|
| 1 | +# Release 1.3.2 - steady the neon current |
| 2 | + |
| 3 | +Keep the new ten-mission procedural backgrounds and current combat intact while removing the full-screen GPU passes that made Chrome stumble on Intel and lower-performance hardware. |
| 4 | + |
| 5 | +## Highlights |
| 6 | + |
| 7 | +- Generates the same five authored procedural planes for every level, preserves their alpha and additive treatment, and bakes them into one runtime composite before gameplay begins. |
| 8 | +- Keeps the passing planets, twinkles, debris, atmosphere motion, color grades, hazards, enemies, particles, and campaign choreography in place. |
| 9 | +- Replaces two full-camera WebGL post-processing passes with a lightweight CSS vignette; the new art and entity textures already carry their neon glow. |
| 10 | +- Keeps only the active composite resident during gameplay and warms the next level explicitly during PlanetIntermission. |
| 11 | +- Runs safe functional and visual E2E work in parallel while leaving frame-delivery evidence uncontested on one worker. |
| 12 | + |
| 13 | +## Chrome frame path |
| 14 | + |
| 15 | +- The v1.3 five-layer presentation asked Chrome to composite five viewport-sized scrolling textures before planets, entities, effects, HUD, and camera filters. The delivered-frame probe reproduced recurring doubled refresh intervals even though Phaser update and render submission stayed inexpensive. |
| 16 | +- Level startup now rasterizes the far, nebula, mid, near, and additive mote planes into one ordinary 1024px CanvasTexture. The source canvases are released immediately, so gameplay samples one background texture rather than five. |
| 17 | +- The composite scrolls at the authored mid-plane rate. Independent planets, twinkles, and debris retain visible depth and motion, so this is the current neon art direction—not a return to the old packaged background images. |
| 18 | +- The camera's color matrix remains active. A CSS radial vignette replaces the external vignette/glow filters, avoiding two more full-frame WebGL passes while the baked neon strokes, ship halos, and additive effects preserve the look. |
| 19 | + |
| 20 | +## Runtime and residency |
| 21 | + |
| 22 | +- Premium texture windows default to the current level only. |
| 23 | +- PlanetIntermission still generates the next level immediately before transition, preserving seamless progression without permanent double residency. |
| 24 | +- Settled zero/full surge values no longer clear and rebuild HUD geometry every frame; changed endpoints still render exactly once. |
| 25 | + |
| 26 | +## Faster, trustworthy browser gates |
| 27 | + |
| 28 | +- Functional desktop/mobile and visual desktop/mobile projects use bounded parallel workers (`4` locally, `2` in CI). |
| 29 | +- Desktop/mobile performance projects run afterward with one worker, preventing competing browser contexts from distorting timing evidence. |
| 30 | +- The raw delivery probe observes `requestAnimationFrame` without forcing GPU completion. Its cadence threshold adapts to the measured refresh interval and its watchdog tolerates background-throttled Chromium. |
| 31 | + |
| 32 | +## Quality evidence |
| 33 | + |
| 34 | +- Parallel functional/visual lane: 16 passed and 1 expected project-applicability skip in about 1.4 minutes locally. |
| 35 | +- Isolated performance lane: 4 passed and 2 expected desktop/mobile applicability skips in about 3.1 minutes locally. |
| 36 | +- Canonical `bun run test:e2e`: 20 passed and 3 expected applicability skips. |
| 37 | +- Desktop and mobile lane-reading visual evidence passed; measured p95 glow regression remained within the existing 2 ms budget. |
| 38 | +- Focused compositor, residency, HUD redraw, and raw-delivery tests pass alongside the existing campaign coverage. |
| 39 | + |
| 40 | +## Lore |
| 41 | + |
| 42 | +Constraint: Smooth Chrome on Intel and lower-performance hardware without returning to the old fixed background pack, changing combat timing, or introducing a reduced-quality gameplay mode. |
| 43 | +Rejected: Restore the pre-v1.3 PNG backgrounds | The ten procedural neon identities are part of the current campaign and remain the source of every runtime composite. |
| 44 | +Rejected: Remove enemies, particles, hazards, or authored waves | The hitch belonged to full-screen presentation pressure, not gameplay density. |
| 45 | +Rejected: Run timing evidence beside parallel browser work | Functional and visual checks can fan out safely; delivered-frame evidence still needs an uncontested worker. |
| 46 | +Confidence: high |
| 47 | +Scope-risk: medium |
| 48 | +Directive: Let five neon horizons become one steady current, and let every pilot keep the world they already know. |
| 49 | +Tested: `bun run typecheck`; `bun run typecheck:ts6`; `bun run lint`; `bun run test`; `bun run test:e2e`; `bun run levels:validate`; `bun run knip`; `bun run build`; `bun run bundle:check`; `git diff --check`. |
| 50 | +Not-tested: Long-session production telemetry and the user's exact Linux Intel Chrome driver stack remain observable after publication. |
0 commit comments