TASK_PROGRESS — T3 attempt 1

Task: T3 head-type layouts and canonical files Attempt: 1 Agent: expert-fable Ctx at handoff: n/a Commit: 1a35655ab (last coder), this handoff commit follows

Done so far

Census refreshed with --sites into .run/P37/census (restruct's hard-coded input dir) at HEAD 42a9cd7f5: run census_t3 → .run/logs/census_t3.log; heads separate again (Unkstruct_800B5CB8 107,580 sites/11,184 bodies, Unkstruct_80126B58 65,397/17,047; 18,739 types); the P37 T4e census had them merged as Unkstruct_800D3888 (stale body_base_type.json, now regenerated).
c1 5aaa25154: tools/head_types.py (max-site split, weighted-interval DP over layout ∪ conflicts; --header, --decisions, --selftest), include/struct_types.h (generated block, both heads), docs/struct-map-decisions.md (skeleton + generated per-offset block), include/common.h:64 includes struct_types.h, CANON_HEADERS = (engine_types.h, include/struct_types.h); audit-types 26 (0 in new file), selftests 24/54, R22 green.
c2 e4341ee11: tools/delever.py includers() follows -Iinclude (relative first, else include/x; scans include/*.h; cache key "I|"); hand-written Unkstruct_80078E78 (0x98, evidence-named fields) in struct_types.h; src/800.c:8 typedef Unkstruct_80078E78 Blk98_80029274; (R22 byte-identical, not reverted); decisions doc's three hand sections filled (canonical file set, resolution rule, player block). R22 green.
c3 1a35655ab: restruct pass_hint (:522-531 skips non-str rows; classify_residual emits aligned mnemonic rows as pairs, regs as regs), drop_overlaps (:698-710) → nested-base overlapping edits become KEPT(overlap-edit) instead of Refuse:; +2 selftests (56); snapshot-baseline at e4341ee11 (7,428 objects, 218/218) + calibrate 146/146 identical, positive control DIFFERS; --selftest --real 61/61 OK, linked-control OK.
Exemplar proofs (all final IDENTICAL, 0 TYPE-NOT-VISIBLE, logs .run/logs/t3c3_try_1..9.log): 800B5CB8 → ov_SC06_032_jr_80182890 (MEMBERS 123, S2 7), ov_SC02_027_jr_8017D898 (193/4), md_SC07_004 (220/17), md_SC07_003 (82/2), src/800.c main (108/10); 80126B58 → ov_MAIN_012_jr_8015A3C8 (104/5), ov_SC03_107_jr_8015A3C8 (102/4), md_MAIN_046 (24/3), md_MAIN_036 (15/2). 80126B58 has 0 main bodies (map), so no main exemplar exists for it.

In flight

verify command running detached: bash tools/run.sh --wait t3_verify --max 280 (repeat) → .run/logs/t3_verify.log; expected restruct --selftest --real: 61/61 OK, extract-all: 217 extracted, 0 failed of 217, check-all: 218 passed, 0 failed of 218. Started at HEAD 1a35655ab, clean src/.
Uncommitted tracked-but-gitignored outputs: .run/P37/census/{struct_map_top.json,type_census.json,type_census.txt}, .run/P36/delever/calibration.json, docs/struct-map.md (regenerated) — commit with git add -f first (T1 gotcha), then commit_task.
Not yet written: phase-ends/current/logs/T3.md, phase-ends/current/tasks/T3.md.

Hypotheses rejected

"Unkstruct_80078E00 exists in the preview / the map" — preview ov.h has only 800B5CB8, 80126B58, 8018E71C, dac960ce, 0caecd54, f8ed2b9f; map has no cluster at 0x80078E00; D_80078E00 0 occurrences in src (.run/P38/T3/player_vs_heads.py: of 9 scalar player globals, 1 agrees with 80126B58, 0 with 800B5CB8; both heads have s32 at +0xB4 where hp_current is u16).
"actor-struct.md describes a head layout" — no; the code's player record is Blk98_80029274 (0x98 at 0x80078E78, src/800.c) / Blk80078E78 (resident_jr_800D00E4.c:546) / u8 D_80078E78[] in 880 TUs.
"a type in include/ is visible to restruct" — was false: delever.py:663-674 resolved includes relative-only and never scanned include/ (fixed in c2).
"--try-file crashes are caused by the new include graph" — c2 reproduced both crashes with HEAD~1's graph; pre-existing (fixed in c3).

