Chapter 30: Longer boats and a line to the quay Frozen 12 October 2026 at 22:42 BST (game main at 36011bd). Covers the first four parts of docs/decor-controls-ui-plan.md: R5a fixes and document sync (9 October), R1a boat length grip and parametric hull (10 October), R1b boat styles and R1c moorings and the quay-edge magnet with its fix pass (12 October). The R5a and R1a entries are committed; the R1b and R1c entries and the 12 October rulings in artifacts/decor/owner-decisions.md were read from the uncommitted working tree. Later work (the next part, a jetty, and anything after it), the queued crisp-grips request and the castle plan beyond one sentence are not covered. Status was checked against docs/decor-orchestration-handoff.md section 1 (not quoted here). Original source terminology is preserved. Line numbers refer to the frozen copies in research/update-2026-10-12-P10close/. SOURCE: docs/visual-handoff.md (frozen 12 October 2026, working tree), lines 5-20 (Decor R1c) 5: **Decor R1c Moorings and the quay-edge magnet (12 October 2026):** Opus 5.5 on this Mac, 19:34-20:36 (docs/decor-controls-ui-plan.md part "R1c", stage R1 Water craft, step `29-R1`, part c). A boat released near a quay lies parallel to it 0.35 m off (Ctrl places as drawn); a moored boat gets derived mooring lines to a bollard, mooring rope, lantern post or houseboat link post, a boat alone on open water a buoy, and fenders hang on the quay side only; nothing derived is saved or enters undo. Snapshot: `build/backup-feel/29-R1-c-pre/` (orchestrator's, 19:34, with R1b uncommitted; the live tree equalled it at part start, no foreign hunks). Scoped backups `c/backup/`; scratch tree `c/pre-tree`; part diff `c/part.diff` (959 changed lines; below, `c/` = `artifacts/decor/29-R1/c/`). 6: - **Implemented.** `scripts/scene_state.gd`: `EDGE_REACH` 1.5, `water_edge_at(p, reach)` -> {point, normal (land to water), tangent, height, distance, cell} for the nearest land-cell side or corner within reach, {} on land or with none; it reads `land_at` only (any finish but stepping stones), so building, bridge and every other placement is untouched (no stop). `scripts/magnets.gd`: `QUAY_REACH` 1.5, `QUAY_GAP` 0.35, `QUAY_PARALLEL` 0.985, `quay_edge`, `quay_berth` (edge within reach of the hull side; turns to the edge tangent nearest the heading; {} unless the berthed hull is clear) with a "quay" guide along the edge. `scripts/main.gd`: `object_candidate` applies the berth to a boat afloat, carried by nothing, when snapping is on (place, G and body drag share it); the turn grip and the inspector Length row stop at the quay (below); the drag replay gains a `body` step (`replay_body`). `scripts/drag_harness.gd`: `body` in `STEP_TYPES` with its validation (the same harness change in `c/pre-tree`, `c/drag/pre-tree-harness.diff`). `scripts/composition.gd`: lived stage 3 is the moorings (`moorings`, `_moorings`, `mooring_anchors`, `boat_afloat`, `boat_quay_side`, `_over_hull`); the lived cache keys on the terrain when a boat is present. `scripts/architecture.gd`: `mooring_line` (two-box cleat, `_link_cable` rope 0.025, a rope turn at the anchor), `mooring_buoy` (red and cream floater, iron ring); `boat(..., fender_side)` drops the fenders on the far side for cabin, narrowboat, barge and houseboat (rowing and sail have none; side 0 draws as before). `scripts/world_view.gd`: `lived_moorings`, drawn with the links, held and hidden by owner during a drag, recomputed on release; decoration key `_Fq` for a one-sided boat. `scripts/creatures.gd`: nap spots (carried item, below). 7: - **Design readings (for the review):** (1) anchor reach 1.2 m is measured from the hull rectangle in plan, and each end takes the anchor nearest its own cleat, so on a 14 m Amsterdam houseboat the stern line can run diagonally along the hull to an anchor near midships (boat 55 stern to lantern 34, 7.9 m; boat 56 stern to bollard 39, 8.5 m): a spring line, never over a hull (**superseded by the fix pass**: no spring lines, both lines dropped); (2) the fender side needs a quay within reach on exactly one hull side and the hull parallel to it (0.985), so a boat in a 2-3 m channel with both banks in reach keeps both fenders (Courtyard boat 47, `moorings.json` cabin 143); (3) the magnet turns a boat parallel whatever its angle (boat 136 in the drag turns 90 degrees to lie along the pier). 