refactor(zwave): extract scenes groups and associations by AlCalzone · Pull Request #4732 · zwave-js/zwave-js-ui

AlCalzone · 2026-07-11T22:01:26Z

Extracts scene, group, virtual-node, and association ownership from ZwaveClient.

Keeps public facade methods, persistence, socket projections, and association behavior compatible.
Uses live ports and generation fencing across driver and virtual-registry replacement.
Review focus: stale async work, virtual-node identity, and scene/group persistence ordering.

Part of #4722. Stacked on #4731.

coveralls · 2026-07-11T22:04:04Z

Coverage Report for CI Build 29574997057

Warning

No base build found for commit ed906c3 on alcalzone-backend-refactor-schedules-templates.
Coverage changes can't be calculated without a base build.
If a base build is processing, this comment will update automatically when it completes.

Coverage: 52.094%

Details

Patch coverage: 43 uncovered changes across 1 file (338 of 381 lines covered, 88.71%).

Uncovered Changes

File	Changed	Covered	%
api/lib/ZwaveClient.ts	73	30	41.1%
Total (5 files)	381	338	88.71%

Coverage Regressions

Requires a base build to compare against. How to fix this →

Coverage Stats


Relevant Lines:	7384
Covered Lines:	4030
Line Coverage:	54.58%
Relevant Branches:	4625
Covered Branches:	2226
Branch Coverage:	48.13%
Branches in Coverage %:	Yes
Coverage Strength:	33.38 hits per line

💛 - Coveralls

Layer 10 cross-stack comment audit for the scene/group/association refactor (PR #4732). Removes decorative section dividers, restated JSDoc that duplicates adjacent code, and history/changelog framing (finding numbers, 'used to be', 'original ZwaveClient did X'). Preserves and in several spots restores non-obvious domain facts: generation-fencing/cancellation semantics, stale-driver-across-await re-resolution, virtual-node lifecycle ownership, deep-equality comparison rationale for object-valued CCs, and the still-unresolved Delegates to X comments trimmed to their misleading-signature facts (getGroups' void-return-but-mutates-in-place quirk) while pure one-line-body restatements are removed. Two comments on ports.ts (hasPhysicalNode, getZWaveNode) are corrected for factual accuracy against their actual ZwaveClient.ts implementations. No runtime/logic changes; comment-only. Co-authored-by: Copilot App <223556219+Copilot@users.noreply.github.com>

…ociation services Adds lazy-resolved port interfaces (persistence, driver, node store, virtual node registry, socket, utils, logger) that SceneService, GroupService, and AssociationService will depend on. No runtime behavior changes yet - ZwaveClient still owns all scene/group/ association state and logic until the following commits wire it up. Co-authored-by: Copilot App <223556219+Copilot@users.noreply.github.com>

Introduces SceneService, a strict-clean class that owns scene collection state and all scene lifecycle logic: create/remove scene, bulk set/get, per-value get/add/remove, and activation (including the per-value delayed-write timeout scheduling and error handling). Structurally/behaviorally identical to the logic currently still living in ZwaveClient.ts - ZwaveClient is wired to delegate to this service in a later commit. Co-authored-by: Copilot App <223556219+Copilot@users.noreply.github.com>

Introduces GroupService, a strict-clean class that owns the user-defined ("multi-target"/virtual) group collection and its full lifecycle: CRUD with physical-node validation (rejects broadcast/ long-range/own-node/duplicate/virtual ids, requires >= 2 valid physical nodes), virtual node materialization/reuse/teardown against the driver's multicast group, per-node virtual-node refresh throttling, group-aware value projection for virtual nodes, and the associated socket emissions. Structurally/behaviorally identical to the logic currently still living in ZwaveClient.ts - ZwaveClient is wired to delegate to this service in a later commit. Co-authored-by: Copilot App <223556219+Copilot@users.noreply.github.com>

Introduces AssociationService, a strict-clean class that owns all device-native (Association CC) association flows: reading/refreshing a node's association groups, checking/adding/removing single or bulk associations, removing all associations from a group, and removing a node from every other node's associations (used on node removal). Preserves the endpoint/lifeline/multi-channel-capability handling, driver-null guards, and error logging exactly as ZwaveClient's original implementation. ZwaveClient is wired to delegate to this service in a later commit. Co-authored-by: Copilot App <223556219+Copilot@users.noreply.github.com>

