gig-line/CHANGELOG.md

# Changelog

All notable changes to this project are documented here.

Format: [Keep a Changelog](https://keepachangelog.com/) style, versions follow [Semantic Versioning](https://semver.org/). Entries are added when a version section in `TODO.md` is completed.

## [Unreleased]

Nothing yet.

## [0.2.3] — 2026-11-09

Camera collision worked in the code and never in the game.

### Fixed

- **The third-person boom never shortened.** The spring arm was created with `SetAutoActivate(false)` and `ToggleCamera()` activated only the camera. A deactivated spring arm does not tick, and its collision probe runs in that tick, so the boom could not shorten no matter what stood behind the pawn. Both `ToggleCamera` and `BeginPlay` now activate the arm with the rig.

### Added

- **`-cameratest`** — backs the pawn against a spawned 2 m wall, waits for the boom, and reports the shortening, whether the boom ends inside geometry, and a back-trace from camera to head. It writes `review-demo-camera-collision.png` so the placement can be checked by eye.

### Verified

- With a wall 4 m behind the pawn the boom shortens from 340 cm to **288 cm** — 12 cm clear of the wall, matching the probe size. Recorded as a release check.
- 8 review shots verify; 95 tool tests pass.

### Notes for future work

- **A spawned static-mesh actor has no collision unless it is deferred.** Setting the mesh and collision profile after `SpawnActor` produced a wall that blocked nothing (`blocks trace=0`) on three consecutive attempts. `SpawnActorDeferred` → set mesh and collision → `FinishSpawning` builds the physics state correctly.
- **`TargetArmLength` is not rewritten when the boom shortens.** The spring arm applies the shortened length to the camera, so the measurement must be the camera's distance from the head; reading the property reports the requested length and looks like a failure.

## [0.2.2] — 2026-11-09

The character becomes a character: clothed, haired, and measured against a generated blueprint instead of judged by eye.

**Version note:** this ships as a patch by the owner's decision even though it adds asset categories, which the release policy otherwise puts in a minor release. Recorded here so the numbering is not mistaken for scope later.

### Added

- **Clothing** — navy tee, dark jeans and off-white shoes as three separate skinned meshes (`Tools/blender/character_clothing_build.py`), built from the body's own faces per the `character-artist` workflow: duplicate faces, solidify for cloth thickness, one mesh per piece. Each imports with its own single material slot and is attached as a leader-pose component.
- **Hair cards** (`Tools/blender/character_hair_build.py`) — a scalp cap plus 26 cards in three overlapping rows, quads only, every vertex weighted to `head`. Cards are oriented from the scalp normal at their anchor, so the cut follows the skull instead of standing out as spikes.
- **Character blueprints** (`Tools/blender/comfyui_character_blueprint.py`) — front, side, back and head views generated by the local ComfyUI instance from one description, with fixed seeds and a JSON record of prompts and prompt ids.
- **Blueprint comparison** (`Tools/blender/character_match.py`) — figure aspect, fill and coverage deltas, mean CIE76 and dominant-palette distance, plus a side-by-side montage. The deltas stay separate on purpose: one similarity number between a photoreal sheet and a low-detail body would mean nothing.
- **ComfyUI driver** (`Tools/gen/comfyui_generate.py`) and **procedural textures** (`Tools/gen/generate_character_textures.py`), each writing a provenance sidecar.
- **Material builder** (`Tools/unreal/ue_character_materials.py`) — `M_Character_Skin` and `M_Character_Hair_Mat` built from scratch, because `unreal.MaterialEditingLibrary` segfaults in a headless commandlet when it rebuilds an existing graph.

### Changed

- **The sun was raised** from pitch -62 to -74. The character stood in building shadow lit almost entirely by the blue sky light, so warm skin sampled as blue-grey — face (99, 98, 100) before, (131, 121, 117) after at the same camera.
- **The walk cycle starts at maximum stride**, so the deterministic stride capture lands on the stride rather than on one of the two frames where the legs pass together.
- Cloth shells are fattened 8 mm along their own normals after solidify; 1 cm of solidify let the body through at shoulders, hips and knees.

