Draw static states as the completed braille path

[lvwerra]

Follow-up to #82 for the operator's status-mark feedback: non-working states should read as the working braille spinner's completed path, and that mark must render identically in every surface that consumes it.

What changed

The working sequence remains ⠋⠙⠹⠸⠼⠴⠦⠧. Waiting, idle and stopped draw the same square-cornered hollow rectangle over the six-dot extent:

state	shape	treatment
working	moving braille perimeter	accent
waiting	completed hollow path	accent
idle	completed hollow path	muted
stopped	completed hollow path	muted at 50% opacity

That is intentionally one static silhouette. Waiting, idle and stopped are not shape-distinguishable in grayscale; they use the requested accent / muted / dim treatment and surrounding state text where present.

Context parity

#84 places the same .status inside .ph-title, whose text is weight 600. Before this revision the mark inherited that weight: the pane-title ⠋ rendered 6.00 × 5.875px / 16.34px² alpha area, versus 5.50 × 5.375px / 10.04px² at weight 400 in the sidebar and reader. That confirms the operator's heavier, incompletely-disappearing dots were a real font-weight change, not a second mark implementation.

The shared contract now includes --mark-weight: 400. Every renderer owns its family, 12.5px size, weight, and line-height, so a 600- or 700-weight parent cannot change it. This covers state marks, Overview progress/group marks, and the reader's working mark.

The cell was centred before, but its ink was not. The natural braille ink sits 0.90625px above the cell centre; the 600-weight 12px pane title's x-height centre is another 0.5px below that centre. I chose optical alignment to the adjacent x-height rather than box-only alignment: moving and static marks receive the same 1.40625px downward correction. The render test measures the header baseline/x-height and pins their centres within 0.2px.

Measured geometry

With bundled GeistMono.woff2 at 12.5px in Chromium, rasterised at 16× and measured at alpha ≥ 16:

• six-dot ⠿ ink: 5.50 × 8.8125px
• one dot: 1.875 × 1.875px
• natural ink origin: 1.8125px left, 1.25px top
• optically corrected rendered origin: 1.8125px left, 2.65625px top

Chromium quantises CSS border widths to whole CSS pixels: 1.875px paints as 1px, so the path uses the nearest paintable stroke, 2px. If the bundled font fails and a fallback has different braille metrics, the animated glyph can drift relative to the path; the path deliberately remains tied to the shipped font's measurement.

The base .status rule remains untouched. Usage provider dots and Settings ready dots remain ordinary 8 × 8 caller-coloured dots. No #84/header component code or Sidebar markup is changed here.

Verification

• npm test — green.
• npm run test:render — green. It embeds the bundled font, re-measures the ink, reproduces normal/700-weight/600-weight contexts, proves all marks compute to weight 400, checks shared working/static geometry, measures pane-title x-height alignment, exercises the reader/Overview spinner variants, and pins the bare-dot regressions.
• npm run build — green.
• Built app exercised in light and dark themes with working/waiting/idle/stopped on real sidebar rows, real Overview tiles/list cards, and a local pane-title fixture reproducing Trace download on the session, and the reader's facts behind one i #84's mono 600-weight centred container.
• Normal animation remains ov-spin in the before and after builds; the captures use reduced motion only to hold one comparable frame.

Before/after parity sheet — both themes, all three surfaces in each frame →

[lvwerra]

Updated the static states per the dev Space feedback.

• waiting, idle, and stopped now share one square, hollow rectangle—the completed six-dot path. They differ only by accent / muted / dim treatment; the working braille animation is unchanged.
• The bundled Geist Mono measurement at 12.5px is 5.50 × 8.8125px for ⠿, with a 1.875px dot. Chromium floors a fractional CSS border to 1px, so the implementation uses the nearest paintable stroke, 2px, while keeping the measured box and origin unchanged.
• The state geometry remains single-sourced beside ov-spin; fallback-font drift is documented. Bare Usage/Settings dots are untouched. No Sidebar markup changed.
• npm test, npm run test:render, and npm run build are green. The built app was checked in both themes on sidebar rows, Overview tiles/list cards, and the two bare-dot surfaces.

Before/after sheet (republished at the existing URL): https://lvwerra-agent-artifacts.static.hf.space/pr87-braille-status-marks.html

[lvwerra]

Updated #87 for the pane-header parity feedback at f5f82ed.

• Confirmed the cause against Trace download on the session, and the reader's facts behind one i #84’s typography: moving .status under .ph-title changes its computed weight from 400 to 600. At 12.5px, bundled Geist Mono’s ⠋ grows from 5.50 × 5.375px / 10.04px² alpha area to 6.00 × 5.875px / 16.34px².
• The mark now owns Geist Mono, 12.5px, weight 400, and line-height 1 in every context. The reader, Overview progress/group spinners, and state mark consume the same type/geometry contract. No Trace download on the session, and the reader's facts behind one i #84 component code was touched.
• I aligned the ink—not merely the outer cell—to the pane title’s measured x-height centre. Moving and static states receive the same 1.40625px downward optical correction; the render test pins the centre delta within 0.2px.
• Full unit suite, expanded Chromium render suite, and production build pass. Both themes and all states were exercised across real sidebar rows, real Overview tiles/list cards, and a local weight-600 pane-title fixture.

Republished comparison sheet; each frame shows all three surfaces together: https://lvwerra-agent-artifacts.static.hf.space/pr87-braille-status-marks.html