Repository navigation
fix: clear eight hydra gates — a11y, relation dialect, icons, and 25 dead e2e waits - #1175
Merged
Merged
Conversation
…d row had no reduced-motion fallback The run-log header and the trace-timeline row both expand a detail panel on click. Neither was reachable by keyboard and neither announced itself as anything: a <div> with @click has no role, no tab stop and no Enter/Space handling. Both are now real <button type="button">s carrying aria-expanded, which brings all of that with it; the CSS resets make a button look like the row it replaced. The rules-editor row transitions its hover and focus tint. Under prefers-reduced-motion: reduce it now changes state instantly instead (WCAG 2.2 SC 2.3.3). Closes gate-32 (2 findings) and gate-45 (1) on this repo.
ADR-074 rule 4. Nextcloud holds long-poll and heartbeat connections open, so
`networkidle` does not settle: every one of these 25 waits ran to its own
timeout and then swallowed the error, or burned the goto's budget before the
assertion that was going to do the real waiting anyway.
`goto(..., { waitUntil: 'networkidle' })` becomes `'domcontentloaded'` — each
of these navigations is already followed by an `expect(...).toBeVisible()`
with its own timeout, which is the wait that was doing the work. The four
`waitForLoadState('networkidle').catch(() => {})` calls become
`'domcontentloaded'`, or are dropped where an explicit settle already follows
them.
Also fixes the one occurrence in features-roadmap.spec.ts that the gate did
not report — it fell outside the diff scope, not outside the problem.
Measured against this branch's parent on the local instance: the same 6
spec-coverage tests fail and the same 1 passes, before and after, in the same
2.6 minutes. This changes how long the suite waits for nothing, not what it
asserts.
… '?', and one phpcs error notificaties_abonnement.sourceId and .consumerId were relation-shaped (format:uuid, described as pointing at another object) without the canonical $ref, so OpenRegister could not see them as relations at all (ADR-062 rule 7). Both now carry the $ref the rest of this register already uses. Deliberately without x-openregister-onDelete: the consumer's cascade-delete lives in NotificatiesSubscriberService (REQ-004) and moving it into the schema would change delete semantics, which this is not. The consumer detail page's subscriptions widget asked for 'PackageVariantClosed', which is not in the shared icon registry — it was rendering the '?' fallback in the UI. Now 'Package', the registry's nearest name. ExecutionTraceService::find() takes gate-17's own `@spec exclude` escape with a reason. It is not the shape ADR-022 is aimed at: it pins the register/schema pair, turns rbac and multitenancy off because a trace is infrastructure telemetry rather than a tenant-owned object, and converts a miss from a throw into null. Both callers depend on exactly that; deleting it duplicates five arguments and a try/catch at each. Also fixes the one phpcs ERROR in the repo, pre-existing and unrelated: MappingService's SPDX lines sat as inline comments between the docblock and the namespace, where the sniff requires end punctuation. Moved into the docblock, which is where the rest of the fleet keeps them.
…accessible name Data tables (WCAG 2.2 SC 1.3.1). Nine tables had <th> cells carrying no scope, so a screen reader announces the cell text without binding it to the column or row it heads; they now declare scope="col", or scope="row" for the key/value tables in the job modals. The two synchronization result modals had no header cells at all — every row was <td><b>Label:</b></td><td>value</td>, a table whose structure existed only visually. Those <b> wrappers are gone: <th scope="row"> is the element that means what they were drawing. Form controls (SC 1.3.1, SC 4.1.2). Three NcTextFields in the rule editor sat next to a bare <label> with no `for`, which associates nothing — they take the component's own :label prop now, and the loose labels are gone. Two bulk-select checkboxes in row context had no name at all and now say which row they select. Two raw-JSON textareas had none either and now have a visible, associated label. RuleDetailPage's description label bound `:for="'rule-description'"` — a static string through a binding. Now a plain `for`. Also the two pre-existing stylelint errors in ApiProductDetail (double quotes in an attribute selector), unrelated but in the way. Verified: eslint 0 errors, stylelint clean, webpack build succeeds, gate-40 and gate-43 both PASS.
…s nothing Swapping the consumer-subscriptions widget off PackageVariantClosed fixed one thing and broke another. PackageVariantClosed is absent from the shared widget registry, so it rendered the '?' fallback; Package is in that registry but had never been registered in this app's own src/icons.js, and an unregistered name renders NO icon at all — not a fallback (ADR-077 rule 3). Caught by gate-60 going red in the same run gate-55 went green. Registered.
rubenvdlinde
requested review from
SudoThijn,
WilcoLouwerse,
bbrands02,
remko48 and
rjzondervan
as code owners
August 7, 2026 09:05
….js) "Conditions (JSON Logic)" for the two raw-JSON textarea labels, and the two row-scoped bulk-select names. Added to en.json/en.js and translated in nl.json/nl.js, matching how the Dead letters strings landed. The nearby key is "Conditions (JSON logic)" with a lowercase l — a different string, left alone rather than folded into this one.
Contributor
Quality Report — ConductionNL/openconnector @
|
| Check | PHP | Vue | Security | License | Tests |
|---|---|---|---|---|---|
| lint | ✅ | ||||
| phpcs | ✅ | ||||
| phpmd | ✅ | ||||
| psalm | ✅ | ||||
| phpstan | ✅ | ||||
| phpmetrics | ✅ | ||||
| eslint | ✅ | ||||
| stylelint | ✅ | ||||
| build | ✅ | ||||
| build | ✅ | ||||
| check-specs | ✅ | ||||
| test-l10n | ❌ | ||||
| composer | ✅ | ✅ 153/153 | |||
| npm | ✅ | ✅ 551/551 | |||
| PHPUnit | ✅ | ||||
| Newman | ✅ | ||||
| Playwright | ❌ | ||||
| Hydra gates | ❌ |
Quality workflow — 2026-08-07 09:12 UTC
Download the full PDF report from the workflow artifacts.