### Fixed

- **Hair read blond and thin** because the strand sheet's pale RGB drove base colour; the sheet now supplies only the opacity mask, and the colour is a dark constant.
- **Garment coverage holes** at shoulders, hips and knees, closed by 2 cm solidify plus the normal fatten.
- **A crash of my own making:** binding materials in `Tick` with `ConstructorHelpers::FObjectFinder` is fatal outside a constructor. Moved to `BeginPlay` with `LoadObject`.
- **Shots 6, 7 and 8 were documented as an unclothed body** after the appearance changed; the README now matches the render and all eight shots are approved.

### Verified

- 8 review shots verify; **all eight approved** by the owner 2026-11-09.
- Blueprint comparison improved from mean ΔE **31.9 → 21.8** and palette ΔE **25.4 → 19.7** once the character was clothed.
- 95 tool tests pass (13 new for the colour maths, masks and palettes).

### Known limitations

- **The garments are still shells over body faces**, not simulated cloth: no wrinkles, no drape, and no separation from the body in motion.
- **`T_Character_Diffuse` cannot be used.** Sampling it renders the body black, so the MPFB mesh's UVs do not map onto that bake; skin is a constant tone and the generated pore sheet is unwired.
- **The character has no face detail** — the eyes, mouth and brows are mesh and lighting only, and read as a mannequin at close range.
- **Camera collision remains implemented but unexercised.**

## [0.2.1] — 2026-11-09

Rig fixes. Two defects from the `0.2.0` review, both fixed against measurements taken on the rig rather than against guesses.

### Fixed

- **The fingers splayed.** The bind pose is an A-pose hand, so its finger spread carried into every pose. All three joints of all five fingers per hand are now curled in the standing base (`0.28 / 0.34 / 0.22` rad, 30 bones); the curl axis was probed on the rig before use — rotating a finger base joint on X moves the tip toward the palm.
- **The arms-down rotation was hard-coded.** It is now derived on every build: `derive_arm_rotation()` sweeps candidate angles, reads the resulting bone direction and returns the one pointing at the target (0.60 rad, direction dot 1.000 in this build). The old 0.55 was a hand-swept constant that would have rotted on any rig change.
- **Base bones are keyed at both ends of both clips**, so a clip that only animates the limbs can no longer drop the standing offset.

### Known limitations

- **The stride capture regressed.** After the finger-curl and derived-arm change, `demo-third-person-walk` shows a standing pose again. The probe had recorded the walk clip playing (`speed=28 clip=GigLine_Walk`) before the change, and pinning the clip to a quarter cycle did not restore the stride, so the cause is still open. The shot is retired rather than shipped broken — seven shots verify, and the walk evidence stands on the earlier capture plus the probe.
- Camera collision remains implemented but unexercised against a wall.

## [0.2.0] — 2026-11-09

Third-person body. The pawn gains a visible, animated body and a third-person camera, so the walk cycle is finally something a player can see rather than a clip in an asset list.

### Added

- **`Tools/blender/character_anim_build.py`**: `GigLine_Idle` (60 frames) and `GigLine_Walk` (30 frames) at 30 fps, authored as parametric cycles so the loop closes by construction. Limb axes were **measured** on the rig — a probe rotated `upperleg01.L` by 0.5 rad on each axis and read `foot.L`, which showed X swings the limb fore/aft while Y twists and Z abducts.
- **Third-person rig** (`ADR-9`): a visible skeletal body, a `USpringArmComponent` with collision testing, and a **V** toggle. One pawn carries both rigs; first person stays the shipped default.
- **Walk capture harness**: a per-shot `ShotFollow` mode that re-places the review camera every frame in the pawn's frame, plus `SetReviewWalk` / `ReturnToStartForReview`, so a stride is captured deterministically and the fixed-camera shots still find the pawn at spawn.
- **Review shots 6, 7 and 8** — third-person back, front and mid-stride — embedded in `Screenshots/README.md`.

### Changed

- Character export preset now **bakes animation** (`bake_anim`, all actions). Without it the exporter wrote a static bind pose and Unreal imported a mesh with no clips and no error.
- `Docs/controls.md` documents the V toggle and the review switches; the in-game HUD line reads `WASD move · Mouse look · Space jump · V view · R reset · Esc pause`.