…tion services Replaces every scene, group, and association method body in ZwaveClient with a thin delegate to _sceneService/_groupService/ _associationService, constructed lazily in init() so ports always resolve the current driver/controller/nodes/value registry/socket/ store across a restart. Every public and underscore-prefixed method keeps its exact existing signature and position in the 110-entry allowedApis list - this is a pure internal refactor with no wire/API/ socket/MQTT/HA/persistence behavior changes. Removes now-unused imports (AssociationGroup, AssociationCheckResult) and now-duplicated module-level constants (GROUP_NAME_MAX_LENGTH, GROUP_ID_MIN) that moved into GroupService. ZwaveClient.ts: 8185 -> 7605 lines (-580). Co-authored-by: Copilot App <223556219+Copilot@users.noreply.github.com>

…tions Adds deterministic, isolated-fake-port unit tests for SceneService (22 tests), GroupService (35 tests), and AssociationService (27 tests), covering success paths, error/edge paths, stale-driver/no-driver guards, persistence rollback on failure, virtual node materialization/ reuse/teardown, throttling, and concurrency-sensitive scenarios (stale scene/group replacement mid-flight). All three exceed the 90/85/90/90 statement/branch/function/line coverage bar. Co-authored-by: Copilot App <223556219+Copilot@users.noreply.github.com>

…cade extraction callApi.test.ts, inboundApis.test.ts, and outboundProducers.test.ts seeded scene/group state by poking the now-removed zwave.scenes/ zwave.groups public fields directly. Since TypeScript doesn't type-check test/ files against tsconfig's api/**-only include, these silently compiled as no-op ad-hoc property assignments and stopped seeding anything once SceneService/GroupService encapsulated that state, without failing to compile. Repoints the seeding at the private service instances' internal arrays (zwave._sceneService._scenes / zwave._groupService._groups) instead of the real _setScenes/_createGroup delegates, preserving the original in-memory-only, disk-independent seeding semantics each test relied on (using the real persisting delegates would leak seeded state across tests sharing the same isolated on-disk store). Co-authored-by: Copilot App <223556219+Copilot@users.noreply.github.com>

Several test files imported store-dependent modules (jsonStore, Gateway, MqttDiscoveryManager, mqttScopedSubscription, configuration template routes/services) via static top-of-file imports, which resolve and start reading module-level state before per-test STORE_DIR/env isolation is established. Under parallel or shuffled test runs, this raced against the real repo store/ directory and could read/write actual on-disk artifacts. Convert the risky static imports to type-only imports plus dynamic import() calls inside beforeAll, strictly after ensureTestEnv() (or an equivalent harness that calls it) has redirected STORE_DIR/env for that file. Confirmed via repeated shuffled-seed runs that no repo store artifacts are generated and behavior is unchanged. Files: - test/lib/Constants.test.ts - test/lib/Gateway.test.ts - test/lib/hass/MqttDiscoveryManager.test.ts - test/lib/hass/mqttScopedSubscription.test.ts - test/lib/http/configurationTemplates.test.ts - test/lib/jsonStore.test.ts - test/lib/zwave/ConfigurationTemplateService.test.ts Co-authored-by: Copilot App <223556219+Copilot@users.noreply.github.com>

Found while proving the full backend suite is order-independent across shuffled seeds (see logger order-independence work): - settingsConstructorBoundary.test.ts: the mocked ZnifferManager was missing a close() method. POST /api/restart awaits oldZniffer.close() on whatever zniffer a prior test in this shared harness left attached, so whichever test happened to run before another restart call - order dependent under shuffle - threw 'oldZniffer.close is not a function'. Add the missing close() mock. - auth.test.ts: 'fails with a generic error when there is no logged-in session user' relied on the route's users.find(u => u.username === user.username) throwing on the missing session user, but Array.prototype.find never invokes its predicate for an empty users array - so under default declaration order (an earlier sibling describe happens to seed a user first) the test passed for an accidental reason. Under shuffle this test can run before any user is ever seeded, short-circuiting to the same observable failure for the WRONG reason. Seed an unrelated user directly so the real precondition holds regardless of run order. Co-authored-by: Copilot App <223556219+Copilot@users.noreply.github.com>