8: - **Carried items.** Turn grip: `start_resize` records whether the boat's hull was clear; each turn step bisects (Adapt.clamp_fraction) from the last valid turn toward the asked one and stops at the last clear angle with "Turned as far as the water here allows." Length row: a growing length that would leave the hull unclear is clamped to the longest clear length (0.001 m) with "%.1f m — as long as the water here allows."; both one undo step, tested. Nap spots (creatures.gd ~:615): a houseboat's box (grown 0.25 m) and its base corners now go into `ctx.hulls`, and `_nap_keep_some` drops any spot inside another hull or within NAP_CLEAR of its points; the houseboat's own roof spots skip their own hull, so the roof stays a nap surface (fixture: a quay-wall ledge beside a moored 6 m boat; the ledge spot goes for cabin and houseboat, the houseboat roof spot stays). composition.gd ~:2067: the comment now says a flag at the other end gives bunting. Houseboat end already strung: a mooring line ties at the link post's foot on the roof (`mooring_anchors`: the post minus HOUSEBOAT_ANCHOR - HOUSEBOAT_ROOF - 0.04) while string lights or bunting hang from its top, so one post serves both; tested (a rafted boat moors to the post foot and the string stays). Amsterdam night: `--lighting=night --time=22` in the capture args (Main reads `--time=` after `--lighting=`; the scene's own hour 18.5 otherwise wins); `c/cap/after/ams-night`, `ams-close-night` looked at: night sky, lit windows and lamps. 9: - **Tests** (test-first logs `c/testfirst/`, after `c/suites/`, runner `c/run_suites.sh`). Test-first on the unchanged code: test_decor 21 of 285 failing, test_links 1 of 411 (the R1c block is gated on moorings existing), test_creatures 2 of 250, test_chemistry 1 of 414. Test fixes after test-first: the creatures fixture moved to a quay-wall ledge (a quay spot beside a hull was already removed by the blocked shore cells, so the first fixture passed on old code); the body-drag check runs in a fresh main (the earlier `test_click_at` checks reframe the test viewport so the press and target were 4 px apart, under the drag slop). After: test_decor **308/0**, test_links **449/0**, test_creatures **250/0**, test_chemistry **425/0**, test_adapt **640/0**, test_scene **1709/0**, test_starters **393/0**, test_cloth **116/0**, test_creature_life **399/0**. New checks: edge finder (ten Harbour probes on the east quay, pier and south quay, open water, reach, land, corner; channel banks, finishes 1-5, stepping stones as water, writes nothing), magnet (30 degrees 1 m off -> parallel, centre 0.35 + 0.75 off, one undo step; Ctrl keeps 30; out of reach at 6 m; guide; body drag in a fresh main lays it at x -0.9), clearance (turn grip and Length row), fenders (one-sided metal per style, R1a side-0 hash pin, sides -1/+1/0/0, never saved), Amsterdam (each houseboat turned 12 and -25 degrees is laid back parallel at z 3.1, 6 of 6), moorings (one line, two at 10 m, nearest per end, none at 2 m, buoy at local x 3.3, rope and lantern anchors, houseboat post, never over a hull, absent from snapshot and history, links off, beached; Harbour/Courtyard/Amsterdam counts; WorldView held during drag steps, gone on release, undo, `_Fq` key). 