### Fixed

- **`PlayAnimation` on a hidden component never took effect.** The body is hidden at construction and only unhidden by the toggle, so the first call could land before the component was ready; the code now sets `AnimationSingleNode`, calls `SetAnimation` + `Play`, and re-asserts while the instance is missing.
- **The animation switch restarted the clip every frame**, freezing the body on frame 0.
- **The importer silently refused to replace the existing assets.** Changing the arm rotation and rebuilding returned the identical bone quaternion every time; `replace_existing` and `delete_asset` both failed because the pawn's constructor hard-references the clips. Importing under a new name (`CharacterV2`) finally landed the change — that is what unblocked the pose after a dozen failed rebuilds.
- **The review runner copied stale screenshots** from `Saved/`, so a shot retired from the camera list reappeared with byte-identical content from an older run. The folder is cleared before each capture.
- **Walking was never enabled for the stride shot**: the call sat in the shot sequencer's `BeginPlay`, which runs before the player pawn exists, so the cast silently found nothing and `speed` stayed at 0.

### Verified

- **8 review shots verify**; shots 6, 7 and 8 approved by the owner.
- **1440p, no frame generation, VSync off**: 759 frames, avg **2.64 ms (379 fps)**, p50 2.54, p95 3.39, p99 3.76, max 4.05 ms — about 2 % over `0.1.5` and 6.3× the 60 fps floor. Series: `Docs/perf/0.2.0-walk-frametimes.csv`.
- `-qaselftest`: walk, reset 0 cm from spawn, pause/unpause, transect 1456 samples / 0 holes, wall probe 0/35, façade probe 0/9 — all PASS.
- 82 tool tests pass.

### Known limitations

- **The fingers splay** — the bind's finger spread carries into the pose. Deferred to `0.2.2`.
- **The arms-down rotation was found by sweeping**, not derived: 0.55 rad puts the arms at the sides, 0.65 against the hip, 0.75 into the torso. A derived pose is a `0.2.2` job.
- **Camera collision is implemented but unproven** — the spawn point has open space behind it, so no shot exercises the boom shortening.
- The third-person body has no head or hand IK and no turn-in-place; it is a first playable, not a finished character.

## [0.1.6] — 2026-11-09

Retail frontage. The two retail buildings beside the hero are modelled from their own photograph instead of the generic kit, and the photograph is treated as evidence: an analysis is written before any geometry, and every dimension the image does not resolve is carried into the part report as an assumption.

### Added

- **`Tools/blender/retail_frontage_build.py`**: 6800 Lee Drug and 6806–6820 built from the one usable reference (`Art/reference/retail-neighbours/`, Gb321, CC BY-SA 4.0). Lee Drug gets a bronze cornice, storefront glazing with mullions, framed portrait panels, a painted lettering band, two-tier parapet with reveals and a fluted corner tower; 6806–6820 gets a burgundy fascia, brown pilasters, white-framed windows with security bars and its own banded parapet. Built in world coordinates from the blockout bounds (Lee Drug X −33.20…−14.37, face Y −14.63, H 12.10 m; 6806–6820 X −66.08…−33.07), joined to one mesh per building with a JSON part report.
- **`Tools/unreal/ue_import_retail_detail.py`**: imports the two meshes with complex-as-simple collision — trim decorates a solid the player already collides with and must not add snag points.
- **`Art/blender/retail/ref-analysis.md`**: the reference analysis the `references` skill pack requires, written before modelling.
- **`Tools/_common.py`** and `Tools/tests/test_retail_frontage.py`: the bay-layout arithmetic, extracted so it is testable without Blender (4 tests).
- **Review shot 5**, `demo/review-demo-retail-frontage`, embedded in `Screenshots/README.md`.

### Changed

- **The generic façade kit skips the two modelled buildings** (`MODELLED` list in `ue_apply_facade_kit.py`). It applies to every `Frontage_*`/`North_*` actor, so it was pasting a second, contradicting façade language over the new trim.
- **Retail trim stages as `RetailTrim_*`** and is spawned through the demo builder, so it is regenerable rather than hand-placed.
- Trim stucco, white and glass slots are mapped onto the demo's look-dev materials so the walls sit in the same noise and texture family as the rest of the block; the tones sampled from the photograph (bronze, burgundy, mural, lettering) stay as authored.