…endence module()/customTransports() memoize the transports list and register every named logger in the module-level logContainer singleton - correct for production (a single process never needs to un-create a logger) but this leaked across describe/it blocks that exercise several module names/configs in the same test process, making logger.test.ts order-dependent. Under --sequence.shuffle seed 4732 this manifested as an assertion failure on a test that expected default logging to be enabled, because an earlier block's config had already mutated the shared/cached logger state. Extract closeCachedTransports() out of setupAll() and add a production-inert __testHooks.reset() seam (closes cached transports, stops the clean job, and closes every registered winston logger via logContainer.close()) so tests can reset to a clean slate between suites. Nothing in the production entrypoint calls __testHooks. Rewrite test/lib/logger.test.ts to call __testHooks.reset() in beforeEach/afterEach. Reproduced the pre-fix seed-4732 failure via git stash before implementing the fix, and confirmed seed 4732 plus 10+ other shuffled seeds now pass consistently. Co-authored-by: Copilot App <223556219+Copilot@users.noreply.github.com>

…ciationService Several methods cached this._driver.getDriver() into a local BEFORE an await (CC-value refresh, a preceding add/remove call, a loop iteration), then kept using that captured reference afterwards. If the driver is torn down and replaced by a restart while the method is suspended, the stale reference would silently keep operating against a destroyed driver instead of the new one. Add a private _requireDriver() helper and call it fresh at every point the original code re-read this._driver, including immediately after every await boundary (refreshCCValues, a prior addAssociations, each association-array loop iteration) across all 7 public methods: getGroups, getAssociations, checkAssociation, addAssociations, removeAssociations, removeAllAssociations, removeNodeFromAllAssociations. Extend AssociationService.test.ts (24 -> 35 tests) with mutable-port deterministic deferred-driver-swap regressions for get/add/removeAll, asserting the exact old-driver-then-new-driver call sequence. Co-authored-by: Copilot App <223556219+Copilot@users.noreply.github.com>

…g to GroupService GroupService received ZwaveClient's _virtualNodes Map by reference at construction time. ZwaveClient.init() replaces _virtualNodes with a brand-new Map on every restart, so a GroupService instance built before a restart kept reading/writing an orphaned, abandoned Map instead of the one the rest of ZwaveClient (and the next GroupService generation) actually uses. Replace the direct Map reference with a closure-backed groupVirtualNodeRegistryPort (has/get/set/delete) in ZwaveClient.ts that resolves this._virtualNodes fresh on every call, matching the pattern already used for the other Group*Port objects. Also add GroupServiceGeneration, a cancellation-token class: ZwaveClient.init() cancels the previous generation and mints a new one before constructing the next GroupService. createGroup/updateGroup/ deleteGroup/removeNodeFromGroups check the generation immediately after their persistence await and abort further virtual-node mutation if it was cancelled while suspended - so an in-flight call left over from a pre-restart GroupService instance can no longer mix its bookkeeping with the new generation's driver/node state. Extend GroupService.test.ts (35 -> 40 tests) with deferred-persistence generation-fencing regressions for restart-during-create/update/delete and current-registry-restoration assertions. Co-authored-by: Copilot App <223556219+Copilot@users.noreply.github.com>

…acade ScenePersistencePort.put() and SceneService's addSceneValue()/ removeSceneValue() were typed to return Promise<unknown>, discarding the actual Promise<ZUISceneRecord<V>[]> shape produced by every code path (jsonStore.put<T>() resolves to the T it was given, and both methods only ever return that or throw). ZwaveClient's _addSceneValue()/_removeSceneValue() facade methods pick up the correct Promise<ZUIScene[]> return type now that the underlying chain is exact end-to-end, matching their doc comments and their callers' expectations. Add expectTypeOf compile-time assertions plus runtime result-shape checks to SceneService.test.ts proving the return type is exactly ZUISceneRecord<V>[], not a wider/unknown type. Co-authored-by: Copilot App <223556219+Copilot@users.noreply.github.com>

…age floor Adds zwaveClientServiceWiring.test.ts (20 tests): production-adapter integration coverage for the ZwaveClient wiring touched by this PR's review findings - association delegate methods, group create/update/ delete/get through the real facade, buildVirtualValueId, startup group restoration, node-removal group-index delegation, and group virtual-node socket projection using valid real VirtualValueID shapes. Exercises the actual ZwaveClient class and production ports/adapters end-to-end rather than the extracted services in isolation (already covered by their own dedicated unit-test files) - deliberately not a broad api/** coverage pool. Add an exact-file 'api/lib/ZwaveClient.ts' coverage threshold entry to BOTH vitest.config.server.ts and vitest.config.ts, set to the measured baseline with this file's 20 tests (22.81/17.02/31.14/22.91 statements/ branches/functions/lines) as a non-regression floor - separate from, and below, the 90/85/90/90 bar the fully-extracted services keep. Empirically proved the gate actually fails by temporarily raising the thresholds above the measured baseline (confirmed npm run coverage:server fails with explicit per-dimension errors), then restored the passing baseline values. Co-authored-by: Copilot App <223556219+Copilot@users.noreply.github.com>

