293a82c5e8
9 Commits
| Author | SHA1 | Message | Date | |
|---|---|---|---|---|
m3ultra
|
293a82c5e8 |
R42 THE LOOK: five lane-runs — the pub is dressed, the decks read, and the town is cast
John looked at R41's proof shots and named three faults. All three are fixed, at zero draw cost, and every lane corrected somebody's numbers on the way — including its own. E-LOOK (§42.1+§42.2a) — THE PREMISE WAS WRONG. "The pub props have no texture" (my brief, quoting E's own R41 note) is false: all 37 carry COLOR_0 and three.js renders it. The colour was BLEACHED — jukebox at saturation 0.02, 15 assets under 0.05; what the source library called colour was baked AO. So the fix was a palette imposed on colour already present, in place: 46 assets, draw delta +0 and tri delta +0 on every one, proved the honest way — the concatenated POSITION/NORMAL/TEXCOORD/index bytes are sha1-identical to R41, so footprints and origins provably did not move. (A first-pass bbox check ignored node rotations and "found" 3 phantom violations; thrown out for the byte comparison.) The banquette was 1990s office teal and went burgundy; the beer umbrella was white and went brewery green. THE DECK, diagnosed before being touched: 7 greys, chassis 0.05 luminance on a 0.033 bench = 1.21:1, and the one high-contrast feature was INSIDE-OUT (white ring around a dark disc; a turntable is a dark record on a light platter). Now Technics silver at 6.74:1, black vinyl, red start/stop. Node contract re-read from the shipped GLB: 4/4 + 1/1 PASS. Filed to R43, not fixed: deck_1200_rigged has NO PLINTH (bottom 43mm of 119mm lost to R41 join damage). E-CAST (§42.3) — 19 of 24 ship. Rejected: comical_luchador_01 (Mixamo Ch43, purple mask reading EL CHUPACABRA, identified from the file's own material name), a 3-heads-tall cartoon child, a robed wuxia elder, a WW2 infantryman — and dj_phrtt_01, AN UNCLOTHED BODY whose only texture is a bare-skin atlas, which was in the DEFAULT crowd and live on the depot. It is the R41 deny-list's failure class through a different door: an innocent name from another pipeline. Only rendering all 24 and LOOKING found it. Ruling 2 (prefer CC0) was REFUSED ON MEASUREMENT: Quaternius is 25 nodes / 23 joints and ZERO of 64 canonical tracks bind — it would have loaded silently and never animated. Replacements came from Mixamo's Ch## pool, filtered to the contract-clean 65-bone family (candidates ran 65..99 bones; named characters carry facial/hair bones). THE LICENCE GAP WAS UNREAD, NOT MISSING: every source FBX carries ApplicationVendor "Mixamo, Inc." internally. Adobe royalty-free, no attribution owed. B's amber rows are now real. B (§42.5) — MY ASSIGNED FIX WAS USELESS AND B PROVED IT. document.fonts.ready shipped, then measured: the atlas canvases are already byte-identical across boots, the game ships NO webfonts (document.fonts.size === 0), and ready fires 50ms BEFORE the first fillText. The real cause of unreproducible bookmarks is the wall-clock CROWD — 983 changed px of 921,600, ALL in the pedestrian band; 0 px with the crowd out. New DBG.freeze() takes stable bookmarks 2/5 -> 5/5. Also found: ?pop=0 boots 140 citizens (0 is falsy against the default). THE BENCH: D's filing correct on both counts. Front-vs-road 90.0deg -> 0.0deg, min=med=max over 60/60 instances; the streetlight arm now reaches 1.05m over the road where it reached 0.00. Both used a base yaw that pointed +Z down the street while their own comments claimed otherwise — the side term was already right, which is why it survived 41 rounds of reading. Also: on any box with CDN access 9 of 60 benches are the depot GLB, 90deg out from the house law — normalised at the geometry, once, at load. benchStops(plan) exported as one truth. C (§42.2b) — 122 vs 123 WAS THE CAMERA ASPECT, and nobody was wrong. Draw counts are