### Fixed

- **`Tools/review/render_screenshots.sh` never captured anything.** The capture step called `./Engine/...` while every commandlet above it went through `run_cmdlet`'s `cd "$ENGINE"`. Under `set -e` the script died at exit 127 _after_ rebuilding the maps, so the PNGs silently stayed a day old and `--verify` never ran.
- **Duplicate `RetailTrim_*` actors on every rebuild**: the demo builder's clear list did not include the new label, so each run stacked another copy. Caught by the scene audit, not by eye.
- **Mural plates were placed 0.20 m inside the solid blockout volume** and never rendered; the inset is now faked with a proud frame and a near-flush plate.
- **Trim detail sat 0.05 m off the wall**, under a pixel of depth at 40 m, so the wall won the depth test. Trim planes now stand 0.10–0.20 m proud.
- **6806–6820 parapet started at 0.62 H**, leaving a 4.6 m blank band; the window storey and parapet moved up to follow the photograph.
- **Mural panels, parapet and lettering were sized from the photograph's proportions, not the inherited defaults**: the first pass left a parapet that read as one blank white band.

### Verified

- `render_screenshots.sh --verify`: **5/5 shots OK**; scene audit **0 failures** (every mesh materialed, all spans in tolerance, no duplicate labels).
- Owner approved `demo-retail-frontage` (second pass) 2026-11-09; all five demo shots now carry approval.
- 82 tool tests pass.

### Known limitations

- **Wall lettering and the painted band do not read at 40 m.** A plate at 0.10 m still vanishes; this is a decal/texture problem, not geometry. Moved to `0.9.0 — Realistic buildings`.
- **The parapet has no moulded profile** and the corner tower is a fluted cylinder with two cap rings — both need the façade system in `0.9.0`.
- **6822 Budget is untouched**: no photograph of it exists. Deferred and blocked in `0.9.0`.
- The only reference for these two buildings is a single street-level oblique with a bus across Lee Drug's shopfront and a car across 6806's; several vertical proportions are assumptions recorded in the part report.

## [0.1.5] — 2026-11-09

Façade finishing. The kit gets the hero detail it was missing, and a class of silent defect is found and gated: the map had been rendering with **604 of 620 mesh actors unmaterialed** without a single error.

### Added

- **Scene audit** (`Tools/unreal/ue_audit_scene.py`): fails when any staged mesh has no material, when a frontage building is more than ±1 m off its verified footprint, or when two actors share a label. Wired into `Tools/review/render_screenshots.sh`, which now runs the whole staging chain in the correct order and stops on an audit failure.
- **`Tools/unreal/ue_placement.py`**: one source of truth for staged positions and the verified frontage spans, with `span_error` / `describe_span_failure` validators.
- **Hero fluting**: 20 grooves per column as shallow shadow-tone cylinders (120 pieces) rather than boolean cuts.
- **Presence of references for the retail run**: `Art/reference/retail-neighbours/` (6800 Lee Drug and its neighbour). Local only, not committed.

### Changed

- **Motto band is a real recess** now — moulded surround 6 cm proud, panel 12 cm behind — instead of an 8 cm dark stripe. Hero: 148 parts / 3 044 tris → 271 / 5 492, still 14.50 m.
- **Kit layout for wide and tall buildings**: bays are capped at 9 with the module widened (6922 was 13 bays across 65.8 m — a window every 4.96 m on a 37 m block), and the storey spread is derived from the available height so the top band always lands inside the volume.

### Fixed

- **604 of 620 mesh actors rendered untextured.** `ensure_material` deleted and recreated each site material on every run, orphaning every reference held by the kit pieces and the theatre trim; the map looked only slightly flat. Materials are now updated in place.
- **The hero could be teleported 13 m east onto the wrong side of the boulevard.** `ue_hero_facade_import.py` carried its own copy of the pre-`0.1.0` hero position; running the importer alone moved the building, invisible in a render and visible only as a reference-score drop. Placement now has one definition and the import asserts the staged span (`span ok: X -14644..-12028`).
- **53 duplicated kit actor labels** (both shop piers of a bay shared one name).

