Repository navigation
Release: merge development into beta - #3726
Open
github-actions[bot] wants to merge 883 commits into
Open
github-actions[bot] wants to merge 883 commits into
github-actions[bot] wants to merge 883 commits into
Conversation
rubenvdlinde
requested review from
Rem-Dam,
SudoThijn,
WilcoLouwerse,
bbrands02,
remko48,
rjzondervan and
rubenvdlinde
as code owners
October 11, 2026 07:21
A batch and a triage rule are private to the people the case type names. Two checks read them under the caller's own access and so saw a slice: a document could go into a second open batch, and applying the rules as a colleague of the assignee applied none and answered zero. Both now read every record of the case; listings stay under the caller's access.
The embedded table held 16 interventions that differ from the 16 shipped grid cells on every position. Decision 213: the grid cells are right. The embedded table now holds the grid's interventions, and LhsLookupAnswersTheGridTest reads the shipped files and asks all 16 positions from each source (stored table, register cells, embedded table), with a control that reports exactly one changed cell. Reapplies #3774, which train 3 reverted (c9cfb5d).
…ADR-118) No class in lib/ injected KvkHandelsregisterAdapterInterface. The four client classes, the registrar and the contract fixture go. The KvK card stays Not available and says dossiq ships no KvK client.
A form can now offer the named OpenRegister record templates of its schema. RecordTemplateReader reads one template as the signed-in caller, and RecordTemplateSelect lists them from /apps/openregister/api/record-templates. Both are generic: any schema, any case type. The search plan of a request is the first form that uses them. Until the plan is recorded the dialog offers the templates of the request configuration. A template fills whose files, the systems and the terms, never the period. The recorded configuration names the template in copiedFrom, and a plan that names a template the caller may not use is refused. woo-request-corpus-collection task 5.2.
Both sides added a method and a test at the same place: both kept. The predicate is renamed isAdmittedReader().
… openregister development openregister#4583 and #4592 landed in openregister train #4595, and dossiq#3761 in train 3. The tick for 5.2 records what was read on openregister development: the review annotation is kept and validated, the 403 decision-owned answer, and the two events dossiq listens to. 2.2 is ticked because ReviewBatches now calls the reconcile. The Newman collection the tasks name does not exist yet.
Characterisation for termijnbewaking-op-engine-timers 4.2, part 3, green on the code as it stands: the share of the statutory term that is gone (CaseTermsService::progressFor) and how long a request for more information has been open (AanvullingsverzoekService::daysOpen). The other five term calculations in this part were already pinned.
caseSearchDeclarations.spec.js asks every case property for an input control or a place on its silent list. Both fields are looked up by exact match.
A reviewerAccess.records entry carries assigneeMay and reviewersMay, each read (the default) or update. CaseRecordReaders::grantsFor() answers core's permission bits per reader, and ObjectReadGrants::reconcile() makes the record's user grants equal them: a grant with other permissions than its reader is due is revoked and granted again. Delete and share are never granted (REQ-CRA-010). Configured for the Woo case type: the assignee may update a review batch and a triage rule. Register 0.20.39, Woo case type object 1.4.5.
termijnbewaking-op-engine-timers 4.2, part 3 of 4 (REQ-TOT-007). Term consumed and days left (CaseTermsService), days of a pause used, days in a phase, the calendar overrun of a first response and the days a request has been open are counted through DayCount. No count changes. DeadlineExtensionService and TermijnTimerService stay on the allowlist: both sit at the PHPMD coupling limit and own their day math. Allowlist 19 to 14.
…ices and declares how its date is searched
…rs, before the class is split
CaseRecordReaders::grantsFor() tripped the ratchet against a second permission evaluator. It decides nothing about who may read: it answers what dossiq asks OpenRegister to grant. Renamed to toGrant().
documentReview carries the private block. The Woo case type lists it: the assignee reads and updates every review of the case, a batch reviewer the reviews of their own batch, supervisors all of them. Every rule that reads reviews now reads the whole case (DocumentRelevance::forCase, REQ-CRA-011), so what is outstanding, which pages are unseen and what a rule may mark do not depend on who asks, and a marking never makes a second review for a document. The relevance route, the pages-seen route (ReviewWrites) and the triage apply reconcile the case when they create a review. A reconcile that cannot run after a write is logged and answered as failed. Register 0.20.40, documentReview 1.0.3, Woo case type object 1.4.6. An administrator runs POST /api/cases/reviewer-access/reconcile once after the upgrade.
src/settings.js, the admin entry, never imported @conduction/nextcloud-vue/css/index.css (src/main.js does), so /settings/admin/dossiq drew every library component unstyled.
…-once fix(portal): a Woo request sent twice with the same submission key makes one case, and the dossier is linked after the answer
…ever scroll (#3937) DqWerkbord draws the columns side by side in a grid, a few cards each and a dashed "Toon 12 meer". The board was one row of every status, 16,116px wide, with a scroller per column and cards wider than their column. - Columns sit in a grid of at least 240px each and wrap to the next row. - A column draws three cards; the rest stay in its list behind Show N more, so drops, selection and the keyboard move still see every case. - No column scrolls on its own; the window scrolls the page. - Card headers wrap instead of running past the card. - The drag-to-pan helper went with the strip it panned.