post-frustum-cull and the sweep camera inherits the window aspect: 1.25->116, 1.60->122 (C's harness), 1.778->123 (F's), 2.37->124. Pinned MEASURE_ASPECT=16/9 with a 1.25 control. E's "keep the bench dark" advice was measured and declined with reason (deck-vs-bench was already 4.73:1, over threshold). "Reads as a black box" is a claim about pixels, so C measured pixels: near-black in the booth crop 57.5% -> 26.7%, distinct colours 228 -> 389. Bench rebuilt 3->7 boxes with an open record bay; headphones on the mixer; faceDoor aims booths at the player's entry — alignment within 45deg went 49% -> 92%. Two bugs found by looking: a fitting could be planted INSIDE A WALL (0.77m of a 0.86m half-depth in plaster), and every booth part sat 25mm sunk into its own worktop. D (§42.4) — five swaps IN PLACE at the same index (pickRig indexes these arrays and pedIndex is in the identity signature and the impostor atlas; removal would renumber the town). Identity signature pre- vs post-swap: 0 differing lines of 150 — who each citizen is did not move, only which body they wear. dj_phrtt_01 verified gone AT THE WIRE, not from source: a fresh default boot fetches 32 ped GLBs, none retired. THE CONTROL ARM FOUND A 41-ROUND-OLD BUG: the ped being RETIRED, comical_boy_01, binds only 3,132 of 3,460 tracks — six thumb bones lack the mixamorig prefix. It had been silently dropping six tracks per clip since it shipped. Testing the thing being thrown away found it. Imported B's benchStops mid-round: sitters now 0.0deg to the road, signatures byte-identical across the rewire. Corrected F's filing: the browse smoke was flaky-by-construction (a 16-25s race against a live sim), not deterministically red. F (§42.6) — qa.sh --strict 24 passed / 0 failed / 0 warn / 0 skipped. New Gate 11: period law (0/5 banned; CONTROL: R41's roster trips all five) and roster licence (24/24, 0 unverified; CONTROL: an amber row turns 20 red). r41_shots.py promoted warn -> strict. R41'S BYTE-HASH BAN IS REVERSED AND ITS STATED REASON WITHDRAWN AS WRONG. A gate may pin a hash under five in-run assertions (classic+flags boot, freeze().stillLive all-false, a second freeze removing 0 draws, settle until DBG.info() stops changing, hash with the dbg panel hidden): 5/5 bookmarks now pin to one sha256 over 3 fresh boots. Two findings got it there — the chunk streamer (a fixed settle is a bet on load) and the fps readout, which with everything else frozen was the last thing moving at 53-62 px. ?pop=0 fixed (Number.isFinite) — a no-op for 35 rounds. Three red flags_check gates fixed. Street: noon 282 · NIGHT 291/300 — the nine-draw margin INTACT, every R41 row reproducing to the digit. Interior at pinned 16:9: 123 @ dept/auto, margin 227. selfcheck 157,647/157,647, 0x5f76e76 unmoved. No tag — John's playtest rules the epoch. Co-Authored-By: Claude Fable 5 <noreply@anthropic.com> |
||
|
m3ultra
|
78f49f7113 |
Lane D R41 §41.3: the town stops walking — 99.3% to 78.1%, at zero draws
THE ROUND IN ONE MEASUREMENT (12 samples x 146 active, ?clips=0 vs default — a new flag that turns off the library and nothing else): walking 99.3% -> 78.1% · bench-sit 0 -> 9.5% · lean 0 -> 7.1% · stopped in own idle 0.4% -> 5.1% · DISTINCT CLIPS ACROSS THE CROWD 4 -> 20. The town was 99.3% people walking because standing still had nowhere to happen. Wiring: new postures.js + clipbank.js. idles.glb (10/10) drives a per-citizen deterministic idle on every near-tier actor plus the seeded shopkeeper. locomotion gives 33.7% of walkers a shopping bag. sitlean (8/8, lazy) puts 4 sits on Lane B's ACTUAL benches and 4 leans on shopfront walls. browse (5/8, lazy) is a real BROWSE state at C's browse points, seeded per (shopId, slot). venue (5/6, lazy) widens the gig crowd, plus a publican pouring and a record keeper in headphones. social (0/8) is never fetched — two-person conversation needs a paired state machine, filed to R42. Cost: boot = 4 requests, 1.24 MB / 16 clips resident; the rest lazy on first need; heap delta +3.34 MB; mixer median 0.1 ms both arms. ?clips=0 / ?classic=1 / ?noassets=1 fetch ZERO clips — not even clipbank.js (dynamic import). No shell edit needed. DRAWS: +0 on every bookmark (street_noon 193, crossroads 108, night_crowd 128, market_square 94, night_neon 111, interior 110 — identical both arms). Ruling 4 respected exactly. DETERMINISM: 150 citizens, two fresh contexts, byte-equal posture signature. Controls: seed+1 differs; EVERY clip GLB delayed 2 s -> identical signature (posture is a pure function of (citySeed, id), never of residency). 6 new streams collide with none of the 12 pre-R41 keys. TWO FINDINGS THAT CHANGED THE DESIGN: the idle pool was INVISIBLE — wired only to the R17/R29 node loiter, so only 0.8% of citizens were ever stopped; and the lean never fired at all (0 in a 9 s run). Both moved to the patronage stride check. Bench stations are GATED not trusted: 14/14 derived stations coincide with real instanced geometry within 2 cm, and the control (same stations offset 2 m) matches 0/14. Filed to B: one benchStops(plan) export retires the mirror, and furniture.js puts the bench's front ALONG the street rather than facing the road, contradicting its own comment. Leak: +0 geometries, +1 texture over 6 enter/exit cycles. Goldens 157,647/157,647, 0x5f76e76. Co-Authored-By: Claude Fable 5 <noreply@anthropic.com> |
||
|
m3ultra
|
3e1ff312a5 |
Lane D R40 §40.5: the peds come in the front door — 100% on every real town
Verification first, and B's R39 heading fix did NOT clear the gate alone. Measured today (tools/qa/door_footpath_check.mjs, new): synthetic 447/493 foot (90.7%) · katoomba 71/72 · fitzroy 138/139 · bowral 28/30 (93.3%) — from R39's 0/493, 0/72, 0/139, 0/30. Two residue species, handled differently ON PURPOSE: 1. A 0.85m kerb miss on real towns — plan_osm's setback puts the facade past the corridor edge, so the shell's d/2+0.6 door point lands just off the verge. Fixed by new D-owned door_snap.js, applied in sim.setShops: clamps each door into the registry vergeBand of its nearest footpath-bearing edge, 6m cap. Corpus: fixes 4, breaks 0. Live-shell verified (493 points, 2 moved, max 5.41m — matches node exactly). Real towns now 72/72, 139/139, 30/30. 2. The synthetic's 36 strict misses are market stalls and arcade shops WHOSE FRONT IS THE MARKET SQUARE. An uncapped snap would 'pass' them by teleporting a door 59.85m to a street they do not front — passing the gate by breaking the truth it measures. The gate instead scores walkable front ground (footpath ∪ market/arcade block ground) = 473/493 (95.9%) and reports the strict number alongside. The check ships its own falsifiability CONTROL (the R8-R39 back point must score under the floor: reads 0.4-13.3%) and a fix-only arm (clamp breaks 0, asserted forever). Deterministic. Remaining 20 synthetic misses are artefacts of the fixed nudge at index.html:433-437 — not D's file; measured, exact fix documented for F. Goldens untouched. Co-Authored-By: Claude Fable 5 <noreply@anthropic.com> |
||
|
m3ultra
|
888335d633 |
Lane D round 12: gig band + crowd + patronage surge (?gigs=1)