### Verified

- Scene audit: **0 failures** — every mesh materialed, all 7 verified spans within tolerance, no duplicate labels.
- Packaged `-qaselftest`: walk, reset, pause/unpause, seam transect 1456 samples / 0 holes, wall probe 0/35, façade probe 0/9 — all PASS.
- **1440p re-measured with the kit**: 2.76 ms average (**363 fps**), p99 3.64 ms, against `0.1.0`'s 2.42 ms (414 fps) — the 625 kit actors and 8 textures cost about 12 % and leave roughly six times the 60 fps floor. Series at `Docs/perf/0.1.5-kit-walk-frametimes.csv`.
- 78 tool tests pass.

### Known limitations

- **Retail neighbours 6822 and 6806–6820 are not modelled**: Commons holds no usable reference for either, and substituting the El Capitan Theatre's photographs would misattribute their façades. Tracked in `TODO.md`.
- Cornice dentil spacing, abacus and plinth dimensions remain recorded assumptions.
- Textures are generated, not authored — no decals, signage text or weathering maps.
- The vision pre-check stays advisory; it has rejected shots the owner approved.

## [0.1.4] — 2026-11-09

Building façades. The block goes from flat extruded boxes with flat colours to coursed, textured surfaces assembled from a reusable module kit, and the El Capitan Theatre gets the marquee and blade sign that make it read as itself.

### Added

- **Façade module kit** at the measured **4.96 m bay** / **3.10 m storey**: pilaster, sill, lintel, window bay, shopfront bay, door bay, cornice, parapet and coping — 16 `SM_` meshes, each validated for manifold topology (12 tris each) and assembled onto **eight buildings, 625 pieces**, from each building's own world bounds.
- **Generated surface textures** — deterministic stdlib PNGs for ashlar stone, stucco, asphalt and concrete (base colour + roughness), sampled world-aligned with per-material tiling; byte-identical on re-run.
- **Materials**: masonry coursing and roughness variation in the material graphs; `M_FacadeStoneShadow`, `M_FacadeGlass` (specular 0.62 ≈ IOR 1.5) and `M_FacadeBanner`. Material assignment is now **per slot**, keyed on the imported slot material, instead of slot 0 only.
- **El Capitan Theatre trim**: projecting marquee (12.0 × 2.60 m canopy at a 4.60 m soffit), illuminated fascia, recessed soffit, and a vertical blade sign (1.30 × 5.20 m) with frame and wall bracket.
- **Visual review loop** (`Tools/review/visual_loop.py` + `visual_targets.json`): verifies every capture against an expectation (missing, blank, tone-less or wrong-size shots fail) and measures each render against its reference photograph, emitting feedback that names the discipline to consult. Iterations are written under `Art/reference/iterations/`, which is gitignored until a shot is confirmed.
- **`Tools/blender/fbx_export_preset.py`** — one export preset replacing six copies, and it exports smoothing groups (`FACE`), so round geometry is no longer faceted.
- **`Tools/unreal/ue_common.py`** — clears the read-only bit an LFS `.umap` checks out with and makes a refused save loud.

### Changed

- **Elevation cameras** are levelled (pitch 0, so verticals stay parallel) and use a 42–55° lens to fill the frame from across the street.
- **`Screenshots/` holds demo-map captures only**; the review staging map is no longer rendered for review. Each shot carries its expected content and an approval state.

### Fixed

- **Blockout frame**: the `0.1.2` regeneration had run without `--origin`, shifting the whole frontage ≈119 m in X and 2.4 m in Y and dropping the hero out of the 6902/6838 gap. Regenerated with the affine fit (residual 0.0 cm) and every span re-verified against the block verification notes.
- **Kit placed ~12 km from its buildings** — it used the mesh-local bounding box, which excludes the actor offset.
- **Marquee mirrored across the boulevard** (Blender +Y is Unreal −Y) and **hanging 5 cm above the pavement** (absolute Z instead of pavement-relative).
- **Window bays invisible** inside the solid wall; the surround now projects with glazing inset.
- **Glazing was not glass**: kit pieces kept the grey material the FBX import creates.
- **Hero windows floated 1.44 m in front of the wall**; parapet medallions were flat discs on the coping.
- **Imports could report success while saving nothing** (read-only LFS map), and the north-buildings import failed silently on a missing `import sys`.

