0 Fork 1

mirror of https://github.com/jkdansereau/goldeneye-pc-port.git synced 2026-10-23 15:52:42 +02:00

Files

T

James Dansereau 9d118e0b13 M-13: 21-level sweep matrix + D123 (C1 fix) + D124 (C2 Jungle fix)

WS4 level sweep driven to completion: 12/21 solo levels PASS
(load + render + no-crash). Full matrix in docs/LEVEL-STATUS.md,
reusable runner tools_pc/level_sweep.sh, WS6 checklist docs/LEVEL-PLAYTEST.md.

Fixes:
- D123 (tools_pc/d88_propdefs.py): crash class C1 on 6 levels
  (Dam/Runway/Frigate/Statue/Streets/Cradle). D122 fallout - the widened
  Vehichle/AircraftRecord.ailist slot was zeroed instead of carrying its
  read-before-write integer AI-list id, so prop.c -> ailistFindById(0) ->
  GAILIST_AIM_AT_BOND -> ai() ran a CHR aim list with a NULL chr ->
  chrIsNotDeadOrShot(NULL). New OBJ_ID_WORDS arm. 4 levels -> PASS,
  Runway falls through to C2. BUNKER1/Silo unregressed.
- D124 (port/src/gimgfixup.c): crash class C2 on Jungle. Compiled
  explosion-DL pointer-sync keyed on an IMAGESEG marker that texLoad()
  had already erased -> compiled globalDL_0xNNN arrays kept link-time
  IMAGESEG words -> 0xABCDxxxx into fast3d on first explosion draw.
  Latent on every level. Detect the slot from the compiled array instead.

Diagnosed, NOT fixed (write-ups in PCPortResearch.md SF):
- C2-GDL (D124-Facility addendum): Facility/Runway - model texture-pool
  path yields a garbage tex->data; every model file loads at an odd
  bank base (mempAllocBytesInBank does no alignment). Suspect
  d43_emit.py/pcmodels sidecar texture-blob offsets (N64 8B vs PC 16B Gfx).
- C3/C6 (D125): Aztec/Bunker2/Surface2 - converted propDefs blob in RAM
  doesn't match d88_propdefs.py offline output (garbage after record 0);
  runtime sizepropdef() walk drifts ~100 records -> wrong door/modelid.
  d88_propdefs.py convert_stream + sizepropdef() PORT strides verified
  correct; suspect a dest-offset/pd_end bug in d88_emit.py propdefs emit.

Remaining: 9 crashes / 6 classes (C2-GDL, C3/C6, C2m Jungle G_MTX,
C5 Control BG portal, C4 Depot BG tile/room, C7 Surface1 sndSetupSound).
Priority order in docs/LEVEL-STATUS.md.

Also folds in the in-progress docs restructure (HANDOFF/AGENTS/CLAUDE,
PCPortResearch D121/D122 SF/SH entries, PLAN-linear-level-sweep,
HANDOFF-ARCHIVE, BACKLOG).

2026-08-29 16:36:18 -04:00

4.5 KiB

Raw Blame History

BRIEF C2-GDL — model/prop GDL runtime relocation misaligns dst (Facility, Runway)

Repo: REPO_ROOT (git master, Windows).

READ FIRST

docs/HANDOFF.md top, docs/PORT-LEARNINGS.md §B (16-byte PC Gfx vs 8-byte N64 — this is the class) + §D.
docs/PCPortResearch.md §F D124 (the C2 split — this brief is the unfinished Facility half), D80, D82, D83, D85, D95, D58, D50.6 (Gfx-stride / GDL history — jump via §F index).
AGENTS.md non-negotiables. rsp/graphics/gmain.s = GBI ground truth.
PD analogue: PD_PORT_CHECKOUT — its model-GDL / texture relocation path. Check before writing anything.

SYMPTOM (WS4 re-sweep — `docs/LEVEL-STATUS.md`)

Facility -level_34 (fault 0x72181ee8) and Runway -level_35 (fault 0x721b8ee8) both crash in import_texture_i8 (port/fast3d/gfx_pc.cpp:821) — fast3d fed a garbage G_SETTIMG w1.