10: - **Changed expectations (old -> new):** no existing check changed (the test diffs only add). Derived, not saved, so every pin and document md5 is unchanged: Harbour gains `buoy:136` (bow of boat 136 to a buoy at (16.5, 0.15, 2.2)) and boat 134's fenders go to the quay side only (side -1); Courtyard ("Rain on the canal") gains one line, boat 47 to bollard 140 (both fenders kept); Amsterdam gains six lines (55: bollard 38, lantern 34; 56: lantern 36, bollard 39; 57: lantern 37, rope 43) and its three houseboats fender on the quay side only; the quarter (`artwork.json`, boat 241) and the boatless journey scenes derive nothing (`c/probe/journey.log.out`). Harness: a `body` step type (both trees). Houseboat nap rule from R1b (`if _boat_roof(o): continue`) -> the box is an obstacle for others' spots and the roof stays a nap surface. A shipped Amsterdam houseboat stays at z 3.3 as saved; the magnet berths one at z 3.1 only when it is moved. 11: - **Full suite and --verify** (`tools/suite_all.sh c/suite` a, b, then `VERIFY_RES=2880x1800` v; battery 42%, Low Power Mode 0): **48 suites, 26,866 checks, 1 failure** (`c/suite/suite_all.out`): test_composition "scene_state.gd does not call the composition layer", a source-text check tripped by my comment naming `Composition.moorings` in scene_state.gd. Comment reworded (no code change); test_composition rerun **155/0** (`c/suite/fix/test_composition.log`); so 26,866 checks, 0 failures on the final code. R1b's 26,754 -> 26,866: +112 = decor +57, links +39, chemistry +12, creatures +4. `--verify` 2880x1800 exit 0 on try 1 (27 s), `PIXELPUNK_INTEGRATION []`, only ERROR the inherited `terrain_flowers.svg` (`c/suite/verify.log`). test_creature_reactions 6,871/0 and its rerun 6,871/0 (the known flaky check did not fire). 12: - **Captures** (`c/caps.sh`; `c/cap/pre-1`, `pre-2` from `c/pre-tree`, `after` live; camera passed after `--scene` in the args; 1600x1000; power `c/cap/power.txt`, battery 44-42%, Low Power Mode 0). Scenes from `c/probe/scenes.gd`: `c/scenes/moorings.json` = Harbour without its boats plus bollard 139 and two mooring ropes; houseboat 142 (10 m, dropped askew at 0.3 rad off and laid by the magnet along the east quay; bow to rope 141, stern to bollard 132), cabin 143 in the 2 m channel (bow to bollard 131, both fenders), narrowboat 144 (8 m, pier south face; bow to rope 140, stern to bollard 139), sail 145 (5 m, open water, buoy); `amsterdam.json`, `amsterdam-turned.json` (8, -6, 10 degrees), `amsterdam-magnet.json`. Views: `moor-day`, `moor-night`, `moor-close-day`, `moor-close-night`, `harbour-day`, `harbour-night`, `harbour-lane`, `ams-boats`, `ams-night`, `ams-close`, `ams-close-night` (pre and after, same camera), `ams-turned`, `ams-magnet` (after only). Sheets `c/sheet/r1c-moorings.png`, `r1c-amsterdam.png`, `r1c-harbour.png`, contact `r1c-after-contact.png`, all looked at, with `c/cap/after/moor-day/courtyard.png`, `moor-close-day/editor.png`, `ams-close/editor.png`, `ams-close-night/editor.png`, `harbour-day/editor.png`: the narrowboat's bow line to the pier rope, the sail boat's line to its red buoy, the Amsterdam bow line to the lantern post and the line to the bollard, the Harbour buoy by boat 136. The houseboat 142 sits partly under the Harbour stilt house in these cameras. Metrics `c/metrics.sh` -> `c/metrics/`: noise pre-1 vs pre-2 (rain) changes 3-49% of the frame (harbour-night 49.4%, moor-close-day 12.8%, >24 at most 3.6%); ab pre-1 -> after is of the same order (moor-close-day 14.3%, >24 2.9%; harbour-night 49.6%), so the frame figures do not separate the lines from rain; the counts do: identical pre-1 vs pre-2 in every view, after: moor views +2 draw calls and objects, harbour views +3, ams-boats +8, ams-night +2, ams-close +3, ams-close-night +2. 13: - **Drag measure** (`c/dragall.sh`, `c/drag.sh`; `c/drag/boat_quay.json` on the Harbour fixture: boat 134 dragged by its roof 2 m along the east quay, held 0.35 m off by the magnet on every step, and boat 136 dragged toward the pier and turned parallel to it at z 2.1; traced headless in `c/probe/probe4.gd`, `probe4.log.out`; pre and live alternating, 3 runs each, folded by `tools/feel_drag_report.js` into `c/drag/pre/`, `c/drag/live/`; power `c/drag/power.txt`, battery 37-36%, Low Power Mode 0, load 3.7-4.8 (corrected by the orchestrator from `c/drag/power.txt`, which reads up to 4.76)). Body step total p95 **pre 20.985 ms, live 21.477 ms** (+0.49 ms), against the 53.7 ms handle baseline (q/drag/live); median 18.244 -> 18.398; release median 34.2 -> 35.9 ms; 48 samples each. Each tree's three runs end on one document (live `ebd822c76fae`, pre `69f4f9490089`; the report prints "RUNS DISAGREED" only because the script has no undo or redo step). Pre run 3 first stalled three tries without a report (exit 0) and was rerun at 20:31 after live run 3. A first script (`c/drag-v1/`, p95 pre 19.624, live 20.134) is superseded: its legs ended over land or out of reach, so neither tree moved a boat to a berth (equal final digests). 