545092d376
phase-37: T3 ☑ — the tools: tools/struct_layout.py (the o32 layout engine + the canonical-type writer in the final style + the naming invariant), tools/restruct.py's full form (rungs S/S2/S+A/X/R/D/L, the ledger, inflight restore, selftest 48 + --real 53), delever_oracle's LINKED mode (the build's own ld on the candidate object; f3 known-true + negative; the snapshot guard that caught T2's contaminated build/ object), include/common.h's reinterpret macros (proven equal on cc1's assembly), tools/restruct_cycle.sh (detached); rung D banked on ov_SC04_011 (258 units: 242 canonical / 8 K&R marked / 16 kept with causes); SETUP §P37 S107, dictionary rows, kit corpus, decision log + accelerators P37 S107, cookbook §458 addendum, story §10 + retrospective §7 + timeline | R22 check-all: 218 passed, 0 failed of 218 (r22_t3c) | 🛑 T4 NEXT (the declaration layer; P6 rules check first)
2026-09-12 10:04:16 -06:00
0bd784c64e
phase-36: T5 (header step) — the GTE consolidation: tools/gte_consolidate.py (signatures by the build's own maspsx→as tail: bytes + operand counts + clobbers; one canonical text per signature under Sony's names, Sony's clobbers canonical even when every definition carries the steer; lever variants <name>_m tried as canonical first; direct statements → canonical calls; --sweep), include/gte_inline.h (50 macros for 9,102 definitions) included from common.h — R22 218/218; the census's gte-lever class + per-TU definition count + strict gate; the cycle's MODE=gte; dictionary + SETUP rows; kit corpus; log entry
2026-09-09 05:28:50 -06:00
0265712916
feat(phase-33): B1 make disc-extract — the rom->decoder step in the build: extract.py --expect-manifest (compare against the committed oracle, never write it; mismatch -> .run/extract/ + diffs, exit 1) + --allow-missing-audio (explicit PARTIAL for Track-1-only dumps); disc-extract = probe (0.7 s no-op) -> disc presence -> redump SHA1/CRC32 -> extract+compare -> verify (15.7 s full); extract/extract-all call it; check-env WARN-on-absent EXE + oracle self-consistency; help rewritten; the 4 splat preset headers TRACKED (clean keeps them); .gitignore re-tightened to H1 (dumps/*.bin, ghidra/, tools/psyq/, session archive/, ghidra-ext zips, brave.exe; EXE re-include dropped); controls: no-disc exit 2, truncated disc FAILs with the oracle untouched, regenerated tree byte-identical to the previous (1,801 files)
2026-09-06 19:02:34 -06:00
c8c9fd20c1
feat(phase-16): S5 driver+supervisor+safe-exit + NULL fix + sig_unify pipeline
...
- common.h: +NULL (byte-neutral; recovers 54/106 m2c CC1-fails that were just NULL-undeclared)
- tools/auto_driver.py: never-stop worklist loop (m2c -> sig_unify -> byte-gate -> propagate ->
commit), STOP-sentinel safe-exit at every fn boundary, heartbeat.json
- tools/auto_supervisor.sh (pure-bash babysitter, relaunch-on-crash, reap permuters, MCP-off),
auto_stop.sh (touch .run/auto/STOP), auto_status.sh (remote check-in)
- p16_known_answer.py: +--gate (whole-binary capability test)
- YIELD REALITY (honest, P9): macro+sig_unify whole-binary = 4/16 (25%) on KNOWN-matchable,
~1% on the unmatched hard tail. The permuter is the untested differentiator -> overnight test.
- CURRENT_PHASE: full night's findings + Drew's never-stop/safe-exit/graduated-ladder refinements
2026-06-18 23:15:31 -06:00
7141e9752b
feat(phase-16): S0 close + S2 foundation — m2c-compat in common.h (byte-verified)
...
- S0 (cookbook §15): ran m2c on real ov_SC01_077 stubs (R14). Finding: m2c
--valid-syntax output COMPILES via m2c_macros.h (M2C_FIELD = byte-faithful cast);
30/30 sampled targets use only byte-faithful macros. Compiling != matching: the
residual is regalloc/schedule -> decomp-permuter is the byte-closer; struct types
are an enhancer, not the sole gate (S3/GATE-B measures the lift). ML parked (owner).