Layer 10 cross-stack comment audit for the scene/group/association refactor (PR #4732). Removes decorative section dividers, restated JSDoc that duplicates adjacent code, and history/changelog framing (finding numbers, 'used to be', 'original ZwaveClient did X'). Preserves and in several spots restores non-obvious domain facts: generation-fencing/cancellation semantics, stale-driver-across-await re-resolution, virtual-node lifecycle ownership, deep-equality comparison rationale for object-valued CCs, and the still-unresolved Delegates to X comments trimmed to their misleading-signature facts (getGroups' void-return-but-mutates-in-place quirk) while pure one-line-body restatements are removed. Two comments on ports.ts (hasPhysicalNode, getZWaveNode) are corrected for factual accuracy against their actual ZwaveClient.ts implementations. No runtime/logic changes; comment-only. Co-authored-by: Copilot App <223556219+Copilot@users.noreply.github.com>

Layer 10 cross-stack comment audit, test side. Removes decorative section dividers, restated JSDoc/inline comments that duplicate the assertions directly below them, and history framing ('Finding #3', 'mirrors the assertion above'). Condenses repeated store-isolation rationale (dynamic import after ensureTestEnv(), since production modules read STORE_DIR at module-evaluation time) to one sentence per call site while keeping each file's unique nuance (Gateway.ts's fs.watch(), MqttDiscoveryManager's CustomDeviceRegistry constructor-arg asymmetry, jsonStore/Constants' import-alone-triggers-the-write case). Removes 2 comments in callApi.test.ts that stated a case the code doesn't handle ('scenes default to [] - no seeding needed'), which is a documented CLAUDE.md style violation, not a lossy trim. zwaveClientServiceWiring.test.ts's 82-line file header is condensed to ~22 lines: most of the removed per-section rationale already lives unchanged on that section's own describe-block title (verified via diff), and the store-isolation mechanics are delegated to outboundProducers.test.ts's own doc comment, which already covers them in full. No test subjects any harness/fake/fixture/builder/cleanup-helper mechanics in isolation; no test removals. Every changed/added test still exercises production entry points (SceneService, AssociationService, GroupService, ZwaveClient facades). All 1089 server tests / 1141 combined tests pass unchanged. Co-authored-by: Copilot App <223556219+Copilot@users.noreply.github.com>

An independent code-review pass flagged two precision issues in the comment audit (both informational, not semantic loss): - GroupService.ts's _virtualNodes field comment established that broadcast virtual-node lifecycle belongs to ZwaveClient, but didn't name where - add a pointer to _updateBroadcastNodeValues() so the fact is locally discoverable without cross-file searching. - zwaveClientServiceWiring.test.ts's file header described the nodeId -> groupIds index as 'only rebuilt from _groups at construction', which reads as though construction were the index's only rebuild trigger. It's actually rebuilt after every mutating call too; reworded to state precisely why the node-removal test seeds via disk (a direct post-construction _groups overwrite leaves the index stale until the next mutation), without the ambiguity. Co-authored-by: Copilot App <223556219+Copilot@users.noreply.github.com>

A second independent code-review pass caught a factual regression the audit itself introduced: Constants.test.ts, Gateway.test.ts, jsonStore.test.ts, and logger.test.ts originally correctly said their production file 'statically imports storeDir/logsDir from config/app.ts' (each does - verified via direct grep of their import statements), but condensation had swapped this to 'transitively imports' across all four, apparently applied as a uniform substitution without re-checking each file's actual import graph. Revert to 'statically' in all four. logger.test.ts also had two more issues from the same review pass: - Its second paragraph originally described two distinct module-level singletons (the transports-list array and the winston.Container), but the condensed wording read as if the transports list lived inside the Container. Restore the distinguishing 'and' clause; verified via logger.ts that transportsList (a standalone array) and logContainer (new winston.Container()) are in fact separate structures. - Restore that setupCleanJob()'s un-awaited clean() promise makes its own logger.info(...) calls through this same file's shared transports cache - a non-obvious link between the rotation-cleanup background work and the exact cache this file's tests reset, dropped as part of the DISABLE_LOG_ROTATION paragraph's condensation. Co-authored-by: Copilot App <223556219+Copilot@users.noreply.github.com>

