99ae0afe17
docs(phase-31): S63 T5.2/T5.3 — lane G 23 banked free; wave t5a 43/48 (89.6%) incl. 2 recovered; the (binary,fn) card defect measured at 48/48 and fixed; falsifier cleared 6x
2026-08-26 20:21:07 -06:00
41b259e528
fix(wave): cards keyed by (binary,fn) — t5a received 48/48 ANOTHER binary's card (R48); t5_cards.py builds the target's OWN fuel (tu_ref 88% / decl_prior 98%) via wave_card_fuel.py extracted from build_wave_atlas (R33); t5_bank.sh recovers un-banked drafts via fix_tu_ret_decls (2/2 on t5a); distill workflow generalized (P31 S63 T5)
2026-08-26 20:19:41 -06:00
827b0b156f
feat(t5): wave t5a recovery — 2 banked (func_8017EFC4, func_80185380); both byte-correct bodies refused only by a TU 'extern void' decl on an s32-returning fn, retyped by fix_tu_ret_decls (P31 S63 T5.3)
2026-08-26 20:14:26 -06:00
bfd591c9cc
feat(t5): Claude wave t5a — 41 banked (sonnet 39/44, opus 2/4); fleet 213 passed, 0 failed of 213 (P31 S62 T5)
2026-08-26 20:12:09 -06:00
3591ac6566
feat(decomp): free A-prop sibling lane (maintenance pass) — zero model tokens
2026-08-26 19:42:23 -06:00
f8cc4c77d3
feat(tools): T5 wave draw/bank halves — t5_targets.py (routed, ledgered, band-stratified draw) + t5_bank.sh (judge → R22 sweep → commit); claude_wave_draft.js generalized; packs/judge refuse duplicate fn names (R43/R48); sweep #8 213/213 (P31 S63 T5.1)
2026-08-26 19:35:54 -06:00
20e7005855
docs(phase-31): S62 FINAL checkpoint — no-DeepSeek decision, T5 first-wave recipe on the promoted harness, residue ledger, rule candidates
2026-08-26 19:18:23 -06:00
5962bb70ed
fix(api_agent): prior_draft accepts a warm-start body only when its symbols overlap the target .s (law 1c) — 22/60 T4 agents had received another overlay's same-named function; fleet ledger regenerated (2,380 stubs) (P31 S62)
2026-08-26 19:02:07 -06:00
6a64eee98f
fix(tools): wave_judge.py takes the wave dir from WAVE (default .run/t4)
2026-08-26 19:01:57 -06:00
74ec441d60
feat(tools): the Claude-subagent wave harness promoted from the T4 probe — claude_wave_packs.py (identical api_agent packs), workflows/claude_wave_draft.js + claude_wave_distill.js (the 60-agent draft and the R16 distill scripts), wave_judge.py (per-arm whole-binary gate with git-diff bank detection, tree reset, --union); SETUP row (P31 S62)
2026-08-26 19:01:41 -06:00
44f062a978
docs(phase-31): S62 checkpoint refresh at the T4/T5 boundary
2026-08-26 18:32:51 -06:00
194cce7570
docs(phase-31): T4 DONE — routing rule (<=50: Sonnet+DeepSeek parallel, Opus residue; 51-120: Sonnet, Opus escalation; >120: Opus; haiku retired; M-extend-tell -> wall); frontier-s61 Addendum 5; judge artifacts (P31 S62)
2026-08-26 18:32:51 -06:00
ad273e7b69
feat(decomp): T4 probe union — 17 fns banked (the best/cheapest arm per fn: haiku/sonnet/deepseek/opus drafts from identical packs), each binary byte-identical from a clean rebuild; func_8017FB38 func_80180F5C func_8017FF74 func_80184FB4 func_801810A4 func_8017E014 func_8017EF90 func_8017F590 func_8017E54C func_8017F3F8 func_80181894 func_800CAF9C func_801804C0 func_8017E7D0 func_8017F608 func_800CB578 func_801808E8 (P31 S62 T4)
2026-08-26 18:31:10 -06:00
02c5d757c2
docs(phase-31): T4 Claude-arm table (haiku 8 / sonnet 14 / opus 17 of 20 at the whole-binary gate), the warm-start homonym confound, distill yield (P31 S62)
2026-08-26 18:14:48 -06:00
05d13e52e7
docs(cookbook): §306 (hazard nop before a div-result store = statement order; maspsx --expand-div splices it) + §306a T4 distill addenda (case-body order at 3 nodes, narrow-cast copy+andi cell, dropped-edge sign rule, REGALLOC-PERM via a hoisted clamp local) + the warm-start homonym harness note — from 31 byte-proven T4 transcripts, 6 novel claims, 5 banked (P31 S62)