…k from The closing move's resultTypeId input now carries a label (Result / Resultaat) and options, one per result type the engine already put on the offered move (the case type's own and its parents'), labelled by name. The stages widget's transition dialog renders that as a picker instead of a bare required text field asking for a UUID. Needs openregister to publish input label/options and nextcloud-vue to render options; on older releases the keys are dropped or ignored and the dialog shows the text field as before.
…result fix(lifecycle): a closing stage offers the case type's results to pick from
#3940) ICache types its parameters from Nextcloud 35 on and leaves them untyped in 32 to 34. InMemoryMemcache named `string` on get, set, hasKey, remove and clear, which PHP refuses on 32 to 34: an implementation may widen a parameter type, never narrow it. Loading the class was a fatal error, so every PHPUnit cell on development stopped before it ran a single test. The five parameters lose their type, which satisfies both shapes. ncad is typed identically in every supported version and keeps its types. The twelve missing docblocks phpcs reported on this file are filled in, since the file was open anyway.
…g team A case filed while its case type had no handling block was written without assignedGroup and kept none after the case type gained one. A repair step fills it from handling.defaultGroup on open cases that have none. Assisted-by: ClaudeCode:claude-opus-5-5
…mat gate (#3943) `npm run format` (prettier --check) is red on development: three vitest specs carry the same over-long `toMatch` assertion on one line. Prettier wraps each into a multi-line call with a trailing comma. No behaviour changes. Assisted-by: ClaudeCode:claude-opus-5
…an read the case type 2.84.0 ships schema 2.96.0 with `@context.<name>.<field>` and the detail page's `visibilityContext` (nextcloud-vue#1602), which the case panels on this branch use. Re-vendors tests/schemas to 2.96.0, and turns the captions tripwire in structureProfile.spec.js round: the library now keeps a caption through the relocation step unless the relocations emptied it.
…type feat(case-detail): the family plan and the projects panel show only on the case types they belong to
…ters (#3946) * refactor(termijn): a letter's language is an object, not a flag on every date phpmd flagged TermLetters::dayOf() and ::shownDay() for taking a boolean flag argument. The flag was the letter's language, threaded through every date and defaulting to Dutch, so each call site decided the language again and two of them decided it by leaving the argument out. LetterDays holds the language once and prints days in it. A caller names the language through a factory (forLocale, or dutch for the letters whose wording exists in Dutch only), so no boolean crosses a call site. termStartLines takes the locale the class already passes around instead of its own boolean. Wording and dates are unchanged. * refactor(termijn): the letter's language is a constructor argument, not a static factory The ruleset that flags the boolean flag also flags StaticAccess, and it exempts FleetAppId alone, so LetterDays::forLocale() traded one phpmd finding for another. A caller now names the language as the locale it already holds: `new LetterDays(language: $locale)`, or `'nl'` for the two letters whose wording exists in Dutch only.
) * test(ci): the coverage metadata names the classes these tests run phpunit.xml sets beStrictAboutCoverageMetadata and failOnRisky, so a test that executes a class its docblock does not name is risky, and a risky test fails the run. 49 tests across 15 files did that. They were invisible until #3940, because the PHPUnit cells fataled before running anything. Each class comes from the run's own report, so the list is exact rather than guessed. Six classes account for all of it, ReadableDay for 23. Annotations only: no test and no production code changes. * test(ci): the Woo intake test names the job and queue it runs
…e shipped snapshot (#3944) testTheCheckNamesAGroundThatDiffers overwrote the tracked lib/Settings/woo-refusal-grounds.snapshot.json and restored it in a finally, so it needed write access to a file the app ships. That failed wherever phpunit runs as a user other than the checkout's owner. In the dossiq dev container, which runs as www-data against a host-owned bind mount, file_put_contents() was refused, the check script read an untampered snapshot, exited 0, and the suite reported only "Failed asserting that 0 is identical to 1". That reads as snapshot drift and sends a reader to audit a snapshot that is correct. It also risked the shipped file: a finally does not survive SIGKILL, and filinq and opencatalogi vendor that snapshot as their redaction fallback. The tool resolves its paths from its own location, so the test now copies the three files it reads into a temporary directory and tampers there. Both assertions are unchanged. A control run of the untampered copy asserts exit 0 first, so a copy that cannot be read cannot read as drift, and the shipped snapshot is asserted byte-identical afterwards. Assisted-by: ClaudeCode:claude-opus-5
…case-type-team fix(repair): open cases without a team take their case type's handling team
