Pre-existing E2E flakes surfaced by Dependabot PR review (photoAnnotation, budget-source-move mobile, deposit menu clipping)

[steilerDev]

Context

While reviewing Dependabot PRs #1527 and #1511, multiple E2E shards failed in ways that the product-architect verified are NOT caused by the dependency bumps — they are pre-existing app/test bugs that surface intermittently on `beta` builds.

Failures observed

PR #1527 (post-bump baseline)

• `e2e/tests/budget/budget-source-move.spec.ts:563` (`@responsive` mobile) — "selecting a line with hasClaimedInvoice shows a warning block and disables confirm until 'I understand' is checked." Confirm button stays disabled after the "I understand" checkbox is checked (14 retries, mobile-only).

PR #1511 (post-bump baseline)

• `e2e/tests/invoices/invoice-deposits-ux.spec.ts` — Deposit OverflowMenu portal clipping: `Menu bottom edge must be within viewport height: Expected <=1080, Received 1152`. UI/CSS bug.
• `e2e/tests/invoices/invoice-budget-line-create-and-link.spec.ts` — Times out on German-locale "+ Add Budget Line" button; likely translation/locale-leak across tests.
• `e2e/tests/photos/photoAnnotation.spec.ts` — Arrow/Line tools: `line[data-shapeid]` / `polyline[data-shapeid]` element never appears (pre-existing canvas/SVG render bug, both desktop & mobile).
• `e2e/tests/dashboard.spec.ts` — "Card re-enable" predicate timeout.

Action

Triage and fix or skip-with-issue-link the above. These were not blocking the Dependabot merges (only `Quality Gates` is required on `beta`), but they will block the next `beta → main` promotion PR because `E2E Gates` IS required on `main`.

Evidence

• Architect review on PR chore(deps): bump the prod-dependencies group across 1 directory with 5 updates #1527 (run: 26124384591)
• Architect review on PR chore(deps-dev): bump the dev-dependencies group with 9 updates #1511 (run: 26069561845)

🤖 Filed by `/dependabot`

[steilerDev]

[orchestrator] Triage breakdown of the bundled failures (status as of this comment):

• budget-source-move mobile (budget-source-move.spec.ts:563, @Responsive) — confirm button stays disabled after "I understand". Still actionable here — likely a real mobile code/CSS bug (touch target / event).
• deposit OverflowMenu portal clipping (invoice-deposits-ux.spec.ts) — menu bottom exceeds viewport. Still actionable here — CSS/portal positioning bug.
• dashboard "Card re-enable" predicate timeout (dashboard.spec.ts) — still actionable here (investigate flaky predicate vs real bug).
• invoice-budget-line-create-and-link German-locale "+ Add Budget Line" timeout — now understood as the Create Budget Line not linking regression → tracked as BUG: "Create Budget Line" in invoice picker does not link new line to invoice (eagerLinkInvoice regression) #1611 (Critical). Handle there, not here.
• photoAnnotation Arrow/Line *[data-shapeid] never appears — this is the Konva-<canvas> DOM-inspection limitation; those scenarios are test.fixme. Tracked as Re-enable (or formally retire) the Konva-canvas photo-annotation E2E test.fixme scenarios #1616. Not actionable in this issue.

So the remaining scope of #1575 is: budget-source-move mobile, deposit menu clipping, and dashboard card re-enable. The other two are covered by #1611 and #1616.

[steilerDev]

Consolidated pre-existing E2E flake inventory (from WP-0 PR janitorial, 2026-05-30)

Per orchestrator decision, the print/dark-mode flake surfaced during PR #1623's CI review — plus the broader inventory surfaced during WP-4's PR triage — belongs here rather than in separate issues. When picking up #1575, include all of these in the fix sweep:

1. e2e/tests/budget-overview-print.spec.ts — "Dark mode: print resets CSS variables to light values"

Source: PR #1623 (diary noValidate) CI runs.
Diagnosis (from e2e-test-engineer): Chromium's CSS custom property resolution under page.emulateMedia({ media: 'print' }) is non-deterministic across parallel shard workers. The test reads --color-bg-primary via getComputedStyle on documentElement immediately after emulating print media; the resolved value depends on whether Chromium has finished re-applying the print-only CSS rules by the time getComputedStyle runs.
Pre-existing: confirmed failing on beta runs 26113360964, 26124163930, 26128428796 (2026-05-19) — well before any of the recent PR work.
Fix options: (a) add retries: 2 at the test level; (b) relax the assertion to tolerate Chromium's timing (e.g. expect.poll with a short retry budget); (c) await a frame after emulateMedia before reading computed style.
Impact: hits up to 5 of 16 shards intermittently.

2. e2e/tests/budget-source-filter.spec.ts — debounce coalescing flake

Source: WP-4 (Dependabot review) inventory.
Diagnosis brief: filter-input debounce events occasionally coalesce in a way the test doesn't tolerate.

3. e2e/tests/i18n/i18n.spec.ts — DE heading render after language switch

Source: WP-4 inventory.
Diagnosis brief: heading text in German doesn't always render in time after the i18n language switch, causing assertion timeout.

4. e2e/tests/invoice-auto-itemize-page.spec.ts — SuggestionBadge

Source: WP-4 inventory.
Diagnosis brief: SuggestionBadge visibility / animation timing.

5. e2e/tests/navigation/dashboard.spec.ts — Customize button after dismiss

Source: WP-4 inventory.
Diagnosis brief: Customize button re-appearance after a dismissal isn't reliably detected by the test.

6. e2e/tests/invoice-linked-budget-line-edit.spec.ts — mobile WI edit

Source: WP-4 inventory.
Cross-reference: also relevant to WP-2's work on #1626 — the WI (work item) edit flow on mobile viewports has overlap with that fix. Coordinate with WP-2 before rewriting if #1626 lands first.

Three originally-scoped flakes (per #1575's existing description)

The three flakes already named in #1575 (photoAnnotation, budget-source-move mobile, deposit menu clipping) remain in scope. Address all of the above as one sweep.

Suggested process

1. Reproduce each flake locally with sharded Playwright (npx playwright test --shard=N/M --repeat-each=10) to confirm the failure mode.
2. Pick the lowest-friction fix per case (test-level retries, locator strengthening, async polling, or production code adjustment when the flake reveals a real race).
3. Land fixes incrementally — each can be a separate PR or grouped — but reference this issue from each.

cc orchestrator session 8cdee825-a788-4c51-8019-5b78364e9795 (WP-0).

[github-actions]

🎉 This issue has been resolved in version 2.7.0-beta.46 🎉

The release is available on GitHub release

Your semantic-release bot 📦🚀

[github-actions]

🎉 This issue has been resolved in version 2.7.0 🎉

The release is available on GitHub release

Your semantic-release bot 📦🚀