Background
While working on #2084 (Cypress/Behat coverage for the "remember my choice" epic), the pre-existing it.only(...) in tests/e2e/cypress/integration/skeune/wayf/wayf.general.spec.js was removed to allow other tests in the file to run again.
Running the full file exposed 20 failures (not the ~12 previously assumed). #2084 fixes the subset that is well-understood and cheap:
- A test missing its own
cy.visit() (breaks under Cypress's default testIsolation: true, which resets to a blank page between tests).
- Raw connected/disconnected IdP count assertions that don't account for
WayfController's addDiscoveries query param defaulting to true (which injects 2 fixed synthetic "discovery" IdPs into every render).
The remaining failures are left re-.skip()'d (with a comment pointing at this issue) because they look like a distinct, deeper problem needing dedicated investigation:
- Visibility assertions tied to the
wayf__idpList--cutoffMet CSS class (Should show no connected IdPs when cutoff point is configured, Should show no disconnected IdPs when the flag is false) — the element has the expected class but isn't actually hidden, suggesting either the CSS rule is missing/stale in the built theme assets for this test environment, or the class/behaviour contract has drifted.
search__submit / search__reset visually-hidden class toggling not happening as expected when typing in the search field (Should show no results when no IdPs are found, Should be able to search for an idp).data-weight attribute assertions (6 tests) — findWeight.js / checkFullMatch.js / checkPartialMatch.js scoring constants (215/82/100/60/8/7) don't match any currently-rendered element; could be legitimate algorithm drift since these tests were written, unrelated to the discovery-idp noise..remainingIdps__defaultIdp banner not found / not toggling correctly in several tests (Check if the defaultIdp is present, Should reset the search text..., Should hide/show the IdP link when search term is provided).
No console/JS errors were observed during a quick browser check, which weakens the "broken build" theory somewhat, but this needs a proper investigation with fresh eyes — likely tracing whether the built CSS/JS actually matches current source for the container used to run these tests, and re-verifying the weight-scoring constants against current fixture data.
Task
Investigate and fix (or intentionally rewrite) the remaining suppressed test cases in wayf.general.spec.js, removing the .skip() added in #2084 once each is understood and passing for the right reason.
Background
While working on #2084 (Cypress/Behat coverage for the "remember my choice" epic), the pre-existing
it.only(...)intests/e2e/cypress/integration/skeune/wayf/wayf.general.spec.jswas removed to allow other tests in the file to run again.Running the full file exposed 20 failures (not the ~12 previously assumed). #2084 fixes the subset that is well-understood and cheap:
cy.visit()(breaks under Cypress's defaulttestIsolation: true, which resets to a blank page between tests).WayfController'saddDiscoveriesquery param defaulting totrue(which injects 2 fixed synthetic "discovery" IdPs into every render).The remaining failures are left re-
.skip()'d (with a comment pointing at this issue) because they look like a distinct, deeper problem needing dedicated investigation:wayf__idpList--cutoffMetCSS class (Should show no connected IdPs when cutoff point is configured,Should show no disconnected IdPs when the flag is false) — the element has the expected class but isn't actually hidden, suggesting either the CSS rule is missing/stale in the built theme assets for this test environment, or the class/behaviour contract has drifted.search__submit/search__resetvisually-hidden class toggling not happening as expected when typing in the search field (Should show no results when no IdPs are found,Should be able to search for an idp).data-weightattribute assertions (6 tests) —findWeight.js/checkFullMatch.js/checkPartialMatch.jsscoring constants (215/82/100/60/8/7) don't match any currently-rendered element; could be legitimate algorithm drift since these tests were written, unrelated to the discovery-idp noise..remainingIdps__defaultIdpbanner not found / not toggling correctly in several tests (Check if the defaultIdp is present,Should reset the search text...,Should hide/show the IdP link when search term is provided).No console/JS errors were observed during a quick browser check, which weakens the "broken build" theory somewhat, but this needs a proper investigation with fresh eyes — likely tracing whether the built CSS/JS actually matches current source for the container used to run these tests, and re-verifying the weight-scoring constants against current fixture data.
Task
Investigate and fix (or intentionally rewrite) the remaining suppressed test cases in
wayf.general.spec.js, removing the.skip()added in #2084 once each is understood and passing for the right reason.