Per the D124 investigation: the bad G_SETTIMG (fmt=4 siz=2 w=0, preceded by an unconsumed 0xba GE tex-macro opcode) sits in a model/prop GDL relocated at runtime by texLoadFromGdl() via sub_GAME_7F0762E0 (objecthandler_2.c:82) — NOT a room GDL, NOT the global bank. A GE_C2 probe showed texLoadFromGdl writing to non-16-byte-aligned dst (0x701eac01, 0x701eb7d5, …): the replacementgdl / name(=srcsize) offset arithmetic still mixes N64 8-byte and PC 16-byte Gfx strides, so converted commands land mid-slot and a subsequent G_SETTIMG w1 reads as garbage.

12 levels PASS — they either don't hit this model-GDL path or hit it with a GDL whose misalignment happens to be benign.

TASK

Pin the exact stride/offset error in the model-GDL runtime relocation and fix it narrowly (#ifdef PORT, N64 line under #else, §F-documented). Likely spots:

src/game/tex.c texLoadFromGdl() — src/dst pointer advance per Gfx.
src/game/objecthandler_2.c sub_GAME_7F0762E0 — replacementgdl / srcsize math, GDL buffer sizing.
src/game/model.c modelNodeReplaceGdl / rwdata GDL slot stride.
The 0xba (G_GEGBIMACRO_* / GE tex-load macro) decode — fast3d may need to consume its extra word (cf. gmain.s).

Use the existing GE_D86 / model-node probes and add a capped GE_C2GDL probe if needed.

FILES YOU MAY TOUCH

src/game/tex.c, src/game/objecthandler_2.c, src/game/model.c (narrow #ifdef PORT); port/fast3d/gfx_pc.cpp (only a guard / the 0xba consume, if that's the cause); new tools_pc/c2gdl_* probes; docs/PORT-LEARNINGS.md, docs/PCPortResearch.md §F (extend D124 or new Dxx), docs/LEVEL-STATUS.md. NOT src/game/propobj.c / prop.c / tools_pc/d88* (C3 agent owns those), NOT port/src/gimgfixup.c (D124-Jungle, done).

KNOWN-GOOD / RULED OUT

D123 (C1), D124-Jungle (gimgfixup.c) — landed, work.
12 levels PASS — the common GDL path is fine.
Global-bank / explosion-DL sync (D124) — separate, fixed.
HUD/text X-mirror (D114/D116) — parked, do NOT re-trace.

PRE-FLIGHT

export PATH="/c/msys64/mingw64/bin:$PATH" && ./build-pc.sh ntsc-final — green.
Repro ./build-pc/ge007.x86_64.exe -level_34. ge007.crash.log; addr2line -e build-pc/ge007.x86_64.exe -f -C 0x<PC>.
GE_PCDUMP="80-260:40" → ./ppm/; python tools_pc/pixcount.py.
timeout does NOT kill on Windows: ... &, sleep 24, taskkill //F //IM ge007.x86_64.exe.
gdb attach fast (gdb -batch -x cmds.txt -p <winpid>). HW watchpoint on the dst cursor in texLoadFromGdl catches the mis-advance.
Regen sidecars after any converter change (none expected here): python tools_pc/d43_emit.py ntsc-final && python tools_pc/d69_emit.py ntsc-final && python tools_pc/d88_emit.py ntsc-final --regen.

BUDGET

Structural / shared-infra → ~14 build→run→inspect cycles or ~2 h. On expiry: revert temp probes, write up the stride spec + best hypothesis in §F with a confidence rating + files touched. A good half-solved write-up is a deliverable.

CONSTRAINTS

No game-logic / control-flow changes. ABI/layout/format only, #ifdef PORT with N64 line under #else, §F-documented. Revert your own probes. Don't touch the C3 agent's files.

REPORT

Root cause — file:line, the exact stride/offset bug, why 12 levels are unaffected.
Fix diff or why-not.
Verification — -level_34 + -level_35 boot to ≥1 non-degenerate frame (pixcount.py > a few %); -level_09 + -level_20 unregressed (framediff.py); re-check Jungle -level_37 + Streets -level_29 (may improve).
Probes left in tree.
Confidence rating.
Generalisable quirk appended to docs/PORT-LEARNINGS.md.

4.5 KiB Raw Blame History