14: - **Acceptance:** water-edge finder generic, terrain-only, ten Harbour probes: met (tests). Released 1 m off at 30 degrees -> parallel and 0.35 m off; Ctrl keeps 30: met. Lines: one at a bollard, two at 10 m with an anchor on each half of the hull, none at 2 m, buoy on open water, never in the save or undo: met. Lines never cross a hull, their own included: met after the fix pass (as first written this rested on fixtures without the own-hull case, and the review measured 3.35 m and 9.74 m of line over its own hull in two probe cases; now 0.00 m on Harbour, Courtyard, Amsterdam, `moorings.json` and both probe cases, `c/fix1/probe/probe.log.out`, and test_links samples every line every 5 cm against every hull). Lines never under the ground: **not met as first delivered** (the review found 5 of 6 Amsterdam lines, the Courtyard line and all 5 `moorings.json` quay lines 0.12-0.42 m inside the quay masonry), **met after the fix pass**: 0.00 m of rope under the ground on every line; lowest rope over land 0.110 m over the ground, at the coping (`c/fix1/probe/probe.log.out`). Two lines for a boat longer than 6 m: **not met everywhere** under the no-spring ruling: Amsterdam houseboats 55 and 56 (14 m) and `moorings.json` houseboat 142 (10 m) keep the bow line only, every anchor in their reach standing on the bow half (55: 34, 38, 41; 56: 36, 39, 42; corrected by the orchestrator from the recheck) (reported, spring lines not reintroduced). Fenders on the quay side only: met (four styles). Magnet's cost inside a boat drag within the 53.7 ms handle baseline: met, p95 21.5 ms (+0.5 ms over pre, 2.3%, under the 20% stop). Every Amsterdam houseboat parallel to its quay without hand turning: met (shipped parallel; turned 8/-6/10 degrees they are laid back, `ams-turned` -> `ams-magnet`; test 6 of 6). 15: - **Expected journey count deltas:** harbour-day, harbour-night, harbour-lane +3 draw calls and +3 rendered objects each (the buoy and its line for boat 136, boat 134's one-sided fender mesh); quarter and styles views none (no derived moorings there). 16: - Review: three reviews, Opus 5.5, fresh contexts. **(1) Review** 20:35-20:40, 959 changed lines by its own `diff -ru` against `29-R1-c-pre` (equal to `c/part.diff`; no foreign hunks; probe `c/review/probe.gd`, `probe.log.out`, `probe2.log.out`): **must-fix open**, three: mooring lines passed through the quay wall (Amsterdam 5 of 6, worst 1.73 m of rope inside masonry; Courtyard 0.19 m under; every quay line in `moorings.json`), a line could cross its own hull (3.6 m boat 3.35 m, 10 m boat 9.74 m), and the acceptance lines overstated both. Notes: the reworded comment could not be proven byte for byte (no pre-reword copy; the file predates the end of `suite_all.out`; test_composition 155/0 on the final file); `quay_berth` turns a boat to an arbitrary diagonal at a convex pier corner; the 14 m stern spring line (owner question); the Harbour buoy sits 1.2 m off the pier face. Confirmed: `water_edge_at` terrain-only and generic (defined for points over water, so R2b probes from the water side); Ctrl via `snapping_active()`; nothing derived in the save or undo; deterministic; turn grip and Length row clamp at the last clear value, one undo step; the R1b nap finding resolved; the `body` harness step additive; journey deltas +3 match; pins hold. Orchestrator ruling: no spring lines (owner-decisions.md). **(2) Fix pass** (below). **(3) Recheck** 21:06-21:13 (`c/review2/`): the live