2026-08-26 17:07:12 -06:00
1392bf6c93
docs(phase-31): checkpoint — fleet sweep #7 213/213 GREEN at T3 close
2026-08-26 16:02:20 -06:00
e8549eaf10
docs(phase-31): S62 session checkpoint — T0–T3+T5pre done, paused at the R27 gate for T4
2026-08-26 16:00:43 -06:00
ceab8ad725
fix(config): ov_SC03_108 + ov_SC06_011 orders regenerated from their yamls — the failed table-bearing gate attempts left their carve entries in the mk and my 15:43 blanket add committed them (R59, mine); both byte-identical from clean rebuilds (P31 S62)
2026-08-26 16:00:01 -06:00
16d1d01b56
docs(phase-31): T3 log (75 banks; six carve-lane defect classes fixed; residue named) + frontier-s61 Addendum 4 (P31 S62)
2026-08-26 15:57:18 -06:00
5382025752
chore(docs): fleet ledger + backlog regenerated after T3 (2,397 stubs, 98.5% instr, 96.8% distinct)
2026-08-26 15:57:18 -06:00
4e562cfbf4
feat(decomp): ov_SC02_016/func_801810C8, ov_SC02_021/func_8017ECFC, ov_SC03_117/func_8017E6EC banked — the Block4-typed drafts once the isolator could see the lifted typedef; isolate + tail carve; byte-identical from clean rebuilds (P31 S62 T3)
2026-08-26 15:55:57 -06:00
99f601d02a
docs: cookbook §305 addendum — glued DEFINE carrier, attribute typedefs, leading-jr cut, the overlays.mk merge race closed at the root (P31 S62)
2026-08-26 15:49:08 -06:00
cd1e38ee0b
fix(tools): jr_isolate_all recognises lifted "} __attribute__((...)) X;" typedefs as engine types (5 drafts refused on Block4 after the lift) and drops a cut equal to the object start (a jr function that leads its object made an empty region -> out of order); mk_write refuses to merge a block that moved on disk since the caller read it (R43) (P31 S62 T3)
2026-08-26 15:48:50 -06:00
497d462c98
feat(decomp): ov_MAIN_012 + ov_SC02_037 func_8015444C banked (the family's remaining two; isolate + tail carve after the carrier/carve/mk fixes); both byte-identical from clean rebuilds (P31 S62 T3)
2026-08-26 15:43:29 -06:00
7324baebab
fix(tools): diff_autopsy.sh creates its scratch dir (the promoted copy lost the mkdir, so its snapshot cp failed and the restore left four TUs spliced), prints the first build errors and flags a stale cmp when the build fails (P31 S62)
2026-08-26 15:41:20 -06:00
34c94a17c4
feat(decomp): ov_SC03_107/func_8015444C banked — isolated region + tail carve (the func_8015444C family's first bank); byte-identical from a clean rebuild (P31 S62 T3)
2026-08-26 15:38:23 -06:00
fabde98078
fix(tools): three carve-lane defects behind the last CARVE-REFUSED class — jr_isolate_all's decl carrier no longer carries glued DEFINE_…() invocations (it re-instantiated shared bodies as nameless functions in the new region; prototypes carried instead), jtbl_carve treats a full data-side overlap after isolation as the ordinary tail state, and mk_write.write_overlays_mk merges only the caller's changed blocks into the CURRENT file under the lock (base=; the whole-file rewrite raced between parallel carves — ov_SC03_029/105/MAIN_012 drifts) (P31 S62 T3)
2026-08-26 15:38:23 -06:00
0e05821b02
feat(decomp): ov_SC03_029/func_8017FF7C, ov_SC03_105/func_80186DDC, ov_SC04_016/func_8017F03C banked after the type lift (gate-time isolate+carve); two orders regenerated from their yamls after a block restore landed on a successful carve; all three byte-identical from clean rebuilds (P31 S62 T3)
2026-08-26 15:34:11 -06:00
dcbe05cabe
