Motir

Type: bug · Parent: Epic 3 · Surfaces: swimlane board (Subtask 3.3.5 — SwimlaneBoard.tsx) · Status: open · Reported by: Yue · Sibling of: bug-swimlane-lane-header-not-spanning-scrolled-columns (this is the SECOND occurrence of the same shape — collapsed-state was not covered by the prior fix).

On the /boards swimlane view (group-by Assignee / Priority / Epic), when a lane is collapsed (chevron pointing right, body hidden) the lane-header band does NOT span the full column track — it shrinks to roughly the width of the inline header content (chevron + label + count badge), leaving the rightmost columns visually outside the band. The same lane in its expanded state paints the band across the full track correctly (that case was fixed by the prior bug). Asymmetry between expanded and collapsed visual shape is the giveaway.

Repro: sign in as zhuyue@prodect.co / !QAZ1qaz, open the moooon / prodect project → /boards, switch group-by to anything other than none (Epic recommended — matches the original repro screenshot). Click a lane header to collapse it. Observe that the collapsed lane's gray band stops well short of the rightmost column (in the default workflow, the band visibly ends around In Review or earlier — Done and Cancelled sit OUTSIDE the band). Expand the same lane and the band paints full-width as intended.

Root cause (hypothesis to verify in the fix Subtask). In app/(authed)/boards/_components/SwimlaneBoard.tsx, the prior fix landed min-w-max on the lane-row wrapper (the outer <div data-testid={swimlane-${lane.key}}>), so the wrapper grows to its max-content width. That max-content is established by the child cell-row (${track} flex row of w-72 LaneCells over columns.map(...)) — but the cell-row only renders inside the !isCollapsed ternary branch. Collapsing removes the width-establishing child, so the wrapper's max-content drops to the header-only inline width (chevron + label + count), and min-w-max resolves against that smaller value. Net: the band paints behind a narrower wrapper, columns sit outside it. The prior fix's comment at SwimlaneBoard.tsx:122-126 named this exact risk for the EXPANDED case ("the lane-header band collapses to the scroller's clientWidth and stops at the viewport edge") but did not notice that collapsing pulls out the width source.

Fix shapes (decide at fix time — NOT prescriptive; trade-offs to weigh):

1. Render an invisible width-establishing track row OUTSIDE the !isCollapsed ternary — e.g. a zero-height ${track} h-0 pointer-events-none invisible row of w-72 shrink-0 spacer cells. Keeps the wrapper's max-content at full track width regardless of collapsed state. Lowest disruption to the existing sticky-stacking + click semantics; cost is one extra DOM row per lane.
2. Move min-w-max off the wrapper and onto an explicit-width band element, driven from the columns array (width: columns.length * 288 + (columns.length - 1) * 14 + 48 — the same shape the track class encodes). Removes the dependency on a width-establishing child entirely; cost is keeping the explicit-width formula in sync with the track class's w-72 gap-3.5 px-6.
3. Always render the cell-row track, hide it with visibility: hidden h-0 when collapsed — simplest diff. (NOT display: none, which would remove the element from layout entirely and collapse back to the original bug.) Functionally equivalent to option 1. Each option leaves the sticky-pinned label content (sticky left-6 z-[2]) untouched — only the band's width source changes.

Test gap that let it ship. The existing swimlane render tests (tests/components/board-swimlanes-render.test.tsx, board-swimlanes.test.ts) likely assert the band-spans-track shape only in the default (expanded) state. The fix MUST add a collapsed-state assertion: toggle collapse on a lane, measure the header band's offsetWidth (or computed style), and assert it equals the column-track's intrinsic width — not the viewport width, not the header-content width. Same measurement posture as the bug-tree-header-misalignment and prior bug-swimlane-lane-header-not-spanning-scrolled-columns fixes (measure rendered width via getBoundingClientRect, not CSS rules).

Acceptance criteria

• On a swimlane-grouped board, the lane-header band paints across the full column track when the lane is collapsed (chevron right) — including columns beyond the viewport edge.
• The same lane, expanded (chevron down), continues to paint the band full-width (guard against regression of the prior fix).
• Collapsing / expanding a lane does NOT change the band's rendered width — only the cell-row visibility.
• The sticky-pinned label + chevron + count behaviour is unchanged (label still sticks to the left edge as the user scrolls horizontally).
• The catch-all lane (bg-(--el-muted)) and named lanes (bg-(--el-surface-soft)) both render the band full-width in BOTH collapsed and expanded states.
• A render-test regression in tests/components/board-swimlanes-render.test.tsx asserts the collapsed-lane band width equals the column-track width (measure offsetWidth/getBoundingClientRect, not CSS rules) — mirrors the prior fix's test posture.

Context refs

• app/(authed)/boards/_components/SwimlaneBoard.tsx lines ~119-188 — the lane-row wrapper, the min-w-max comment naming the prior bug, the header <div role="button">, and the !isCollapsed ternary that hides the width-establishing cell-row
• bug-swimlane-lane-header-not-spanning-scrolled-columns (sibling Epic-3 bug, same shape, expanded-state-only) — the precedent fix this one extends
• PRODECT_FINDINGS.md #61 — the planner-side finding entry that surfaced this bug
• tests/components/board-swimlanes-render.test.tsx — where the missing collapsed-state assertion belongs
• motir-core/CLAUDE.md — colour via --el-*, shape via element-shape tokens (applies to whatever new wrapper the fix introduces)

Refactor signal (rule of three): this is the SECOND occurrence of min-w-max open-coded on a swimlane row producing a width-establishment bug. If a third occurrence surfaces, extract a shared TrackRow (or useTrackWidth) primitive rather than open-coding min-w-max per row. Not yet — fix this one in place, but note the pattern for the next time.