CydandClaude Fable 5
4cf09917ce
The previous commit's attribution was wrong and is corrected in the roadmap. WRONG: 'the crash is the skeleton walk's' rested on 0 crashes in 2 runs with the walk disabled -- a 25% coin flip presented as evidence. The oldest preserved dump has no [skl] line and ends on the old 'couldn't figure out how to MakeEntityRenderables' fallback, so it predates the walk and faults identically. WRONG: 'it lands at different points each time'. Every dump carries the same numbers to the byte (cr2 7000FA64, EIP 66D9). What varies is how far the log gets, not where the fault is. Resolving the address through btl4opt.map names it exactly: EulerAngles::operator=(const LinearMatrix&)+0x19, a read through the matrix reference. A binary scan finds all three call sites of that operator pass a stack local, so the 1.79GB pointer still has no static explanation -- that is now its own open item rather than a guess. Two theories killed cleanly by the new BT_STACK_LOG probe and a binary scan: ESP drift (drift=0 over 2002 frames; exactly one callee-cleans function exists in the whole binary) and an undersized stack (our PE and the shipped one have identical 1MB/8K geometry). What the probe found matters more: the application is parked in state=2, which is LoadingMission, not WaitingForLaunch. Priority 0 is where the interest manager queues renderer events during load, the gate needs that priority empty, and it never empties -- so the mission never launches and the renderer holds a blank screen by design. The fault arrives ~30s into that wait. Runs that DO launch never print a single [launch] line. So 'crashes half the time' and 'hangs during load' are one event seen twice. A 15x disagreement between two clocks looked like a reconstruction slip -- BTL4.CPP passes GetTicksPerSecond() where ApplicationManager wants a frame rate. Checked against the shipped binary before touching it: same instruction sequence, same kind of static float pushed. Authentic. Documented so nobody 'fixes' it. Also swept every subsystem DefaultData against its real C++ base. Fifteen chain past their immediate parent, but fourteen skip only classes that add no handlers and no attributes, and no class's attribute-ID base disagrees with its index chain -- so there are no gap slots there. The one real defect: Generator is a HeatSink but chained to Subsystem::MessageHandlers, so it ignored every ToggleCooling message. Fixed (compiles next build). Tooling: podrun.sh stages the build over BTL4REC.EXE, exports the host-side VPX board env -- without which the run dies at the iserver handshake rather than merely rendering nothing -- and archives every run's log, marking it -CRASH when it faulted. Before this the only preserved dump was an accident, on a rig where each run costs four minutes. Co-Authored-By: Claude Fable 5 <noreply@anthropic.com>
source410 — literal reconstruction of the missing BT 4.10 game source
Everything in this directory is RECONSTRUCTED, not recovered. The authentic surviving
source lives in CODE/ and is never mixed with this tree. The goal here is an
archival-grade recreation of the d:\tesla_bt\ translation units that exist only inside
BTL4OPT.EXE (the shipped 4.10 binary), written in the 1995 VWE house style against the
1995 MUNGA/MUNGA_L4 engine headers that DID survive.
Ground rules
- Never copy a reconstructed file into
CODE/— the assembled buildable tree (originals + reconstructions) is produced by a script into a build directory, so provenance stays unambiguous file-by-file. - Every
.CPPhere has a.NOTES.mdsidecar carrying the evidence: the binary address map, decomp references, which sibling sources supplied each idiom, and every judgment call with its tier ([T1] decomp-proven … [T4] style guess). The.CPPitself stays clean 1995 style — no modern annotations inside. - Assert line numbers are constraints. The binary's
Fail()sites record the original file+line (BTL4APP.CPP:400etc.); reconstructed files are shaped so those calls land on their recorded lines. A file that can't meet its line constraints yet says so in the sidecar. - Layout mirrors the original tree:
BT/(bt.lib, 36 TUs) andBT_L4/(btl4.lib, 13 TUs + BTL4.CPP), matching the survivingBT.MAK/BTL4.MAK.
What drives the work
- The per-TU manifest:
C:\VWE\BT411\reference\BT410_SOURCE_MANIFEST.md(tooltools/manifest410.py; method logphases/phase-03-bt410-source-manifest.md). - The decomp:
C:\VWE\BT411\reference\decomp\all\part_*.c(+section_dump.txtstrings). - The port's semantic reconstructions:
C:\VWE\BT411\game\reconstructed\(WinTesla-hosted; each literal file here re-hosts one back onto the 1995 engine APIs). - Style/idiom exemplars: surviving
CODE/BT/*.CPP,CODE/RP/RP_L4/*.CPP,CODE/RP/MUNGA*/**(the 1995 engine source itself).