tree equals the snapshot plus `c/part.diff` plus `c/fix1/fix.diff`; the fix touches only the mooring derivation, `mooring_line` and test_links; probes rerun plus `probe4` (quay tops 0.4, 0.7, 3.6, 5.6 m, a set-back post, an askew boat, a narrow canal with the anchor on the far, higher quay, and a fuzz of 813 lines on Harbour, Harbour +2 m and Amsterdam): no rope under the ground (lowest 0.019 m), worst 0.01 m over any hull (sampling): **findings closed**. Notes carried to R1d: every line that reaches land is bent at the lip even where a straight rope would clear it, so on low quays (0.4 and 0.7 m tops, and so on a jetty) the rope dips onto the coping in a V (fuzz 21 of 303 Harbour lines, worst 0.41 m; none on the four shipped scenes); at a glancing angle the cleat-to-lip span can clip the coping by up to 0.13 m (askew boats only); test_links' crossing helper never checks a buoy line against its own hull (the code does; probes 0.00). Wording in this entry corrected by the orchestrator from the recheck (anchors of 55 and 56; bollard 132 amidships). **(4) Adversarial drag review** 21:13-21:20 (`c/review-drag/`, `derived.txt`, `warm-derived.txt`): the drag numbers above match the raw `performance.json` (load range corrected). Second set on the final tree, live first, battery 24 -> 23%, LPM 0, load 3.4-4.3: body-step p95 pre 21.597 [21.44-23.065] (5 runs), live 22.572 [21.901-23.96] (runs 2-5; live run 1 never moved the boat: a focus-out cancel measured silently, see below); median 18.762 -> 19.075. Live is +0.15 to +0.3 ms median and +0.5 to +1.0 ms p95 above pre in both orders, inside the run-to-run spread, level with R1a's 22.6 and well under 53.7: **met**. Not in the entry until now: the first quay-magnet engagement costs one frame of about 66 ms (worst frame of a warmed drag 87.8 ms live against 21.6 pre), likely the guide material compiling on first draw (`world_view.gd:5042-5071`): carried to R1d as a pre-warm. A body drag cancelled by focus-out is measured silently (`main.gd:1298-1320`, `:1696-1698`) and the report always says "RUNS DISAGREED" for a script with no undo step: carried as a harness note. Orchestrator: closed 21:22. 17: - **Fix pass (12 October 2026, 20:41-21:05, Opus 5.5 on this Mac; evidence `c/fix1/`, backups `c/fix1/backup/`, diff `c/fix1/fix.diff`).** The review found (1) mooring lines passing through the quay wall: one straight rope with a parabolic sag from the cleat (0.75) to a post on a quay whose top is 1.6 lay under the quay top where it crossed the lip (`c/review/probe.log.out`: Amsterdam 5 of 6 lines inside masonry, 55 stern 0.42 m under with 1.73 m of rope inside; Courtyard 47 -> 140 0.19 m; `moorings.json` all five quay lines 0.12-0.28 m), visible in `ams-close` as the bow line vanishing under the coping; (2) a line could run over its own hull (an anchor ahead of the bow or astern of the stern was taken from the far tip's cleat and the own hull was skipped by the crossing check: `c/review/probe2.log.out`, 3.35 m and 9.74 m); (3) this entry's "never cross a hull" rested on fixtures without that case. Fixed: `scripts/composition.gd` `_mooring_rope` finds where the line's plan run crosses from the water onto land (bisected to the millimetre), puts a lip point `LIP_OUT` 0.08 m out from the side `water_edge_at` finds and `LIP_RISE` 0.115 m over the quay top (past the coping, whose top is 0.1 over the lip), and returns two spans, cleat -> lip with its own sag and lip -> post with a sag capped by `_rope_sag` so the rope keeps `ROPE_CLEAR` 0.02 m over the ground and does not dip below the lip over the first `COPING_BAND` 0.2 m; a line that cannot be drawn so is not drawn (the next anchor is tried); entries gain `lip` and `sag2`, `mooring_spans(m)` gives the spans; lines over water alone are unchanged. Orchestrator ruling of 12 October applied: no spring lines, the bow line takes an anchor on the bow half (local x >= 0) and the stern line one on the stern half; `_over_hull` now skips only the anchor's own houseboat, never the line's boat (the buoy line too). `scripts/architecture.gd` `mooring_line` draws each span with `_link_cable` (same links mesh, built on the full sync only: no per-frame cost, no new MultiMesh). The drag path is not touched (lines are derived on release). Tests (`tests/test_links.gd`, test-first `c/fix1/testfirst/test_links.log.out`: **15 of 470 failing** on the unchanged code): on Harbour, Courtyard, Amsterdam, the moorings scene rebuilt from its recipe and the saved `moorings.json`, no sample of any span (200 per span) over land more than **5 mm** under the ground, and every line to a post on land bent at a lip at least 0.1 m over the edge height; the two probe cases and a 10 m boat with a bollard 0.8 m astern (one line, from the right tip); no line over any hull, its own included, on all of these. After: test_links **471/0**, test_decor 308/0, test_chemistry 425/0, test_composition 155/0, test_creatures 250/0, test_scene 1709/0, test_starters 393/0 (`c/fix1/suites/`; `c/fix1/scratch/` is the interim test_links run before the Amsterdam count was re-pinned). Review probe copied to `c/fix1/probe/probe.gd` (its scene loop restored and the rope sampled through `mooring_spans`), rerun on the fixed code: **0.00 m of rope under the ground on every line** (was up to 1.73 m), lowest rope over land 0.110 m; **0.00 m over its own hull** on every line and both probe cases (was 3.35 and 9.74 m); deterministic on all four scenes. Full suite (`tools/suite_all.sh c/fix1/suite` a, b, `VERIFY_RES=2880x1800` v, after the last script edit; battery 31-27%, Low Power Mode 0): **48 suites, 26,888 checks, 0 failures** (`c/fix1/suite/suite_all.out`); `--verify` 2880x1800 first run exit 1 on the known held-raise flake ("Holding the raise brush did not build a hill: 2.05", `c/fix1/suite/verify-try1.log`), second run exit 0, `PIXELPUNK_INTEGRATION []`, only ERROR the inherited `terrain_flowers.svg`; test_creature_reactions 6,871/0 and reruns 6,871/0. Captures (`c/fix1/caps.sh`, `c/fix1/court.sh`; pre-1 from `c/pre-tree` and after from the live tree in one sitting, power `c/fix1/cap/power.txt`, battery 31%): the part's views plus the Courtyard sample (`c/fix1/scenes/courtyard.json`, views `court-close`, `court-line`, `court-line-night`; `court-close`'s camera hides the line behind a lantern post, `court-line` sees it). Looked at (`c/fix1/cap/crops/z-ams-close-part.png` vs `z-ams-close-fix.png`, the night pair, `z-courtline-after.png`, `moor-close-day`): the Amsterdam bow line now climbs the wall face, runs over the coping and across the quay top to the rope turn at the lantern post, by day and at real night, where the part's vanished under the coping; the Courtyard line rises up the wall, over the coping to the bollard's foot; the narrowboat's lines climb the bank to the ring and the bollard. Sheets `c/fix1/sheet/fix1-amsterdam.png`, `fix1-moorings.png` (the reviewed part's after | fixed), `fix1-pre-after.png`, `fix1-courtyard.png`. Metrics `c/fix1/metrics.sh` -> `c/fix1/metrics/`. 18: - **Changed by the fix pass (old -> new):** Amsterdam lines **6 -> 4** (`mooring:55:stern:34` and `mooring:56:stern:39` dropped; the pinned check "Amsterdam derives six lines" -> "four lines: 55, 56, 57 bow, 57 stern"); `moorings.json` 5 lines + 1 buoy -> **4 lines + 1 buoy** (`mooring:142:stern:132` dropped: bollard 132 stands exactly amidships, local x +0.0000, and goes to the bow by the x >= 0 tie); Harbour (buoy 136) and Courtyard (47 -> 140) unchanged in count. Test "a 10 m boat beside a bollard gets two lines, bow and stern" (bollard amidships) -> "a 10 m boat with its one bollard on the bow half gets one line, the bow's". Full suite 26,866 -> 26,888 checks (+22, all test_links: 449 -> 471). Draw calls and rendered objects: **unchanged** against the part's after in every view (`counts-partafter-fixafter.txt`: identical), so the expected journey count deltas stand as written above (harbour-day, harbour-night, harbour-lane +3 each); pre -> fixed in this sitting: ams-boats +8, ams-close +3, ams-close-night +2, ams-night +2, moor views +2, harbour views +3, court-close and court-line +2 (`counts-pre1-after.txt`, `counts-court.txt`). 