Skip to content

E2E: eight specs blocked by product gaps found when turning the Playwright gate on #719

Open

Description

@rubenvdlinde

Turning the shared E2E Tests (Playwright) job on (#709) took procest from cancelled at the 45-minute cap after 65 of 122 tests to a completing suite. Most failures were stale test mechanics and are fixed in that PR. These eight are not test bugs — each is a gap in the app or in the CI fixture data, measured on a GitHub runner on 2026-08-04. They are test.fixme referencing this issue.

1-2. The in-app Settings page renders none of its configuration content

pages.spec.ts › Settings page › has schema configuration fields / › has case type management section

/index.php/apps/procest/settings renders only the section chrome — H4:Settings, H2:Settings. Measured: .settings-form count 0, no Case Type Management heading, and no scrollable container, so this is not lazy-mount: the type:"settings" page's section-admin slot (AdminRootView) never renders its body. The same components DO render on the Nextcloud admin page /settings/admin/procest, which is why spec-coverage/admin-settings.spec.ts passes.

The scenario is spec'd as the in-app settings page (admin-settings/spec.md#in-app-settings-page-renders-configuration-sections), so the tests were deliberately NOT retargeted at the admin URL — that would silently change what the scenario proves.

3. Admin "add case type" form never surfaces its Save control

admin-settings.spec.ts › clicking add case type opens creation form — times out even with test.slow() (180s). The admin page mounts fourteen OpenRegister-backed sections and was measured at ~50s to first paint under php -S.

4. Case-email settings expose no "Test connection" control

case-email-integration.spec.ts › settings expose a Test connection control — the button exists in EmailSettings.vue:116 but does not render on the admin page even after all sections are scrolled in. Its sibling test on the same page passes, so the page itself loads.

5. Case list renders neither a table nor cards

deelzaak-support.spec.ts › the case list renders and may show an "N deelzaken" badge. Measured on /cases: table=0, [role=table]=0, .viewTable=0, [class*=card]=0 — the list body renders an empty state, so the assertion is data-dependent on seeded cases.

6. /subsidies falls back to the generic case index

workflow-operations.spec.ts › subsidies index renders the subsidy intake list shell. /index.php/apps/procest/subsidies renders buttons [Settings, Cards, Table, Add Case, Actions] — a generic case list with Add Case, not a subsidy-specific create control. (/subsidieregelingen correctly renders Add Subsidieregeling and its test passes.)

7-8. Case detail never displays the zaaknummer

cases-crud.spec.ts › opening the row shows the case detail with its values and deelzaak-case-email.spec.ts › CaseDetail page renders the case. A case seeded with assigned identifier 2026-0001 renders a detail page containing CASE, the title, Assignee, Case type, Confidentiality — and the identifier is absent from the page text entirely. It DOES render in the case LIST, so the list-view assertions pass.


Each fixme carries the specific reason inline. Removing a fixme should follow from fixing the gap, not from relaxing the assertion.

Activity

  1. rubenvdlinde commented on Aug 10, 2026

    @rubenvdlinde
    ContributorAuthor

    The 300s bootstrap stall is not attached to any one spec — measured on one SHA, two runs

    Working the E2E debt on development I measured the Test timeout of 300000ms exceeded failure that the FIXME(#719) comments sit next to. Posting the measurement here because it settles what the stall is, and rules out two explanations that are currently written down as if they were facts.

    The stall MOVES between spec files on an identical SHA

    SHA a6bfe8b9dd7e0ae2fef1ee7bf677151445cec01a, same seed, same runner image, two runs:

    Run Event Spec that hit the 300s stall Tally
    31388005994 push spec-coverage/admin-settings.spec.ts › case type list renders its management surface and add control 83 expected / 38 skipped / 4 unexpected
    31393761940 workflow_dispatch case-types-tabs.spec.ts › admin settings surface has an add-control for case types 83 expected / 38 skipped / 4 unexpected

    A different spec file stalls in each run while the other three unexpected results (the PDOK trio) reproduce byte-identically. So the stall is not a defect in admin-settings.spec.ts; it lands on whichever spec is running while the shared app bootstrap is slow. Any fix must therefore be verified by the stall no longer moving, not by one spec going green.

    (The PDOK trio was a real product defect and is fixed separately — a service worker that swallowed every /apps/openconnector/api/pdok/* call. It is unrelated to this stall.)

    Two things in the current explanation are wrong

    1. There are no "~14 registerObjectType() round-trips" to batch. src/store/store.js#initializeStores() makes exactly one network call — await settingsStore.fetchSettings(). All 23 objectStore.registerObjectType(...) calls after it are synchronous and in-memory (they seed collections[type] = []; see the note in src/services/formatters.js:80). Batching or parallelising them would change nothing, so please don't spend a sprint there.

    2. Raising the timeout is not a fix and is off the table. A 300s budget that is sometimes met in 4.2s is not a budget problem.

    The candidate that is worth measuring next (named, not asserted)

    useSettingsStore().fetchSettings() (src/store/modules/settings.js:23) has no in-flight dedupe and no early return. It sets this.initialized = true but nothing ever reads it as a guard, and initializeStores() calls it unconditionally. initializeStores() is awaited on mount from 19 components (App.vue, MyWorkCards, DeelzaakList, DeelzaakDetail and the dashboard widgets MyTasksWidget / TaskRemindersWidget / DeadlineAlertsWidget / OverdueCasesWidget, …). Every one of those that mounts on a page issues its own GET /apps/procest/api/settings, and Nextcloud holds a per-session lock, so N identical requests are strictly serial, not concurrent.

    I did not confirm this is the cause and I am not claiming it is: on a fresh unseeded Nextcloud 32 I counted the requests of a dashboard load and saw 0 duplicate /api/settings calls out of 86 requests — the widgets that call it did not mount, because there is no data. The measurement that would settle it is the same request counter run against the surfaces that actually stall (admin-settings / case-types-tabs) on a seeded instance, counting GET /apps/procest/api/settings.

    If it does reproduce, the fix is pure de-duplication and needs no threshold change:

    • memoise initializeStores() behind a single shared promise, and
    • early-return from fetchSettings() when this.initialized is already true (the flag exists, it is simply never read).

    Also worth recording while this issue is open

    The same run reports 38 skipped of 125. A large share of those are self-skips on the default CI state — test.skip(true, 'No cases in the deployed/seeded register …') and '… not present in the deployed build (deploy mismatch)' in related-case-linking, deelzaak-support, workflow-editor-canvas, kanban-board-keyboard-status-transition, handler-vervanging-waarneming, doorlooptijd-dashboard. tests/e2e/ci-seed.sh provisions the register and schemas but seeds no case objects, so those guards are true on every CI run while the @e2e anchors keep claiming the scenarios. A skip that can never not happen reads exactly like a pass on the report.

Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment

Metadata

Metadata

Assignees

No one assigned

    Labels

    No labels
    No labels

    Type

    No type

    Projects

    No projects

      Milestone

      No milestone

      Relationships

      None yet

      Development

      No branches or pull requests

      Issue actions