Files
pipelines/app/notes/arc-quick-switch.md
T

36 lines
2.1 KiB
Markdown

# Arc swap quick switch
`Arc swap` is a new persisted choice beside Swap. The outgoing active chat bows
left while the incoming chat bows right, trading vertical positions together.
The paths use opposite half-sines and a slight planar card rotation. They remain
opaque and settle to the exact prior Swap endpoints. Paint transforms use the
existing affine path; no hinge rotation or new 3D/backdrop path is introduced.
The 440 ms base clock, quintic easing, global speed control, reduced-motion
behavior and Tape default are unchanged. Existing persisted style names remain
valid. The Settings picker enumerates the new choice automatically and its
preview renders the same real workspace/motion code as chat.
Verification:
- 74 focused motion/workspace/preview tests passed (`/tmp/memwatch/out.x37AcW`).
Coverage includes persistence, exact endpoints, finite/bounded geometry,
opposite arc sides, retained live drafts and style selection in the preview.
- The subsequent 13-test style rerun passed (`/tmp/memwatch/out.38HNQA`) after
strengthening the three-pane cycle test to retain every draft and a custom
0.12 peek ratio. Reduced motion settles instantly.
- Analyzer clean (`flutter-analyze.vbo9ds`; final rerun recorded by manager).
- Full-app fake-gateway fixture build passed at `build/probe-arc`.
- Phone-layout browser switch-and-return passed, with the original draft read
back after both exchanges. Inspected contact sheets and returned-state PNG:
`/tmp/review/arc-swap-full-app/`. Script `/tmp/arc-swap-capture.json`.
- Settings browser interaction chose Swap, then Arc swap through the real
dropdown, asserted the stored `arcSwap` value, and filmed Replay. Inspected
selection PNG and preview contact sheet: `/tmp/review/arc-swap-settings-v3/`.
Script `/tmp/arc-swap-settings-capture.json`. An earlier capture overscrolled
the picker under the pinned header; corrected to a 100px scroll.
Software Chrome is qualitative browser evidence, not Android FPS or wall-time
measurement. No native device test, live chat mutation, server restart, pipeline
change, or deployment was performed. Capture drivers completed and cleaned up
owned browsers and loopback servers.