timeline: a dot renders 98px past the track, and the test that catches it has not run in a while #263

e2e/timeline-geometry.spec.ts:158 — "every dot renders inside the track (v0.9.4 P0-3)" — fails:

Error: a dot rendered past the track width
Expected: <= 1800     (track width)
Received:    1898     (max dot offsetLeft)

98px outside the track it is supposed to be clipped to.

Why this surfaced now

The test is in a serial describe, and the test before it (P0-2, lane stack scrolls vertically instead of clipping) had been failing, so Playwright reported P0-3 as did not run. It has not been evaluated for as long as P0-2 has been broken.

P0-2 was fixed on refactor/onboarding-controls — it was measuring the sidebar rather than the lane stack, because document.querySelector('div.overflow-y-auto.min-h-0') started matching the nav once 3394e9a made it scrollable. With P0-2 green, P0-3 runs for the first time in a while, and fails.

So this failure was masked, not introduced by that fix. The fix to P0-2 only changed which element the test resets scrollTop on, which is vertical and cannot move offsetLeft.

Likely culprit

The fixture seeded by the test above it (seed all 18 lanes + an out-of-range marker, line 71) includes:

['marker', { title: 'far-future marker', severity: 'critical', atTimestamp: now + 30 * 60_000 }]

and the comment beside it records what this guards: "Before the fix the domain came from timestamp only". P0-3 exists to catch the time domain failing to account for a marker's display timestamp, which puts that marker off the right edge — which is exactly the shape of this failure.

Worth looking at first: buildTimeMap in src/renderer/src/lib/timelineTimeMap.ts and displayTs in src/renderer/src/lib/timelineDomain.ts.

Open question — is this a regression?

Not yet determined. refactor/onboarding-controls changed zoom in 8109e6e feat(ui): 90% is the interface, not a setting to find and 3394e9a fix(ui): one zoom, the sidebar scrolls, and the HUD row has one height. A zoom change moves both trackW and every dot's position, so it is a plausible cause — but the test has been unevaluated long enough that the defect could predate the branch entirely.

The cheap way to settle it: run this one spec on origin/main and compare. That was not done before filing.

Reproduce

npm run build
npx playwright test e2e/timeline-geometry.spec.ts --reporter=list

e2e runs against out/main/index.js, so a source change needs npm run build first.

Requires the P0-2 fix to be present, otherwise P0-3 is skipped and reports as "did not run" rather than as a failure.

Related

Separate from #262, which covers four specs whose premises the same branch invalidated. This one is a possible product defect rather than stale test coverage.