…tion Co-authored-by: Copilot App <223556219+Copilot@users.noreply.github.com>

…file The write-order comment restored in 3d35ef3 was deliberately removed by base's own comment audit (ce123bc, 87220f8, 8af7ee9), which trimmed this shared file to a single non-obvious comment. Re-adding it broke consistency with 5 sibling toHaveBeenNthCalledWith blocks in the same file that never had such a comment. Base owns this file's content. Co-authored-by: Copilot App <223556219+Copilot@users.noreply.github.com>

…le reset Removes the __testHooks.reset() export added to api/lib/logger.ts during review fixes. It existed solely so tests could clear the module-level winston.Container/transports-list/cleanJob singletons between cases - a production-only test seam per PR #4723's accepted design feedback. logger.ts's singletons are genuine process-lifetime state (issue refactor, so instead test/lib/logger.test.ts now calls vi.resetModules() per describe block and re-imports logger.ts (and winston, so instanceof checks compare classes from the same generation) fresh, giving each block its own isolated module instance instead of resetting shared state through a productio-only API. Verified order-independent: full file passes in default order and under --sequence.shuffle with seed 4732 plus 6 other random seeds. Co-authored-by: Copilot App <223556219+Copilot@users.noreply.github.com>

Rewrite zwaveClientServiceWiring.test.ts per PR #4723's accepted review rules, replacing delegation/identity-mechanics tests with tests that exercise the real, unmocked AssociationService/GroupService through ZwaveClient's real public facade: - Association tests: replace 6 tests asserting _associationService.X spy calls (toHaveBeenCalledWith) with 7 tests that drive a stateful fake controller association table and assert the real observable association state returned by getAssociations()/checkAssociation()/addAssociations()/ removeAssociations()/removeAllAssociations()/ removeNodeFromAllAssociations() - including a genuine Forbidden_SelfAssociation branch, not a canned OK mock. - Group CRUD facade: remove a duplicate _getGroups() 'same array reference' identity test already covered by GroupService.test.ts. - _buildVirtualValueId tests: rewrite 3 of 5 through _createGroup()/ _updateGroup() using a fake driver's real getDefinedValueIDs(), asserting the real emitted ZUIValueId fields instead of calling the private method directly. The remaining 'existing field preservation merge' test is removed: both real call sites (GroupService._updateVirtualNodeValues, ZwaveClient._doUpdateBroadcastNodeValues) reset the target node's values to {} immediately before the rebuild loop reads 'existing' from that same map, so the merge branch is unreachable via any real path today - confirmed by the rewritten test failing until removed. Documented in a code comment rather than silently dropped. - Node removal: remove the removeNodeFromGroupsSpy toHaveBeenCalledOnce/toHaveBeenCalledWith assertions; keep the real, observable in-memory and on-disk group/node state assertions. - Socket projection: replace a double (valueId as any).id / (node as any).id cast with full type inference through TypedEventEmitter.once()'s generic signature, typing only the Promise's own resolved shape. No production code changed. Test count: 20 -> 18 (net -2, reconciled above). Remaining private-field (as any) pokes for driver/service fixture setup match the established, inherited, pre-Layer-10 convention already used in callApi.test.ts/outboundProducers.test.ts/ inboundApis.test.ts/hass/server.test.ts. Co-authored-by: Copilot App <223556219+Copilot@users.noreply.github.com>