19: - **Not done:** cleats on the boats themselves are drawn only where a line ties (no idle cleats); the buoy has no bob of its own (it is a static decoration like the lines); boat 135 in Harbour (open water, edge within 4 m) gets neither line nor buoy, as the rule says; the moorings do not react to the boat's heave (lines tie at the rest pose, as R1b's strings). Windows not run. 20: - Platforms: Mac run; Windows not run. SOURCE: docs/visual-handoff.md (frozen 12 October 2026, working tree), lines 22-35 (Decor R1b) 22: **Decor R1b Boat styles (12 October 2026):** Opus 5.5 on this Mac, 18:40-19:20 (docs/decor-controls-ui-plan.md part "R1b", stage R1 Water craft, step `29-R1`, part b; owner E2: all six styles, 14 m maximum). A boat takes an optional `params.style`: `cabin` (absent: today's boat, byte for byte at every length), `narrowboat`, `barge`, `rowing`, `sail`, `houseboat`; Amsterdam's long boats are now houseboats and ship on. Snapshot: `build/backup-feel/29-R1-b-pre/` (orchestrator's, 18:39; the live tree equalled it at part start, no foreign hunks). Scoped backups `artifacts/decor/29-R1/b/backup/`; scratch tree `b/pre-tree`; part diff `b/part.diff` (below, `b/` = `artifacts/decor/29-R1/b/`). 23: - **Implemented.** `scripts/scene_state.gd`: `BOAT_STYLES`, `BOAT_STYLE_LENGTHS` (cabin 2.4-14, narrowboat 6-14, barge 5-14, rowing 2.4-4.5, sail 3-7, houseboat 6-14), `boat_style_asked`, `boat_style_fits`, `boat_style` (the style drawn: a style whose length is outside its range is drawn and boxed as the cabin boat, never an invalid document), `sail_mast_top`; `dimensions` (length, 1.7, 1.5) except the sail (mast top + 0.12 high); `validate_document` "Unknown boat style." for anything but the six or a building style. **Design reading:** every boat placed in the editor before R1b carries the shared params' `style: "canal"` (`Main.params`), so a building style on a boat reads as cabin and stays valid. `scripts/architecture.gd`: `boat(batch, accent, length, style)` dispatches to `_boat_cabin` (R1a's code, its hull moved into `_boat_hull` with the same calls in the same order) and `_boat_narrowboat`, `_boat_barge`, `_boat_rowing`, `_boat_sail`, `_boat_houseboat` on the shared `_boat_hull`; helpers `_boat_fenders`, `_boat_coil`, `_window_row`, `boat_canvas` (the Palette recolours narrowboat panels, barge tarpaulin, sail cloth, houseboat walls), `barge_perch`, `houseboat_anchors`, constants `HOUSEBOAT_ROOF` 1.46, `BARGE_WHEELHOUSE_TOP` 1.61. Narrowboat: low painted cabin less a 1.4 m fore-deck and 1.0 m stern deck, cream rule lines, square windows every 1.2 m (4 lit), stove chimney at the stern with `batch.vent`, water can, mooring pin, tiller. Barge: wheelhouse at the stern (1 light), hold under a ridged tarpaulin whose side skirt is `cloth_box` (CLOTH_HANG), lashings, windlass. Rowing: open hull, raised gunwales, two thwarts, shipped oars, no light. Sail: pale hull, cockpit, metal mast (1.4 + 0.55 x length), boom with a furled sail (cloth, hang) and a masthead pennant (cloth, flag), stays as cables, rudder and tiller, a riding light drawn with `batch.fixture_order = 1` (the lamplighter's round; `scripts/dusk.gd` needed no change: the round reads the fixture order). Houseboat: full-length flat-roofed cabin, large windows every 1.25 m (up to 6 lit), door, two pots on the roof, a post at each end of the roof for strings. `scripts/composition.gd`: `link_anchors(state, o)` (a houseboat's two posts) joins `_light_links` after the lamps and flags (a scene without a houseboat has exactly the anchors, pairs and keys it had; one boat's ends never pair; a boat-end key names the end). `scripts/creatures.gd`: barge wheelhouse roof is a sparrow perch (kind `wheelhouse`); a houseboat's roof is a cat nap `roof` (`terrace_top` reads `HOUSEBOAT_ROOF` for it, and the boat no longer crowds its own roof). `scripts/world_view.gd`: the decoration group key gains `_S