# Conflicts: # tests/Unit/BackgroundJob/AcknowledgementDispatchJobServiceAccountTest.php # tests/Unit/Controller/PortalWooRequestSubmissionKeyTest.php # tests/Unit/Controller/WOOAssessmentControllerExtensionNoticeTest.php # tests/Unit/Service/AcknowledgementDutyTest.php # tests/Unit/Service/Intake/SystemCaseStoreTest.php # tests/Unit/Service/Termijn/TermLettersTest.php # tests/Unit/Service/TermijnNotificationServiceTest.php # tests/Unit/Settings/WorkflowGuardConformanceTest.php # tests/Unit/Woo/PortalFiledCaseOpensLikeAStaffCaseTest.php # tests/Unit/Woo/WooRequestIntakeReceiveTest.php # tests/Unit/Woo/WooRequestIntakeTest.php # tests/Unit/Woo/WooWritesMatchTheRealSchemasTest.php
…ng two advisories (#3948) `quality / Security (composer)` is red on development. `composer audit` reports two advisories, both in packages-dev: - symfony/http-foundation v6.4.43, CVE-2026-108082: ReDoS in HeaderUtils::split(), reachable through the Cache-Control request header. Patched from 6.4.46. - symfony/yaml v6.4.40, CVE-2026-109413: unbounded recursion in the YAML inline lexer lets a single document crash the process. Patched from 6.4.45. It arrives through consolidation/robo (symfony/yaml ^6). Both move to v6.4.48, the newest 6.x, so nothing crosses a major boundary. composer.json is unchanged: the `^6.4.41` constraint on http-foundation already allowed a patched release, and yaml is a transitive requirement, so the lock alone was behind. The lock's content-hash is unchanged and only these two packages move. Assisted-by: ClaudeCode:claude-opus-5
…ed here A newer version left every row the instance already held as it was, so an untouched row never received a corrected value. After the write and the ledger record, each held row now goes through the shipped-set ledger: an untouched row takes the new version and is stamped with it, a row edited here is kept and logged by id. The outcome reports refreshed and kept. Assisted-by: ClaudeCode:claude-opus-5-5
Seven suppressions landed on development without the baseline moving with them, so `npm run check:debt-ratchet` fails: eslint.inlineDisables 15 against 14, phpmd.suppressWarnings 187 against 182, psalm.inlineSuppressions 254 against 253. Three of the seven hide nothing and are removed. The inline `@psalm-suppress UnusedClass` on LinkCollectionToCaseJob duplicates psalm.xml's blanket handler for that issue class. PHPMD does not apply StaticAccess inside a trait, so the annotation on TranslatesRefusals changed no finding; its reason moves to a comment at the `\OCP\Server::get()` call. PHPMD treats DocumentAssessmentGuard::evaluate()'s $userId as interface-mandated when lib is scanned whole, which is why its three sibling guards carry no such annotation. Four are deliberate and stay, each already reasoned at its site: the title prop nine dashboard widgets declare for the host to draw, GuardRegistry's one parameter per guard kind, and WooPublicationService's coupling and parameter count over five optional collaborators. The baseline rises by exactly those four. psalm.inlineSuppressions returns to 253 and is not raised. Assisted-by: ClaudeCode:claude-opus-5
…without raising it (#3952) * fix(letters): a letter's dates take its language, not a boolean flag dayOf() and shownDay() took `bool $english=false`, which phpmd reports as BooleanArgumentFlag twice (TermLetters.php:414 and :438, from #3884). Both now take the letter's locale, with no default, so every caller names the language it writes in; termStartLines() takes the locale the same way. * test(ci): the dossier link test names the job and queue it runs testWithAJobListTheDossierIsLinkedAfterTheCaseIsAnswered executes LinkCollectionToCaseJob and CollectionLinkQueue; under pcov with --fail-on-risky it is risky without them. The other 49 risky tests are covered by #3942. * fix(quality): the debt ratchet holds at its baseline without raising it eslint.inlineDisables 15 -> 14: FirstTodayWidget declared an unused title prop under an eslint-disable so the host's title would not fall through onto the banner. inheritAttrs: false does that without the prop. phpmd.suppressWarnings 187 -> 182, every one removed suppressed nothing (checked by removing it and re-running phpmd): - TranslatesRefusals StaticAccess: phpmd does not flag a chained \OCP\Server::get()->get() call (control file confirms). - DocumentAssessmentGuard UnusedFormalParameter: the signature comes from GuardEvaluatorInterface, which phpmd honours on a whole-lib run. - AcController::validateAutorisatieScopes CyclomaticComplexity and NPathComplexity, LoadDefaultZgwMappings::getDefaultMappings ExcessiveMethodLength: the methods are under the thresholds. psalm.inlineSuppressions 254 -> 253: UnusedClass on LinkCollectionToCaseJob, which CollectionLinkQueue references. * test(ci): the letter tests name LetterDays, which #3946 put between them and ReadableDay Under pcov with --fail-on-risky, 20 tests in 8 files executed Service\Support\LetterDays (and LetterDaysTest ReadableDay) without naming it.
# Conflicts: # .debt-baseline.json # tests/Unit/Service/Termijn/TermLettersTest.php
…hes-untouched-rows fix(packages): a newer package version refreshes the rows nobody edited here
train: land 23 ready PRs (15 Oct, f)
This branch has not been deployed
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
1 participant
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.
Automated PR to sync development changes to beta for beta release.
Merging this PR will trigger the beta release workflow.