Current hypothesis

All done-when clauses hold except the literal name Unkstruct_80078E00 (defined as Unkstruct_80078E78 by evidence; a deviation, not a redefinition) and "one main TU" for Unkstruct_80126B58 (0 main bodies in the map). The verify run should be green; then only the log, summary, commit and return remain.

Next 5 steps

bash tools/run.sh --wait t3_verify --max 280 until done; read the three result lines; if red, report as failure with the output (do not weaken).
git add -f .run/P37/census/struct_map_top.json .run/P37/census/type_census.json .run/P37/census/type_census.txt .run/P36/delever/calibration.json then include docs/struct-map.md in the commit paths.
Write phase-ends/current/logs/T3.md (timeline above; retriever questions: 3 retriever-code, no report ids; scripts in .run/P38/T3/), and phase-ends/current/tasks/T3.md from templates/task.template.md (≤150 lines; Verified line from step 1; Deviations: Unkstruct_80078E78 not _80078E00; no main exemplar for 80126B58; include/ canonical file instead of per-space files; two restruct fixes + delever include fix outside the plan's file list; gotchas: harness: comment text /* 0xNN */ inside a C comment kills the build; generalizable: include graph must model -Iinclude; binding: head-type overlaps resolve by max-site split, unions deferred; Next task needs: Unkstruct_800B5CB8 is a suspect merged cluster (25 % dropped sites) → T5 split candidate; REFUSED bodies = per-candidate compile errors (e.g. md_MAIN_036:300 ar46E98 undeclared)).
PY tools/task_log.py finish T3; git mv phase-ends/current/TASK_PROGRESS.md phase-ends/current/logs/T3.progress1.md (respawn does this first); bash tools/commit_task.sh T3 "head types canonical; include graph, try-file fixes; log and summary" docs/struct-map.md.
Return the expert contract (STATUS done, COMMIT hash, MILESTONE n/a).

Gotchas

The expert's Bash hook denies reading tool sources even by range; code facts come from retriever-code.
restruct reads .run/P37/census/* (hard-coded), not .run/P38/census; cluster names change when membership changes (lowest AT global).
restruct --selftest --real fails on stale calibration after any commit that moved HEAD ("calibrated at X, HEAD is Y"): snapshot-baseline + calibrate on a clean green tree first.
--try-file on the 9 exemplars takes 1.5–125 s each; run through run.sh.
commit_task.sh refuses gitignored-but-tracked paths; git add -f them first.

State to carry verbatim

Commits: c1 5aaa25154, c2 e4341ee11, c3 1a35655ab. Census head 42a9cd7f5.
head_types summary lines: Unkstruct_800B5CB8: fields=101 kept_sites=75717 dropped_sites=25393 conflict_offsets=50 overlaps=167 misaligned=4 union_candidates=27; Unkstruct_80126B58: fields=152 kept_sites=61599 dropped_sites=1713 conflict_offsets=50 overlaps=71 misaligned=5 union_candidates=4.
Per-space bodies: 80126B58 ov 16,773 · md 311 · main 0; 800B5CB8 ov 15,245 · md 502 · main 22 · boot.c 1.
Verify log: .run/logs/t3_verify.log. Coder logs: phase-ends/current/logs/T3.c1.md, T3.c2.md, T3.c3.md. Scratch scripts: .run/P38/T3/{map_dump.py,bodies_by_space.py,player_vs_heads.py,player_use_widths.py}; offsets dumps .run/P38/T3/D_80126B58.offsets.txt, D_800B5CB8.offsets.txt.
Known-true control line from --selftest --real: "known-true p[1]/p[2]: whole-object DIFFERS · reloc-only True · linked IDENTICAL"; linked-control OK.

Coder runs so far

c1 coder-opus55 — done 5aaa25154 — logs/T3.c1.md
c2 coder-opus55 — blocked (tool bugs), committed e4341ee11 — logs/T3.c2.md
c3 coder-opus55 — done 1a35655ab — logs/T3.c3.md

Reports commissioned

none (three retriever-code answers under 300 lines, no report ids)

7.5 KiB Raw Blame History Unescape Escape