refactor(types): lift Block4, Blk4_E960, Rec80182744, MATRIX_80188114, D80190448_t into src/shared/engine_types.h (lift_types --apply; 12 local defs stripped, the divergent Block4 in ov_SC02_011 kept local) — the naming types jr_isolate_all could not carry for 10 staged drafts; fleet clean sweep 213/213 (P31 S62 T3)
2026-08-26 15:25:58 -06:00
d7d8fdb50c
docs+tools: cookbook §305 (gate-time CARVE-REFUSED = 3 deterministic classes: mangled definition, uncarried naming type -> lift_types, post-isolation BOTH-copies -> tail); jtbl_carve treats a full data-side overlap as tail (P31 S62 T3)
2026-08-26 15:24:41 -06:00
0decb45580
feat(decomp): ov_SC03_028 (7 fns: func_80183A74, func_801847E8, func_80184914, func_80185864, func_80185C14, func_801887A8 + 1) and ov_SC06_030 (func_8017F65C + 1) — 9 resolver drafts banked once jr_isolate_all could partition their TUs; gate-time isolate + carve, both binaries byte-identical from clean rebuilds (P31 S62 T3)
2026-08-26 15:22:11 -06:00
cc8df52299
fix(decomp): ov_SC03_028 — a matched function was defined as FUNC_80180a08 (mangled case, resolves to no address) which made jr_isolate_all refuse every isolation in the TU (the 7 CARVE-REFUSED drafts of this overlay); renamed to func_80180A08, byte-identical (P31 S62 T3)
2026-08-26 15:09:45 -06:00
910101fe21
fix(gate): harvest_verify snapshots/restores ONLY the binary's own overlays.mk block, under the mk lock, against the current file — the whole-file restore raced between parallel gates and resurrected other binaries' rejected carve lines (ov_MAIN_012, S62); block round-trip unit-tested (P31 S62)
2026-08-26 15:07:03 -06:00
7bc10e31c8
fix(config): ov_MAIN_012 — order regenerated from its yaml; the 14:56 resolver gate (-j 10) resurrected a rejected carve's mk line via a whole-file overlays.mk snapshot/restore race between binaries, then blanket-committed it (R59/R60 caught live); byte-identical from a clean rebuild (P31 S62)
2026-08-26 15:06:21 -06:00
21972d9436
feat(tools): T5pre — main_queue_rebuild.py derives main's open set from progress.classify() (147 = 144 stubs + 3 NM; 959 LINKED + 2 blobs excluded) and attaches the newest draft per fn (136 queued, 11 never-drafted listed for the main wave; the old queue was 156/166 stale); campaign_status quotes the honest main denominator (P31 S62)
2026-08-26 15:02:15 -06:00
56542b18d5
feat(decomp): integration-resolver pass — 54 banked of 90 staged (zero model tokens)
...
Drafts the ledgers already recorded as byte-correct (closeness 0 / reloc shape MATCH), re-judged against today's tree by rtu_match + reloc_identity, staged, and gated on the whole-binary SHA. Ledger: .run/resolver/verdicts.jsonl
2026-08-26 15:02:09 -06:00
dfd42d4586
docs: cookbook §304 (self-defining rodata) + T3 log incl. fleet sweep #3 = 213/213 (P31 S62)
2026-08-26 14:56:06 -06:00
e4d1c55dde
feat(decomp): md_MAIN_011/func_800D04F4 banked — the resolver's PLUMBING refusal (undefined reference to D_800CEE3C): the strings lived only in the fn's own .s rodata block, so the C body now DEFINES them at file scope in address order (§304 self-defining rodata); byte-identical from a clean rebuild (P31 T3)
2026-08-26 14:55:49 -06:00
7a3ce3434f
feat(config): ov_SC03_118 + ov_SC03_119 — -O0 region o0d (0x80183178..0x80183830, 9 stubs each, the ds1-mapped island) carved by o0_subsplit; byte-identical after extract+build (class F's overlay members now draft at their real opt level) (P31 T3b)
2026-08-26 14:54:44 -06:00
df8f369eea
fix(tools): jtbl_rodata_pads --derive — empty spec when the TU has no C tables (never refuses what it need not pad), sub-4-byte all-zero gaps before an anchor accepted as assembler alignment, .float/.double sized; 70/70 module binaries byte-identical through the derive stage; T3a log (P31 S62)
