Skip to content

test(e2e): revive the three shell specs disabled since #392 - #1414

Merged
rubenvdlinde merged 1 commit into
developmentfrom
test/revive-shell-specs-392
Aug 24, 2026
Merged

rubenvdlinde merged 1 commit into
developmentfrom
test/revive-shell-specs-392

Conversation

@rubenvdlinde

Copy link
Copy Markdown
Contributor

Closes the disabled half of #392.

What was dark

dashboard.spec.ts, navigation.spec.ts and pages.spec.ts were each
test.describe.skip(...) with // TODO(#392): rewrite for manifest-driven app shell on line 3-5. Three lines, 38 tests, none of which had ever run.

Meanwhile the rest of this suite grew to 284 executing tests against the very
shell these files were disabled for — so #392's premise ("Vue app shell not
mounting in CI") has not been true for some time. The files stayed dark because
nothing reads a Playwright report to notice.

Why un-skipping was not enough

Every assertion described the pre-manifest app:

old assertion reality
heading "Dashboard", level 2 the h2 is "Sales overview"
heading "Open Leads", level 4 widgets are role=group named by manifest widget id
href="/apps/pipelinq/clients" the shell routes on the hash; a path deep-link resets the SPA to the Dashboard, so those gotos would have tested the dashboard over and over
getByRole('radio', {name:'Cards'}) it is a button now
Add Item it is Add Client

And 7 of the 31 tests in pages.spec.ts described pages that no longer
exist
: unify-ticket-supertype folded Requests, Complaints and
Contactmomenten into one ticket index behind a quick-filter strip. Verified
live — #/requests, #/complaints, #/contactmomenten, #/surveys render no
index page. Deleted rather than rewritten; their surface is the Tickets index
plus clickQuickFilter(), which the workflow specs already drive.

Locale

The old files asserted English labels. This instance renders nl: Products
reads "Producten", the view toggles read "Kaarten"/"Tabel", and the dashboard
refresh control's only accessible name is "Dashboard vernieuwen". A spec keyed
on wording passes or fails on the runner's locale rather than on what it claims
to test.

So every assertion goes through the manifest page id
(data-testid="cn-nav-entry-<pageId>") or the widget id — untranslated by
construction. New Lead / New Request / New Client are asserted by name
deliberately: they come from the manifest, not the i18n catalogue, and render in
English even here.

Adds revealNavEntryByTestId() beside the existing label-based
revealNavEntry(), since the label variant is the locale-dependent one.

Measured, not assumed

Every route was read off the running app: 37 nav entries with their hrefs,
11 visible at load, and the page contract route by route. That last
measurement changed the design — #/services, #/resources and #/bookings
are index pages with no primary CTA, so a uniform "every index page has a
CTA" rule would have been wrong and would have earned this suite a second
disabling.

No exact totals

The required entries and widgets are subsets. A manifest that grows must not
turn this file red — only one that drops a page. Exact counts are how decidiq's
registry spec came to skip itself: it wants 29 providers, finds 27, and stands
down instead of reporting the difference.

Result

26 executing tests replace 38 that never ran.

check result
playwright test --list all three files parse; 26 tests in the default project
prettier --check clean
local execution not run — the machine's dev container went down mid-session, and this repo's base-url.ts exists precisely to stop suites being pointed at a shared instance. CI is the executing check here.

Assertion-level evidence came from live inspection of the running app rather
than from reading the source, which is what the old file did and got wrong.

dashboard.spec.ts, navigation.spec.ts and pages.spec.ts were each
`test.describe.skip(...)` with `// TODO(#392): rewrite for manifest-driven app
shell` on line 3-5. Three lines, 38 tests, none of which had run in any
pipeline. #392 is still open, and meanwhile the rest of the suite grew to 284
executing tests against the very shell these files were disabled for — so the
premise that the shell does not mount in CI has not been true for a while.

They could not simply be un-skipped. Every assertion in them described the
pre-manifest app:

  - `heading "Dashboard", level 2`            the page's h2 is "Sales overview"
  - `heading "Open Leads", level 4`           widgets are role=group named by
                                              their manifest widget id
  - `href="/apps/pipelinq/clients"`           the shell routes on the HASH, and
                                              a path deep-link RESETS the SPA to
                                              the Dashboard, so those gotos
                                              would have tested the dashboard
                                              over and over
  - `getByRole('radio', {name:'Cards'})`      it is a button now
  - `Add Item`                                it is `Add Client`

Seven of the 31 tests in pages.spec.ts described pages that no longer exist at
all: `unify-ticket-supertype` folded Requests, Complaints and Contactmomenten
into ONE ticket index behind a quick-filter strip. Verified live — `#/requests`,
`#/complaints`, `#/contactmomenten` and `#/surveys` render no index page. Those
tests were deleted rather than rewritten; their surface is the Tickets index
plus clickQuickFilter(), which the workflow specs already drive.

LOCALE. The old files asserted English labels. This instance renders `nl`,
where Products reads "Producten", the view toggles read "Kaarten"/"Tabel", and
the dashboard refresh control's only accessible name is "Dashboard vernieuwen".
A spec keyed on wording passes or fails on the runner's locale rather than on
the thing it claims to test. So every assertion here goes through the manifest
page id — `data-testid="cn-nav-entry-<pageId>"` — or the widget id, both
untranslated by construction. `New Lead`/`New Request`/`New Client` are asserted
by name deliberately: they come from the manifest, not the i18n catalogue, and
render in English even on this Dutch instance.

Adds revealNavEntryByTestId() beside the existing label-based revealNavEntry(),
since the label variant is the locale-dependent one.

MEASURED, not assumed. Every route was read off the running app: 37 nav entries
with their hrefs, 11 visible at load, and the page contract route by route. That
last measurement changed the design — `#/services`, `#/resources` and
`#/bookings` are index pages with NO primary CTA, so a uniform "every index page
has a CTA" rule would have been wrong and would have earned this suite a second
disabling.

Nothing here asserts an exact total. The required entries and widgets are
SUBSETS: a manifest that grows must not turn this file red, only one that drops
a page. Exact counts are how decidiq's registry spec came to skip itself — it
wants 29 providers, finds 27, and stands down instead of reporting the
difference.

26 executing tests replace 38 that never ran. `playwright test --list` parses
all three; prettier is clean.
@github-actions

Copy link
Copy Markdown
Contributor

Quality Report — ConductionNL/pipelinq @ c4feec6

Check PHP Vue Security License Tests
lint ✅
phpcs ✅
phpmd ✅
psalm ✅
phpstan ✅
phpmetrics ✅
eslint ✅
stylelint ✅
build ✅
check-manifest ✅
check-vue-demi ✅
test-l10n ✅
format ✅
check-schema-l10n ✅
composer ✅ ✅ 106/106
npm ✅ ✅ 635/635
app:check-code ⏭️
info.xml ✅
REUSE ❌
PHPUnit ✅
Newman ✅
Playwright ✅
Hydra gates ✅

Spec coverage: 68% — 212 of 313 enforceable scenarios carry an @e2e reference (threshold 20%, 2362 excluded with a reason).


Quality workflow — 2026-08-24 08:20 UTC

Download the full PDF report from the workflow artifacts.

@rubenvdlinde
rubenvdlinde merged commit 9efea43 into development Aug 24, 2026
45 checks passed
@rubenvdlinde
rubenvdlinde deleted the test/revive-shell-specs-392 branch August 24, 2026 08:27
Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment

Labels

None yet

Projects

None yet

Development

Successfully merging this pull request may close these issues.

1 participant