…test Per PR #4723's accepted review rules, audited AssociationService.test.ts/ GroupService.test.ts/SceneService.test.ts for delegation-mechanics framing and raw-crash-text locking: - GroupService.test.ts: 3 rollback/propagation tests asserted the exact text of an arbitrary test-fixture error ('persist failed' from createPersistencePort()'s fake, 'rejected node set' from an inline driver.getMulticastGroup override) via .rejects.toThrow('<text>'). The real, semantic contract under test is that a downstream failure propagates and rolls back in-memory state - already covered by the surrounding assertions - not the incidental message text a test author happened to pick. Loosened to argument-less .rejects.toThrow(). - AssociationService.test.ts: retitled 'delegates to driver.controller.checkAssociation' to 'returns the real driver check result for the given source/group/association' - the assertions already tested real observable behavior (the real return value plus the real external-driver-boundary call), only the title's framing language was mechanics-flavored. No other changes: all other toThrow(...)/toHaveBeenCalledWith(...) text in these 3 files was verified against the real production throw sites (AssociationService.ts/GroupService.ts/SceneService.ts) and matches deliberate, user-facing semantic error messages, or asserts the external zwave-js driver boundary (the correct level for an adapter/wrapper service with no other observable side effect) - left unchanged. Driver/ZWaveNode/VirtualNode "as unknown as X" casts audited and confirmed structurally required by TypeScript when faking large third-party zwave-js SDK interfaces with minimal objects (empirically confirmed narrower alternatives still require the same "unknown" hop, and the established ScheduleService.test.ts precedent's alternative pattern actually produces latent, currently-ungated tsc diagnostics that this pattern avoids) - left unchanged. GroupServiceGeneration fencing tests already exercise real restart/persistence/notification invariants through public behavior, never asserting ".cancelled" directly - left unchanged. No production code changed. Co-authored-by: Copilot App <223556219+Copilot@users.noreply.github.com>

…tles Two describe-block titles carried forward an internal development-time "finding #N" bookkeeping label with no meaning outside that process: - SceneService.test.ts: 'facade return type precision (Finding #3)' -> 'facade return type precision'. - zwaveClientServiceWiring.test.ts: 'group CRUD facade (...) - real GroupService + jsonStore wiring (the closure-backed registry finding #2 fixed)' -> '... wiring, with _getGroups() reading the live registry after each mutation' - keeps the real, still-useful behavioral fact the title was gesturing at (each test in the block reads back via _getGroups()/disk after a mutation, so it can't pass against a stale registry snapshot) without the process reference. No assertions changed. Confirmed via a full diff of this PR's Layer 10 delta against its pre-Layer-10 base that these were the only two such references anywhere in the delta. Co-authored-by: Copilot App <223556219+Copilot@users.noreply.github.com>

Use structural upstream handles in the extracted-service ports so tests can exercise public service behavior without nominal external-class casts. Remove accessor, identity, type-only, and collaborator-mechanics assertions while retaining restart, persistence, projection, and association behavior coverage. Co-authored-by: Copilot App <223556219+Copilot@users.noreply.github.com>

Delete the integration suite that drove private service and driver state. Preserve the observable scene and group facade compatibility checks through public underscore APIs, isolated persistence, and socket payloads. Co-authored-by: Copilot App <223556219+Copilot@users.noreply.github.com>

…th session req.session.user can be undefined with auth disabled and no prior login; isAuthenticated still calls next() in that case. The unguarded lookup let .find()'s predicate dereference user.username on undefined once users.json was non-empty, throwing instead of falling through to the existing "User not found" response. This mirrors a fix already established and tested elsewhere in this stack's history (1c23c2f, predating this rewritten base) that the base's own comment-only-labeled pass silently undid. Also drops a stale seedUser() call and its pre-guard-era comment from the corresponding test: an earlier conflict resolution in this rewrite grafted 6100f49's shuffle-order workaround (written against an unguarded predecessor of this handler) onto this already-guarded test, where the ternary short-circuits before .find() ever runs and the seed is moot. Co-authored-by: Copilot App <223556219+Copilot@users.noreply.github.com>

…f Layer 10's accepted delta api/routes/auth.ts aside, these were the only 2 files where this rebase's cumulative diff against the new base didn't match the original PR's cumulative diff against its own base (verified file-by-file across the full touched set). Root cause: the replayed range includes non-merge commits from a 'merge updated layer 9 base' step whose own merge commit is (correctly) excluded from a non -p rebase. One of those commits' store-isolation fix touched MqttDiscoveryManager.test.ts and mqttScopedSubscription.test.ts, but the original merge's resolution discarded that change for these two Layer9-owned files, keeping the newer parent's version outright - a choice this base's own tip already matches. Replaying the commit's raw diff without that merge's resolution silently reintroduced the discarded change. Confirmed by diffing this PR's original base against its own final head: both files are byte-identical there, across all 29 original commits. Restores both files to this base's content, matching the original PR's true net effect of zero change to either file. Co-authored-by: Copilot App <223556219+Copilot@users.noreply.github.com>