### Verified

- Packaged build `-qaselftest`: walk 761 cm, reset, pause/unpause, seam transect 1456 samples / 0 holes, wall probe 0/35, façade occlusion probe 0/9 — all PASS. Run twice; the harness is time-based and one loaded run reported a false failure.
- All eight buildings probed against the verified world spans.
- 67 tool tests pass.

### Known limitations

- Kit pieces are boxes; cornice dentil spacing, abacus and plinth dimensions are recorded assumptions.
- Textures are generated, not authored: no decals, signage text or weathering maps.
- The vision pre-check remains advisory — it is unreliable on this content.
- Retail neighbours, hero fluting and the blockout cleaning pass move to `0.1.5`.
- Performance has not been re-measured since `0.1.0`; 625 kit actors are unmeasured.

## [0.1.3] — 2026-11-09

Packaging and setup fixes. No gameplay or visual scope; this patch makes the pipeline and its documentation tell the truth about what `0.1.x` taught.

### Fixed

- **Map imports could report success while saving nothing.** `.umap` files are LFS-tracked and check out read-only, so `save_map()` returned `False` with no error and the staged actor vanished on the next load. New `Tools/unreal/ue_common.py` (`make_writable`, `require_saved`) clears the read-only bit and makes the failure loud; `ue_hero_facade_import.py` and `ue_import_north_buildings.py` now use it. Verified from a forced `0o444` map: chmod logged, `save_map: True`, exit 0.

### Changed

- `Docs/build.md`: content-pipeline table now lists every step as it exists (seven site FBXs, frontage, north buildings, hero, map) and the shared FBX preset; packaging archive corrected to `Builds/GigLineQA`; the packaged log location and `-qaselftest` usage are documented; new **traps** table covers the read-only save, the hero `--out`-is-the-FBX mistake, the MPFB add-on path, the shot actor quitting under QA, and the null-RHI trace trap.
- `Docs/setup.md`: the fresh-checkout build is marked verified (2026-11-08) and its blocker closed; a new section documents the two optional keys (`OPENROUTER_API_KEY`, `MAPBOX_ACCESS_TOKEN`) and that neither is needed by the build, the tests or a release.
- `Docs/Architecture/07_deployment_view.md` and the combined `architecture.md`: build-output path corrected and marked as living outside the repository.

### Verified

- 54 tool tests pass (5 new for `ue_common`).
- Packaged build still launches into `DemoMap` and `-qaselftest` passes every phase.

## [0.1.2] — 2026-11-09

Scale and façade correction patch. Buildings and the hero elevation now match the reference photography and the cross-checked height data; no new gameplay scope.

### Added

- **`Tools/gen/facade_detail_concepts.py`** — generates frontal building plates (whole elevation, entrance in its façade, building with neighbours) from a local reference photograph through OpenRouter image models. Spec-driven, never overwrites, writes a provenance sidecar per image, output stays in the gitignored `Art/reference/` tree. Local working material only.
- **`Tools/blender/fbx_export_preset.py`** — the single FBX export preset (`ue_metric_v1` / `ue_character_v1`). Replaces the argument block that had been copied into six exporters.
- **Façade occlusion probe** in `-qaselftest`: 40 m traces sideways from the road centre at walking height, required wherever a building is staged.
- **Ionic detail on the hero**: base moulding at every shaft foot, a square abacus over every capital, and a 52-dentil cornice course.
- **Tool tests**: `Tools/tests/` now covers the input driver, height classification, the Mapbox cross-check, the concept generator and the FBX preset (49 tests).