- common.h: add s64/u64/f64 + M2C_UNK* typedefs + byte-faithful M2C_FIELD/M2C_BITWISE
macros (non-faithful m2c macros left UNDEFINED = early 'defer' signal). Byte-neutral:
main 143dbb89 (clean rebuild, R22), resident 8e17e02f, ov_SC01_077 d19c9580 all OK.
- CURRENT_PHASE.md: Phase 16 plan + S0-S9 gated tasks + Sun-afternoon timeline +
known-answer oracle test method.
2026-06-18 22:32:03 -06:00
8ffb7f0607
feat(phase-7): -O0 split + 4 matches (42 real) + libcd.h — session B checkpoint
...
- per-file -O0 mechanism: split text c-subseg into boot@-O0 (src/boot.c) + 800@-O2
via splat resegmentation + Makefile target-specific CC1FLAGS; the boot/main/
game-mode-dispatch module (0x80010000-0x800123F0) is -O0, not the pinned -O2
(per-module compiler mixing, SETUP §5.5). Regression-gated byte-identical.
- 4 byte-matches (38->42 real, make check BYTE-IDENTICAL throughout): GameModeDispatch
(-O0 register-ptr far member), DebugMenuHandler (-O0 reserved-slot local), CdQueueBusy
(-O2 if/else order + branch polarity), CdReadRequest (-O2 early-return fall-through)
- PsyQ infra: include/psyq/libcd.h (CdlLOC/CdlFILE from the .gdt) + 4 named symbols
(CdSearchFile/CdPosToInt/CdIntToPos/VSync) in symbols.us.txt; func_* stubs renamed
- loader cluster non-jtbl COMPLETE (6 matched + 2 drafted): NON_MATCHING drafts of
LoaderInitFileTable + ResourceLoadStateMachine (logically faithful, residuals in-source)
- tooling: progress.py/difficulty.py fixed for the multi-file split (glob src/*.c, all
asm dirs, skip forward-decls); deterministic; 42 real / 4 NM
- flywheel: matching-cookbook §6 (-O0 detection + idioms), §7 (PsyQ types/symbols), T4
- build byte-identical 143dbb89f34491258bbc27810d0a12ec8b43a8dd from a full clean cycle
2026-06-14 19:37:05 -06:00
5f937bdda5
feat(phase-6): compiler pinned + matching loop proven (14 matches) — LZSS to Phase 7
...
- splat.us.exe.yaml: text subseg asm->c (INCLUDE_ASM scaffold); include/common.h
(committed prelude); Makefile c-rule cpp->cc1->maspsx->as + -Map
- PIN (G8): gcc-2.7.2-psx -O2 -G0 -mips1 -mcpu=3000 -mgas -msoft-float -fgnu-linker
+ maspsx --aspsx-version=2.56 --expand-div; byte-exact on func_80018F20 +
instruction-identical across 3 idiom classes; --expand-div REQUIRED for div/rem
(SETUP.md §5.4 PINNED, Makefile MASPSX_FLAGS)
- harness: diff_settings.py (asm-differ mipsel object mode), tools/decompile.py
(m2c wrapper), tools/permuter/ (decomp-permuter harness + objdump shim), expected/
- 14 real functions matched (asm-differ score 0): func_80018F20 + 13 accessors;
+14 splat-auto empty no-ops; make check BYTE-IDENTICAL throughout
- 2 instruction-identical near-misses guarded NON_MATCHING (func_80015A74 scheduling,
func_80016714 phantom frame) -> decomp-permuter candidates
- flywheel: docs/matching-cookbook.md, memory + CLAUDE.md step 5 (consult+evolve), R16
- requirements-python.txt: venv freeze (pycparser<3.0 pinned — permuter needs plyparser)
- LZSS spike: needs jump-table-in-rodata workflow (102 tables, mixed blob) -> Phase 7
- venv: asm-differ + permuter deps installed
- bumps project version 1.5.0 -> 1.6.0
2026-06-14 14:47:01 -06:00
Drew T