v3.0-alpha — Friday night at the pub. New web/js/citizens/band.js (GigCrew): - Band trio on C's stage.bandPoses (deckY), distinct fleet peds (seeded per gig), each with a primitive instrument on the chest (guitar/mic/bass; E's GLBs drop in via opts.instrumentFor). - Crowd (<=8) at C's watchPoints on the gate-protected spawnRig path: dance:false = weight-shift idle, dance:true = bounce/sway/step with seeded phase (no metronome). - Deterministic from citySeed; placeholder crew under ?noassets. sim.js: setGig(venueShopId, on) gig-night surge — patronage pulls peds from further (GIG_RANGE) and ducks in harder (GIG_SURGE) at the venue; occupants drain at close per A's closing-time ruling. Guarded by _gigVenue==null so gig-off is byte-identical (prime flag law; goldens unmoved). Verified standalone (A withGigs + C buildInterior + D): band 3, crowd 8, dancing 3, all 11 figures human-sized (1.60-1.83m under a 4.0m ceiling; R10 no-giants holds), band faces audience + crowd faces stage, deterministic, leak-free (8 cycles, 0 delta). Crew ~40 draws. Surge: venue occupancy 1 -> peak 5, identity stable. Shot: docs/shots/laneD/r12_gig_night.jpg. Flag for C/F: band.js flips C's watchPoints/bandPoses ry by pi (RY_FLIP) — the fleet mesh's visual front is local +Z, not the -Z the comments assume. Remove RY_FLIP if C re-fronts the gig poses. Co-Authored-By: Claude Opus 4.8 <noreply@anthropic.com> |
||
|
m3ultra
|
3432fccd1f |
Lane D round 10: fix interior giant rigs + no-giants QA gate
R9 blocker: interior rig figures were ~2x too tall (keeper crown 3.83m in a 3.4m room, clipping the ceiling). Root cause in rigs.js buildFigure: it normalised scale by height/headY, where headY is the head bone's world Y in BIND POSE — but the Mixamo skeleton origin is the hips (~mid-body), so headY is head-above-hips (~half the standing height), not feet->crown. Every rig came out ~1.83-2.0x too tall. (The head does NOT move when posed — bind crown 3.209m vs idle 3.205m — so it's a hips-relative measure, not an animation issue.) Fix: normalise by the feet->crown span (headY - minY, both already computed). buildFigure(rig, h) now lands the crown at exactly h above planted feet. Fixes keepers, browser rigs, near-tier street rigs AND the impostor bake (all ride buildFigure). Verified: all 19 fleet rigs crown==1.75m; live keeper 3.83m -> 1.82m; determinism + dispose-leak soak unchanged. Regression guard (required): new tools/qa/interior_scale_check.py (Playwright) enters a sample of shops, injects browser rigs at browse points, asserts every interior fig crown in [1.4,2.0]m AND < room dims.H AND feet planted. Wired into tools/qa.sh --strict (run_gate => a giant fails qa). Proven bidirectional: PASS on the fix (24 figs/6 shops), FAIL on a reintroduced giant. Full qa.sh --strict: 6 passed, 0 failed. Re-shot laneD/r10_browsers_fixed.jpg. Figures human-sized relative to doors/fittings: yes. Co-Authored-By: Claude Opus 4.8 <noreply@anthropic.com> |
||
|
m3ultra
|
34145e24a7 |
Lane D round-9: interior presence — occupancy truth + browser rigs (C->D->F seam)
Follow a ped into a shop and find them browsing. D owns occupancy truth; C owns browse points;
F wires the handoff. Validated end-to-end in-shell (?stock=real). qa --strict GREEN; v1 untouched.
Occupancy (D1): patronage records who's inside which shop by shopId. New
citizens.occupancyOf(shopId) -> {count, occupants:[{seed,enteredAt,pedIndex}]} (pedIndex = the
exact ped who ducked in, for identity continuity). Cleaned on emerge/chunk-drop/stream-toggle.
Deterministic: 17 shops + occupants byte-equal across two runs (fixed a stale-occupancy leak on
stream toggle).
Browsers: keepers.js gained browse:true (faces the shelf via ry, no player-greet) + pedIndex +
seedKey. Each = a merged ped ~1 draw. 12 enter/exit cycles: count == min(occupancy,3) 12/12,
worst 61 draws <=350 (C's headroom holds), leak-free (0 GPU delta after warmup), disposed every
exit via keepers.disposeAll().
Consistency (D2): inside peds hidden on the street (0 rendered), re-emerge at the door.
F handoff verified by running F's exact interior_mode wiring (enter -> occupancyOf ->
browsePoints.slice(0,min(count,3)) -> keepers.spawn). index.html setShops door points need
shopId:s.id added. Closing-time handled by the dwell model (pending A's explicit ruling).
Evidence: docs/shots/laneD/r9_browsers_in_shop.jpg. Changed sim.js + keepers.js + test page.