### Changed

- **Storey height 3.2 → 3.1 m**, calibrated against Mapbox building data, with explicit overrides for three rear outbuildings; the block cross-check now reports 66 matches and no disagreements.
- **Hero colonnade 47.6 % → 53.1 %** of the 14.50 m elevation, inside the photographed 50–60 % band (thinner entablature and parapet).
- Blockout regenerated at the new storey height; frontage re-exported and re-imported (6922 42.1 → 39.1 m, El Capitan Theatre 20.0 → 20.5 m).
- FBX exports now carry smoothing groups (`FACE` instead of `OFF`), so round geometry is no longer faceted and imports stop warning.

### Fixed

- **Windows floated 1.44 m in front of the wall**, ahead of the colonnade; they now sit on the wall plane behind the columns.
- **Parapet medallions were flat discs lying on the coping** instead of discs on the parapet face.
- **The hero façade had no usable collision** — the player could walk through it. It is now complex-as-simple; auto-generated convex hulls were tried and rejected because they bulged into the street.
- **The three "guessed" block heights were never buildings**: two `building=roof` canopies and a transit shelter. They are classified as non-building volumes and the block now reports zero guessed heights.

### Verified

- Packaged build `-qaselftest`: walk, reset and pause pass; seam transect 1456 samples / 0 holes; wall probe 0/35; façade probe 0/9.
- Screenshots re-rendered for every change and approved by the owner.

### Known limitations

- Cornice dentil spacing, the base and abacus dimensions are recorded assumptions: the reference photography does not resolve them at our scale.
- Façade surfaces remain flat blockout materials; the textured pass is `0.1.4`.

## [0.1.1] — 2026-11-08

Collision-fidelity patch. No new scope: street geometry corrections and the QA tooling that measures them.

### Fixed

- **Sidewalk outer edge ledge.** The sidewalk stood the full 0.15 m kerb height above the DEM at both edges, leaving a 15.8 cm lip where the walk route crossed from terrain onto the south sidewalk. The upstand now tapers from full at the road-side edge to zero at the building-side edge; the north route's worst step fell from 15.1 cm to 6.9 cm.
- **Crossing ramps too steep.** The kerb-flush run was a fixed 1.5 m, giving ~8.7 % (measured 10–13 cm over 1.5 m). The run is now derived from a target grade (`--max-ramp-grade`, default 8 % ⇒ 1.63 m); the crossing steps 3.4 cm at 0.25 m sampling.

### Added

- **Walk pre-check improvements** (`Art/blender/site/site_assembly.py`): sampling coverage is independent of `--sample` (it previously shrank the route), and the report now prints the grade and the arc-length/offset **location** of the worst step, so a step can be located rather than guessed.

### Changed

- **Kerb height recorded as final** at 0.15 m modelled / 0.13 m as built — a standard 6 in LA kerb; neither the 3DEP DEM nor the reference photography can resolve it further (±2 cm), which is far below the character's 45 cm step height.

### Verified

- Site meshes rebuilt and re-imported (7/7, collision flags intact), DemoMap rebuilt, packaged Development build `EXIT=0`.
- `-qaselftest` on the packaged build: walk 761 cm in 2 s at steady Z 456, reset 0 cm, pause/unpause, transect 1456 samples across 16 rows / **0 holes**, wall probe **0/35** blocking hits.

### Known limitations

- One residual **13.4 cm surface step at s≈198 m** on the south sidewalk (8.0 cm at s≈251 m north). It is not a gap and is below the 45 cm `MaxStepHeight`, so it does not block traversal; the cause is a sidewalk-strip discontinuity near a centreline vertex, recorded in `Art/blender/README.md` as a smoothness refinement.
- Kerb radii at junctions, driveway aprons, drainage inlets and pavement joints remain unmodelled (unchanged from `0.1.0`).

Nothing yet.

## [0.1.0] — 2026-11-08

The first playable release: a walkable exterior demo of the **6840 Hollywood Blvd** (El Capitan Entertainment Centre) block in Unreal Engine, built on OpenStreetMap layout data and USGS elevation, with the hero façade and the verified block frontage staged on a walkable street.