Status (2026-07-19)
Compile verification is LIVE: ./compile410.sh --sweep (BC4.52 + the authentic OPT.MAK
flags; the period compiler is the reviewer). backdate.py mechanizes WinTesla→1995 header
back-dating (donors: C:\VWE\BT412\engine\MUNGA\*.h).
| item | state |
|---|---|
| THE WHOLE TREE COMPILES (2026-07-19) | ✅ 11/11: all 10 surviving originals + the BTL4APP.CPP pilot build clean under BC4.52 + authentic OPT flags (./compile410.sh --sweep) |
| BT/BTCNSL.HPP + BT/BTSCNRL.HPP | ✅ reconstructed — console wire IDs recovered from the binary (Killed=9, Damaged=10, ScoreUpdate=13, DeathWithoutHonor=15 [T1]; TeamScore=12 [T4 GUESS — verify]); see BTCNSL.NOTES.md |
| BT_L4/BTL4APP.CPP (pilot reconstruction) | ✅ compiles (18.5KB obj); 12/12 functions, Fail() at recorded line 400 |
| Staged header family (13: mech/mechsub/heat/powersub/mechweap/emitter/mechmppr/btplayer/projtile/missile + btl4mppr/btl4vid) | ✅ [T3] interface-proven by every surviving consumer; layouts parked in reserved[] blocks — see STAGED-HEADERS.NOTES.md + MECH-LAYOUT.md |
| MUNGA backfills (VDATA, ROTATION, VECTOR2D, AUDREND, APP, LAMP, GAUGREND, GAUGE) | ✅ compile-proven (BACKFILLS.NOTES.md) |
Next work package — the mech-family, REVISED (round-3 finding, 2026-07-19): the plan to "reconstruct mech.hpp first" was the wrong order. A full read of the port's mech.hpp shows it is a hybrid (1995 skeleton + Recon proxies + port-only members + relocated fields), and substantial regions of the binary's 0x854-byte Mech object remain UNMAPPED — the port declares "the remaining members up to sizeof==0x854" in its mech2-4 slices. A literal MECH.HPP that precedes those TU reconstructions would be fiction. Corrected order:
- MECH.HPP is the CAPSTONE, grown incrementally with each mech-family TU
reconstruction (mech.cpp → mech2-4 → subsystem TUs), starting as an interface +
known-ordered members with explicit
reserved[]filler for unmapped regions (sidecar-tracked), tightened as each TU pins its slice. - The dependent headers (
mechsub,mechmppr,btl4mppr,emitter+mechweap,btplayer,projtile,missile) ship WITH their TUs against the staged MECH.HPP. - The blocked originals (BTTOOL, GAUSS, PPC, BTREG) and the BTL4APP pilot compile green as those land — BTREG last (it includes the whole game-header world). Offset oracles for every stage: BT411 CLASSMAP.md, decomp-reference.md §3, and the port headers' static_assert-locked members.
Toolchain — IN HAND (2026-07-19)
Borland C++ 4.52, archived at ../../BORLAND/BC45/, chosen by byte-match: the fleet's own
CODE/RP/CW32.LIB is identical to 4.52's (and not 4.5's). BCC32/TLINK32/TLIB/MAKE run natively
on Win11. The shipped BTL4OPT.EXE recipe is CODE/BT/OPT.MAK (plain bcc32,
-DLBE4;DEBUG_LEVEL=0;DEBUG_STREAM=cout, -5 -a4 -ff -k- -V -Jg -x- -RT- + the full -O set,
PCH btopt.csm; link -Tpe -ax with c0x32.obj + dpmi32.lib + cw32.lib = Borland PowerPack
DPMI32, not Phar Lap TNT). Compile verification of reconstructed TUs is therefore possible
file-by-file; the current blocker for full closure is the 1995 engine header gap (vdata.hpp
first — only the drifted WinTesla VDATA.h survives, in BT412/engine). Include order for
builds: CODE/BT/* over CODE/RP/*, -DNO_PRECOMPILED_HEADERS until the mech-family headers
are reconstructed. Remaining gap: TASM32 (JOYSTICK.ASM only).
Link campaign (build410, 2026-07-19)
source410/build410.sh assembles BTL4OPT.EXE: merged-engine mass compile (152/152 green
— the ENTIRE surviving engine compiles), BT TUs (11/11), tlib libs (munga 1.25MB / mungal4
with the prebuilt-1995 objs / bt / btl4), then the authentic tlink32 link rooted by a probe
main (build scaffold). First true link run: 52 unresolved externals
(UNRESOLVED-LEDGER.txt) = the measured remaining gap. Kill-lists: l4gauge.cpp back-date
(14 widget statics), the missing-engine-body back-date wave (~24: rotation, player, team,
l4app, explode, dropzone, terrain, cultural, subsystm-statics, netclient...), the prebuilt
RANDOM.obj fold, and the staged game-TU statics blocks (MECH.CPP et al. — each TU file opens
with its binary-evidenced Derivation/SharedData statics and grows into the full body).
Link-lib arbitration: SOS = SOSDBXC+SOSMBXC (the 32-bit Borland-flat variants; SOSMW*=16-bit),
WATTCP excluded (it belongs to the 16-bit NetNub TSR — the game talks to it via shared
memory), filestrm = the RP-era pair (BT-tree pair is post-4.10 drift; override layer).