Contributor
Quality Report — ConductionNL/openconnector @
|
| Check | PHP | Vue | Security | License | Tests |
|---|---|---|---|---|---|
| lint | ✅ | ||||
| phpcs | ✅ | ||||
| phpmd | ✅ | ||||
| psalm | ✅ | ||||
| phpstan | ✅ | ||||
| phpmetrics | ✅ | ||||
| eslint | ✅ | ||||
| stylelint | ✅ | ||||
| build | ✅ | ||||
| build | ✅ | ||||
| check-specs | ✅ | ||||
| test-l10n | ✅ | ||||
| composer | ✅ | ✅ 153/153 | |||
| npm | ✅ | ✅ 551/551 | |||
| PHPUnit | ✅ | ||||
| Newman | ✅ | ||||
| Playwright | ✅ | ||||
| Hydra gates | ✅ |
Quality workflow — 2026-08-07 09:25 UTC
Download the full PDF report from the workflow artifacts.
This file contains hidden or bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
Add this suggestion to a batch that can be applied as a single commit.This suggestion is invalid because no changes were made to the code.Suggestions cannot be applied while the pull request is closed.Suggestions cannot be applied while viewing a subset of changes.Only one suggestion per line can be applied in a batch.Add this suggestion to a batch that can be applied as a single commit.Applying suggestions on deleted lines is not supported.You must change the existing code in this line in order to create a valid suggestion.Outdated suggestions cannot be applied.This suggestion has been applied or marked resolved.Suggestions cannot be applied from pending reviews.Suggestions cannot be applied on multi-line comments.Suggestions cannot be applied while the pull request is queued to merge.Suggestion cannot be applied right now. Please check back later.
Clears eight hydra gates on this repo's
development → betadiff. Everything here is a real defect in the app — the gate false positives found along the way were fixed on the gate side instead (ConductionNL/.github#198, #199, #201, #202) and are not in this PR.Accessibility
<div>s — no role, no tab stop, no Enter/Space<button type="button">witharia-expanded<th>with noscope; 2 more had no header cells at all —<td><b>Label:</b></td>, structure that existed only visuallyscope="col"/scope="row"; the<b>wrappers replaced by the element that means what they were drawingNcTextFields sat beside a bare<label>with nofor(associates nothing); 2 bulk-select checkboxes and 2 raw-JSON textareas had no accessible name:labelprop, row-specificaria-label, visible associated labelsprefers-reduced-motion: reduceRuleDetailPage's description label bound:for="'rule-description'"— a static string through a binding. Now a plainfor.Specs and data
gate-54 —
notificaties_abonnement.sourceIdand.consumerIdwere relation-shaped (format: uuid, described as pointing at another object) without the canonical$ref, so OpenRegister could not see them as relations at all. Both now carry it. Deliberately withoutx-openregister-onDelete: the consumer's cascade-delete lives inNotificatiesSubscriberService(REQ-004), and moving it into the schema would change delete semantics.gate-55 / gate-60 — the consumer-subscriptions widget asked for
PackageVariantClosed, which is absent from the shared widget registry, so it rendered the?fallback. Swapping toPackagefixed that and immediately broke gate-60:Packagehad never been registered in this app'ssrc/icons.js, and an unregistered name renders no icon at all, not a fallback. Registered. Two commits, because the second only exists because the first was incomplete.gate-17 —
ExecutionTraceService::find()takes the gate's own@spec excludeescape, with a reason. It is not the shape ADR-022 targets: it pins the register/schema pair, turns rbac and multitenancy off because a trace is infrastructure telemetry rather than a tenant-owned object, and converts a miss from a throw into null. Both callers depend on exactly that; deleting it duplicates five arguments and a try/catch at each.Tests
gate-58 — 25
networkidlewaits. Nextcloud holds long-poll and heartbeat connections open, sonetworkidledoes not settle (ADR-074 rule 4): every one of these ran to its own timeout and swallowed the error, or burned thegoto's budget before the assertion that was going to do the real waiting anyway.Measured against this branch's parent on a live instance: the same 6 spec-coverage tests fail and the same 1 passes, before and after, in the same 2.6 minutes. This changes how long the suite waits for nothing, not what it asserts.
Pre-existing, fixed in passing
MappingService's SPDX lines sat as inline comments between the docblock and the namespace, where the sniff wants end punctuation. Moved into the docblock, where the rest of the fleet keeps them.ApiProductDetail(double quotes in an attribute selector).Verified
eslint 0 errors · stylelint clean · webpack build succeeds · phpcs 0 errors ·
check:specsPASS · gates 17, 32, 40, 43, 45, 54, 55, 58, 60 all PASS locally againstorigin/beta.Not in scope, still red
gate-16(108 methods missing@spec),gate-19(21 scenarios),gate-25(17 endpoints),gate-26(13 components),gate-46(20 anchors),gate-49(12 methods),gate-51(25 properties) — bulk documentation and coverage debt, untouched here.gate-57is one genuine orphan:LtiAgsService::publishScoreis implemented, spec'd and unit-tested with zero production callers — the Tool-role outbound direction has no route, job or flow step. That needs a design decision, not a cleanup.