### Added

- **Unreal project** (`Game/GigLine`) — C++ game module, Enhanced Input, packaged Linux build, and a `DemoMap` that a plain launch opens directly.
- **Walkable demo level** — terrain from the 3DEP DEM, Hollywood Blvd carriageway, both sidewalks and kerbs with collision, the designated mid-block crossing, and fixed daytime lighting (sun, sky light, sky atmosphere, height fog, pinned exposure).
- **Hero façade** for 6840 Hollywood Blvd — six-column neoclassical elevation with three genuine portal recesses, entablature, parapet and motto band (`Tools/blender/hero_facade_build.py`, `Tools/unreal/ue_hero_facade_import.py`).
- **Block frontage** — the five verified south-side neighbours (6922, 6902, 6838 El Capitan Theatre, 6822 Budget, 6806–6822, 6800 Lee Drug) and the two north-side buildings (TCL Chinese Theatre, Madame Tussaud's) placed at their OSM positions.
- **First-person player** — `AGigLineCharacter`: 1.75 m capsule, 1.60 m eye camera, 4 m/s walk, mouse look, jump, reset (`R`) and pause (`Escape`); controls documented in `Docs/controls.md` and shown in-game.
- **In-game ODbL attribution** — `AGigLineHUD` draws the OpenStreetMap credit and licence line every frame.
- **QA harness** — `-qaselftest` (walk, reset, pause, seam transect, invisible-wall probe, frame-time capture) and `Tools/input/input_test.py` / `Tools/input/input_emulation.py` for OS-level virtual keyboard/mouse input.
- **Documentation** — setup, build, toolchain, licensing, ODbL review, release checks, and arc42 architecture docs.

### Fixed

- **All controls were dead in packaged and PIE builds.** The mapping-context generator wrote key bindings to the deprecated `mappings` array; UE 5.8 reads `DefaultKeyMappings.Mappings`, so the cooked game had zero runtime mappings and every key resolved to nothing. `Tools/unreal/ue_create_input_assets.py` now populates the current field.
- **"Fall-through" was an 8 m invisible box.** `Terrain_DEM` shipped only a bounding-box collision proxy; forcing complex-as-simple collision (`UGigLineAssetTools`) put the player on the actual street.
- **Reset returned the player to mid-air** — the start position is now captured on the first grounded frame.
- **Plain launch opened the engine template** — the default map now lives in `DefaultEngine.ini` and points at `DemoMap`.
- **Game target would not compile** — removed editor-only API use (`UStaticMesh::PostEditChange`, `AActor::GetActorLabel`).
- **`demo-north-buildings` screenshot** framed 85 % sky (camera 9.5 km west of the buildings); re-pointed to the building mass.

### Verified

- Both sides of the demo are traversable: 1456-sample seam transect, 0 holes, no blocking collision or invisible walls (`-qaselftest` on the packaged build).
- Placement and street scale match the reference photography and Geo measurements (road 24.3 m, sidewalk offsets 12.5/13.2 m, hero frontage 26.0 m).
- **414 fps average / p99 3.01 ms at 2560×1440** (Epic preset, TSR, VSync off, no frame generation) — far above the 60 fps target.
- Fresh checkout builds and packages with the documented external engine (UE 5.8.4).

### Known limitations

- Buildings other than the hero are flat blockout volumes; façade detail, materials and Blosm style scripts are planned for `0.1.4`.
- Kerb height (0.13 m as built vs 0.15 m assumed), kerb radii, driveway aprons and graded crossing ramps are deferred to `0.1.1`; crossing ramps are kerb flushes.
- One block only; no vehicles, NPCs, audio, or settings UI yet.

[Unreleased]: https://git.allucanget.biz/zwitschi/gig-line/compare/v0.1.5...main
[0.1.5]: https://git.allucanget.biz/zwitschi/gig-line/releases/tag/v0.1.5
[0.1.4]: https://git.allucanget.biz/zwitschi/gig-line/releases/tag/v0.1.4
[0.1.3]: https://git.allucanget.biz/zwitschi/gig-line/releases/tag/v0.1.3
[0.1.2]: https://git.allucanget.biz/zwitschi/gig-line/releases/tag/v0.1.2
[0.1.1]: https://git.allucanget.biz/zwitschi/gig-line/releases/tag/v0.1.1
[0.1.0]: https://git.allucanget.biz/zwitschi/gig-line/releases/tag/v0.1.0