2026-08-26 14:50:12 -06:00
c71cd7edb5
docs: cookbook §303 (module island pads derived at build time; island-pads/island-blocked dissolve) + SETUP row (P31 S62 T3a)
2026-08-26 14:48:42 -06:00
0b75da6a36
feat(build+tools): module jtbl pads DERIVED at build time — jtbl_rodata_pads --derive <binary> --tu <tu> (walks the retail island with the emission stream: .s spans from their comments, C data anchors from D_ names, C tables lead/trailing zeros; refuses on any anchor miss); Makefile runs it for every md_* object (no stored spec, nothing to drift); harvest_verify hands module island walls to it instead of isolate/split; the modules.mk probe line removed; md_SC03_076 byte-identical from clean (P31 S62 T3a)
2026-08-26 14:48:17 -06:00
ddd9029467
feat(decomp+tools): md_SC03_076 — func_801F0A9C + func_801F0F28 banked via the §8e pads filter wired for MODULES: jtbl_rodata_pads is now table-aware (acts only on an .align 3 followed by a $L label; const data / includes pass through) and supports trailing pads (0t1 = one zero word after the table, the §154-A island shape); spec 0t1,0 in config/modules.mk; byte-identical from a clean rebuild — the island-pads AND island-blocked refusal classes need no peel (P31 S62 T3a probe, R37)
2026-08-26 14:43:26 -06:00
4b31ceaa13
fix(config): ov_SC06_024 — JTBL_INTERLEAVE order regenerated from its yaml (drift predating today's sweeps; sequential-placement shift of 297k bytes); byte-identical from a clean rebuild (P31 S62 fleet sweep #2 : 212/213 -> this was the 1)
2026-08-26 14:38:59 -06:00
d718cecf9d
docs(phase-31): S62 log — clean fleet sweep found 6 hidden failures (5 healed, md_SC07_003 re-measured); R60 candidate (carve-state files); SETUP: interleave_check --fix
2026-08-26 14:35:06 -06:00
05f13d6c9f
fix(config): fleet clean sweep found 5 more drifted overlays that incremental gates had hidden — ov_SC03_002/ov_SC01_077/ov_SC04_019/ov_SC01_084 JTBL_INTERLEAVE orders regenerated from their yamls (interleave_check --fix), ov_SC05_010 jr_8017C8D0 pads 0,0 -> 0 (the 09:20 maintenance repair was blanket-reverted by the 10:21 ds1 wave commit); each byte-identical from a clean per-binary rebuild (P31 S62)
2026-08-26 14:34:32 -06:00
ab9e5770c7
feat(tools)+docs: T2 knowledge — interleave_check.py + pads_audit.py promoted (derive carve specs from the bytes; SETUP rows), cookbook §302 (the three carve-state invariants; five reds classified), frontier-s61 Addendum 3 (red list 5 -> 0, 39/40 held banks) (P31 S62)
2026-08-26 14:29:35 -06:00
b3e879a95e
feat(decomp): ov_SC04_018 — 18 resolver-held drafts banked after the T2d surgery; re-verified byte-identical from a clean rebuild; T2d/T2e log — red list EMPTY (P31 S62)
2026-08-26 14:28:24 -06:00
00e26c78d8
fix(decomp): ov_SC04_018 GREEN — jr_8017AE2C: (1) pads spec 4 -> 3 entries (the carve's 4th table belongs to func_80181804, reverted to a stub; pads_audit-derived); (2) the TU never compiled as committed: an undeclared D_801B9AA8 whose limbo entry poisoned three later block-scope externs, an address-of-a-cast (&((char*)&D)), a (void) re-declaration of func_8017E7CC after its (int) definition (call cast per §20), and func_8017E7CC's first store with offset/value swapped (retail li 3 / sh 2(s0)); R22 clean rebuild byte-identical; RED LIST EMPTY (1 -> 0) (P31 T2d)
2026-08-26 14:27:15 -06:00
315699b597
feat(decomp): ov_SC02_005 — 1 resolver-held draft banked after the T2e surgery; the gate's §8a carve split TU jr_8018EA04 (new .c + yaml subseg + order, consistently this time); re-verified byte-identical from a clean rebuild (P31 S62)
2026-08-26 14:25:14 -06:00
Drew T