…suites test/lib/Constants.test.ts, test/lib/jsonStore.test.ts, and test/lib/logger.test.ts each imported ensureTestEnv/cleanupTestEnv from a './http/env.ts' that has never existed on this branch; the real helper lives at './shared/env.ts'. All 3 files failed at import time with zero tests running, surfaced only by actually executing a shuffled-seed run rather than by diff-size comparison (a same-size, wrong-target import edit doesn't show up as a line-count anomaly). Traced to 2 of this rebase's own earlier conflict-resolution commits. Constants.test.ts and jsonStore.test.ts have byte-identical blobs at both the old base and this new base, so no adaptation is legitimate for either - both are now byte-identical to the original PR's final content. logger.test.ts's base genuinely evolved independently between the two branches, so only its broken env.ts references were corrected in place, preserving its otherwise-legitimate freshLogger() adaptation. test/lib test/runtime: 50/50 files, 842/842 tests passing after this fix. Co-authored-by: Copilot App <223556219+Copilot@users.noreply.github.com>

Remove the out-of-scope HTTP fix added during Layer 10 replay. Co-authored-by: Copilot App <223556219+Copilot@users.noreply.github.com>

AlCalzone marked this pull request as draft July 11, 2026 22:06

AlCalzone mentioned this pull request Jul 12, 2026

refactor(zwave): extract firmware and inclusion #4733

Draft

robertsLando mentioned this pull request Jul 14, 2026

Landing Plan: backend-refactor & dashboard-neo integration branches (review order, merge train & issue closure) #4745

Open

27 tasks

AlCalzone force-pushed the alcalzone-backend-refactor-schedules-templates branch from 89a3ab8 to d09a565 Compare July 16, 2026 14:19

AlCalzone force-pushed the alcalzone-backend-refactor-scenes-groups-associati branch from 77beb46 to abf39e8 Compare July 16, 2026 16:43

AlCalzone force-pushed the alcalzone-backend-refactor-schedules-templates branch from d09a565 to 978a891 Compare July 16, 2026 20:12

AlCalzone force-pushed the alcalzone-backend-refactor-scenes-groups-associati branch from efea0b6 to ed70f11 Compare July 16, 2026 20:49

AlCalzone force-pushed the alcalzone-backend-refactor-schedules-templates branch from 978a891 to ed906c3 Compare July 17, 2026 09:57

AlCalzone and others added 18 commits July 17, 2026 12:11

AlCalzone and others added 12 commits July 17, 2026 12:18

docs(zwave): restore write-order comment dropped by merge auto-resolu…

95ae4c7

…tion Co-authored-by: Copilot App <223556219+Copilot@users.noreply.github.com>

revert(auth): preserve authoritative Layer 9 behavior

de200c3

Remove the out-of-scope HTTP fix added during Layer 10 replay. Co-authored-by: Copilot App <223556219+Copilot@users.noreply.github.com>

AlCalzone force-pushed the alcalzone-backend-refactor-scenes-groups-associati branch from ed70f11 to de200c3 Compare July 17, 2026 10:53

Provide feedback

Saved searches

Use saved searches to filter your results more quickly

Uh oh!

Uh oh!

refactor(zwave): extract scenes groups and associations#4732

refactor(zwave): extract scenes groups and associations#4732
AlCalzone wants to merge 30 commits into
alcalzone-backend-refactor-schedules-templatesfrom
alcalzone-backend-refactor-scenes-groups-associati

AlCalzone commented Jul 11, 2026 •

edited

Loading

Uh oh!

coveralls commented Jul 11, 2026 •

edited

Loading

Uh oh!

Reviewers

Assignees

Labels

Projects

Milestone

Development

Uh oh!

2 participants

Uh oh!

Uh oh!

Conversation

AlCalzone commented Jul 11, 2026 • edited Loading Uh oh! There was an error while loading. Please reload this page.

Uh oh!

Uh oh!

coveralls commented Jul 11, 2026 • edited Loading Uh oh! There was an error while loading. Please reload this page.

Uh oh!

Coverage Report for CI Build 29574997057

Coverage: 52.094%

Details

Uncovered Changes

Coverage Regressions

Coverage Stats

💛 - Coveralls

Uh oh!

Reviewers

Assignees

Labels

Projects

Milestone

Development

Uh oh!

2 participants

AlCalzone commented Jul 11, 2026 •

edited

Loading

coveralls commented Jul 11, 2026 •

edited

Loading