18 on-demand topic files (routed by CLAUDE.md)
docs/ format specs + reconstruction ledgers + PROGRESS_LOG.md (full history)
reference/
decomp/ raw Ghidra pseudocode — source-of-truth for ongoing recon
ghidra_scripts/ the headless decomp exporter
glossary.yaml term / acronym definitions
phases/ restructuring / investigation logs
tools/ btconsole.py (MP console emulator), map/resource scanners
run/ run.cmd helper
CLAUDE.md knowledge-base ROUTER — identity, protocols, quick-lookup, conventions
Prerequisites
- Visual Studio 2019 BuildTools (MSVC v142, x86). The Community install on the original dev box was broken, hence the explicit BuildTools instance in the configure line below; adjust to your install.
- CMake ≥ 3.20.
- Legacy DirectX SDK (June 2010) — the engine uses
d3dx9/dinput/dxerr, removed from the modern Windows SDK. Default pathC:/Program Files (x86)/Microsoft DirectX SDK (June 2010); override with-DDXSDK=<path>. (The installer may throw a harmless S1023 error — dismiss it; the SDK headers/libs install before the failing redist step.)
OpenAL/libsndfile import libs + DLLs are vendored under engine/lib/; the DLLs are copied next to
the exe automatically at build time.
Build (32-bit / Win32)
cmake -S . -B build -G "Visual Studio 16 2019" -A Win32 ^
-DCMAKE_GENERATOR_INSTANCE="C:/Program Files (x86)/Microsoft Visual Studio/2019/BuildTools"
cmake --build build --config Debug
Must be Win32 — the DirectX SDK link libs are Lib/x86. The link uses /FORCE: the 1995
headers define free functions/globals without inline/extern, so identical symbols appear in
many translation units (~124 LNK2005); /FORCE:MULTIPLE keeps the first. UNRESOLVED tolerates
a dead offline-tool factory in mech3.cpp that is never called at runtime. (Cleanup task: move
those definitions to single TUs + neutralize the dead factory, then drop /FORCE.)
Run
run\run.cmd REM boots DEV.EGG (grass / day)
run\run.cmd DBASE.EGG REM any egg in content/
The working directory must be content/ (the engine resolves BTL4.RES, VIDEO\,
BTDPL.INI, and eggs relative to cwd); run.cmd handles that. Maps available in BTL4.RES:
cavern grass rav polar3 polar4 arena1 arena2 dbase — switch via a copied egg's map= field.
Useful env-var flags (default OFF unless noted)
The authentic stack (gait, collision, real controls) is default-on; set =0 to fall back.
Debug/harness flags: BT_FORCE_THROTTLE=1 (auto-walk), BT_SPAWN_ENEMY=1 (spawn a target dummy),
BT_FORCE_FIRE=1 (auto-fire), BT_HEAPCHECK=1 (whole-heap validation — slow), BT_BSL=0 (legacy
texture decode), BT_DEV_GAUGES=1 (render the cockpit MFDs in a dev window), BT_LOG=<file>.
Interactive: WASD drive, A/D turn, Q/E torso twist, R/F torso pitch aim,
Space / 1-4 fire, X all-stop, V view, M control mode. An Xbox-type controller
works out of the box. All bindings are user-editable in content/CONTROLS.MAP (delete it to
restore the compiled-in WASD default; content/CONTROLS_NUMPAD.MAP is an alternate profile). The
complete env-gate table is in context/decomp-reference.md §6 (routed from CLAUDE.md);
controls/pad details in context/pod-hardware.md.
Multiplayer
Modern path (relay + operator console). Pods make ONE outbound connection to a relay/console, so internet play needs no per-player port forwarding and CGNAT-safe LAN discovery just works. Run the operator station:
python tools/btoperator.py # PySide6 GUI: build the mission egg (validated dropdowns),
# start the relay, watch pods arrive, assign seats, LAUNCH,
# end the timed mission, export player join.bat scripts
Players run one universal join.bat (internet, seat assigned by the relay) or join_lan.bat
(same LAN, auto-discovers the console) or play_solo.bat (offline practice) — see players/.
The pod waits patiently if the session isn't up yet. Full architecture, wire format, and the
D1 relay/UDP design: context/multiplayer.md.
Legacy mesh (two instances, one box), still supported:
instance A: btl4.exe -egg MP.EGG -net 1501 (BT_LOG=mp_a.log)
instance B: btl4.exe -net 1601 (BT_LOG=mp_b.log)
console: python tools/btconsole.py MP.EGG 127.0.0.1:1501 127.0.0.1:1601
-net <port> enables networked mode. Verified end-to-end: full entity/movement replication,
cross-pod combat + kills, per-pilot paint + callsigns, timed missions, 4-pod live sessions, and a
spectator/broadcast camera seat (hostType=1 vehicle=camera) with auto-directed coverage and
a live ranking window.
Status & continuing the work
The engine, renderer, audio, HAL, build, locomotion, collision, damage, render fidelity, the full
cockpit gauge / MFD system (every config binding resolves + every widget builds), and the
projectile / missile weapon families are done. Active fronts: per-subsystem polish (the gyroscope
integrator; the 0xBD3 message manager that gates the valve / status-message control routes) and
cross-pod MP combat. reference/decomp/ holds the raw pseudocode every reconstruction is verified
against.
Start with CLAUDE.md — it is the router into the progressive knowledge base: a quick-lookup
table pointing to the context/*.md topic files (loaded on demand), the evidence-tier and
convention rules, and context/open-questions.md for what's deferred / next. The complete
pre-restructure history is preserved verbatim in docs/PROGRESS_LOG.md; docs/*.md holds the
detailed running ledgers.