kandev/docs/plans/kanban-auto-hide-empty-columns/task-04-e2e-docs.md

id, title, status, wave, depends_on, plan, spec

id	title	status	wave	depends_on	plan	spec
04-e2e-docs	Prove end-to-end behavior and document it	done	4	03-drag-targets	plan.md	../../specs/ui/requirements/kanban-auto-hide-empty-columns.md

Task 04: Prove end-to-end behavior and document it

Acceptance

• Desktop E2E covers default-off compatibility, persisted enablement, empty-column collapse, search stability, drag reveal, cancellation, successful drop, and manual-hidden precedence.
• Mobile E2E covers focused-workflow isolation, a 44 CSS px toggle surface, move-target recovery, and absence of document overflow.
• Public tasks/workflows documentation and the feature-spec index explain the new preference and its distinction from manual hiding.

Likely files

• apps/web/e2e/tests/kanban/auto-hide-empty-columns.spec.ts
• apps/web/e2e/tests/kanban/mobile-auto-hide-empty-columns.spec.ts
• apps/web/e2e/pages/kanban-page.ts if a shared locator is justified
• docs/public/tasks-and-workflows.md
• docs/specs/INDEX.md
• this plan and its task files for final status and command evidence

Verification

(cd apps/web && pnpm e2e:run --host --project chromium tests/kanban/auto-hide-empty-columns.spec.ts)
(cd apps/web && pnpm e2e:run --host --project mobile-chrome tests/kanban/mobile-auto-hide-empty-columns.spec.ts)
make fmt
make typecheck
make test
make lint

Risks

• E2E must wait on causal task/settings updates rather than unconditional sleeps.
• The tests must prove manual and automatic hiding are separate, not merely that a column vanished.
• User-facing documentation must not imply that manually hidden steps become move targets.

Results

• Added auto-hide-empty-columns.spec.ts; Chromium passed 1/1 and exercises the complete desktop acceptance path through the real settings UI and pointer drag surface.
• Added mobile-auto-hide-empty-columns.spec.ts; Mobile Chrome passed 1/1 and verifies workflow isolation, settled 44 CSS px geometry, recovered drop targets, cancellation cleanup, and no document-level overflow.
• Added a stable mobile drop-target test id and the auto-hide field to the typed E2E settings helper.
• Both runs used the production E2E build, real backend, isolated SQLite database, and no raw sleeps.