Co-Authored-By: Claude Opus 4.8 <noreply@anthropic.com>
|
||
|
m3ultra
|
93a3168a99 |
Lane D round-5: chunk-streamed roster (v2) behind ?roster=stream
Implements the R3 design note (was design-only). Default-off; the v1 whole-graph path is byte-identical when the flag is off (shell citizen-0 signature matches the R4 golden exactly). - chunk-local identity keyed (seed, chunkKey, i) — same seed + chunk -> same residents, independent of town size / visit order; home edge picked from that chunk's edges. - windowed spawn/despawn: hook-driven via Lane B's onChunkBuilt/onChunkDisposed (sim auto- detects and stops polling) + a camera-poll fallback that needs zero Lane B changes. - global near-cap + mixer stagger + rig pool + impostor layer unchanged, run across live chunks. - owner-chunk despawn (no cross-boundary re-keying); per-chunk time-of-day thinning (hours-aware). Proven on a 9x9 fixture grid: constant far-field density (~58-129 within 70m at spawn AND 150m out, vs v1's uniform ~12-15); deterministic (507 chunk-keyed identities re-derive from seed); leak-free (2 full grid walks / ~380 chunk cycles -> 0 geom/tex delta, pool capped 30); budget unchanged (near-cap 24 + 1 impostor draw, 2.31ms/frame). qa.sh --strict GREEN. Only web/js/citizens/sim.js + web/citizens_test.html changed. Flag + shell call-site documented in LANE_D_NOTES for Lane F to wire web/index.html. Co-Authored-By: Claude Opus 4.8 <noreply@anthropic.com> |
||
|
m3ultra
|
a5e4b64a9d |
Lane D round-4: in-shell verify (D1) + decimated-ped validation (D2) + impostor tone-map fix (D3)
D1: peds+keepers upgrade placeholder->rig in the shell; 12 enter/exit shops leak-free (renderer.info.memory back to exact baseline); determinism holds across reload; ?noassets fully placeholder with zero ped fetches. D2 (gate-3 critical path): validated Lane E's decimated fleet — all 19 peds <=3k tris (1564-2805, avg 2450), rigs bind+animate (skinning/skeleton preserved), 0 broken meshes under eviction churn, silhouettes+identity intact, atlas bakes clean. Near fleet ~59.6k tris at the 24-cap (was ~1.2M) — F clear to re-measure gate 3. D3: fixed a real bug found only in-shell — the impostor self-tone-mapped (toneMapped:false + in-shader ACES), correct on the direct-canvas test page but DOUBLE-tone-mapped under the shell's EffectComposer/OutputPass. Reworked to a standard toneMapped:true material outputting linear via three's own tonemapping/colorspace chunks, so it matches the near rigs on both render paths. setExposure is now a no-op (exposure is global). Verified noon+night in-shell. Only web/js/citizens/impostor.js changed (+ NOTES/progress/R4 shots). Co-Authored-By: Claude Opus 4.8 <noreply@anthropic.com> |
||
|
m3ultra
|
9eb1acc61b |
Lane D (Citizens): deterministic LOD crowd — rigs/impostors/keepers + test page
Populate the town at city scale, on budget, deterministic per-citizen identity: - rigs.js — ported 90sDJsim rig stack: mixamorig canonicalisation so one shared walk clip drives all 19 peds, head-bone height-normalise, feet-plant, pooled near-tier actors with walk<->idle crossfade - impostor.js — 4-yaw sprite-atlas baker + instanced billboards: whole mid crowd in 1 draw call, tone-matched to the ACES near rigs - sim.js — deterministic roster, footpath lanes off the street graph, near/mid/far LOD with hysteresis + hard 24-cap, staggered mixer budget, time-of-day density - placeholder.js — seeded low-poly box humanoids (frame-one population + ?noassets) - keepers.js — one keeper per shop at the counter, idle + greet-turn - 21 ped GLBs (byte-identical from 90sDJsim), citizens_test.html harness, LANE_D_NOTES.md, 5 beauty shots Verified in-browser: 200 citizens at max mixer 0.2ms (budget 2ms), near capped 24, determinism holds (200 identities match seed), no T-pose, ?noassets crash-free. Consumes Lane A plan.streets with zero adapter. Six adversarial-review bugs fixed (non-deterministic fleet order, shared-resource disposal on pool eviction, impostor colour-management). Co-Authored-By: Claude Opus 4.8 <noreply@anthropic.com> |