Tests like testFindTenantsByTitle create 261 tenants and delete them via
deleteEntitiesAsync, which only waits for HTTP responses — not housekeeper
completion. Each tenant deletion submits ~30 TenantEntitiesDeletionHousekeeper
tasks (~7800 tasks total), which cascade further. The teardown's deleteTenant
then waits for lag==0 with a 90s timeout, which is insufficient for this
backlog and causes ConditionTimeoutException.
Fix: add awaitHousekeeperDrained() (5-min timeout) called at the start of
teardownWebTest so any pending housekeeper work from the test body drains
before per-tenant teardown deletions begin.
Co-Authored-By: Claude Sonnet 4.6 <noreply@anthropic.com>
doGet/doGetAsyncTyped assert HTTP 200 internally, so any non-200 response
throws AssertionError which Awaitility re-throws immediately instead of
continuing to poll.
Add .ignoreExceptions() to three additional await() polling loops that
call HTTP helpers:
- AbstractMqttV5ClientSparkplugAttributesTest: two doGetAsyncTyped calls
polling for attribute keys after NBIRTH/DBIRTH
- AbstractMqttAttributesIntegrationTest: doGetAsyncTyped polling for
attribute values after client publish
Co-Authored-By: Claude Sonnet 4.6 <noreply@anthropic.com>
doGet(url, Class) asserts HTTP 200 internally, so when the Sparkplug
device hasn't been created yet the method throws AssertionError instead
of returning null. Awaitility propagates Error immediately rather than
continuing to poll, causing the test to fail after ~3 s instead of
retrying for up to 200 s.
Add .ignoreExceptions() to both await() calls in
connectClientWithCorrectAccessTokenWithNDEATHCreatedDevices and
connectClientWithCorrectAccessTokenWithNDEATHWithAliasCreatedDevices so
that a transient 404 is treated as "condition not yet met" and polling
continues as intended.
Co-Authored-By: Claude Sonnet 4.6 <noreply@anthropic.com>
After the MQTT transport closes the session on RPC delivery timeout,
the server re-queues the RPC asynchronously. The test was asserting
the RPC status immediately after the client disconnect latch fired,
before the server-side re-queuing had completed — so the status was
still SENT instead of the expected QUEUED.
Replace the direct doGet assertion with an Awaitility poll that waits
up to DEFAULT_WAIT_TIMEOUT_SECONDS for the status to become QUEUED.
Fixes flaky: MqttV5CloseTransportSessionOnRpcDeliveryTimeoutIntegrationTest
Co-Authored-By: Claude Sonnet 4.6 <noreply@anthropic.com>
When building with -T6, multiple JS modules run yarn install concurrently.
The --mutex network flag serializes yarn processes, but if node_modules from
a previous build (or a different branch/version) is left on disk, yarn may
skip reinstalling packages — including devDependencies — treating the stale
state as up-to-date. This causes tools like tsc to be missing at build time.
Configure maven-clean-plugin to delete node_modules during mvn clean for
all JS modules (ui-ngx, msa/js-executor, msa/web-ui). Yarn restores from
its global cache (~/.cache/yarn), so the overhead is relinking only.
Co-Authored-By: Claude Sonnet 4.6 <noreply@anthropic.com>
When Execute RPC is sent to FW Update resource (/5/0/2), the test client's
startUpdating() scheduled the client stop with 0 delay. This caused a race
where the client stopped before the CoAP Execute response (2.04 Changed)
was delivered to the server, resulting in RequestCanceledException and
INTERNAL_SERVER_ERROR instead of CHANGED.
Adding a 1-second delay before leshanClient.stop() ensures the CoAP
response is transmitted and received before the client disconnects, fixing
the flaky testExecuteUpdateFWById_Result_CHANGED test.
Co-Authored-By: Claude Sonnet 4.6 <noreply@anthropic.com>
Tests that read shared static state (e.g. testAllowedUrls with 8.8.8.8)
could run concurrently with tests that mutate it (e.g. testAdditionalBlockedSingleIp),
causing intermittent failures. Class-level @ResourceLock serializes all tests.
Co-Authored-By: Claude Opus 4.6 <noreply@anthropic.com>
- Use ID token claims as the source of truth for Apple OAuth2 attributes
- Added Apple mapper type to OAuth2 client data validation
- Consolidated duplicated validation logic for BASIC, GITHUB, and APPLE mapper types
Co-Authored-By: Claude Opus 4.6 <noreply@anthropic.com>
The lastCommit field was shared across all repository keys, causing
MissingObjectException when multiple repositories were registered.
When onUpdate fired for repo A it overwrote lastCommit, and subsequent
listFiles/getFileContent calls for repo B used repo A's commit whose
tree objects don't exist in repo B's object database.
Changed to a per-key Map<String, RevCommit> so each repository's
resolved commit is stored and retrieved independently.
Co-Authored-By: Claude Opus 4.6 <noreply@anthropic.com>
Move TestDbCallbackExecutor from rule-engine test sources to
common/util main sources as DirectListeningExecutor, making it
available to all modules. Convert to an enum singleton since the
executor is stateless. Widen JpaAbstractDaoListeningExecutorService
service field type from JpaExecutorService to ListeningExecutor to
allow injecting DirectListeningExecutor in tests. Fix
AbstractChunkedAggregationTimeseriesDaoTest NPE by injecting the
direct executor into the spy.
Co-Authored-By: Claude Opus 4.6 <noreply@anthropic.com>
The stack-based check in DefaultTbContext.input() only catches a loop
on the second Kafka round-trip through the same rule chain. When the
target rule chain equals the node's own rule chain (direct self-loop),
the first iteration is always unnecessary and should be rejected
immediately before any enqueue occurs.
Add an eager check in TbRuleChainInputNode.onMsg(): if the resolved
targetRuleChainId equals ctx.getSelf().getRuleChainId() (the rule chain
that contains this node), tellFailure() is called right away.
Together with the existing stack-based check in DefaultTbContext.input()
this forms a two-layer defence:
- Layer 1 (TbRuleChainInputNode): direct loops A→A, caught immediately
- Layer 2 (DefaultTbContext): indirect cycles A→B→A, caught on second pass
Co-Authored-By: Claude Sonnet 4.6 <noreply@anthropic.com>
When a TbRuleChainInputNode has `forwardMsgToDefaultRuleChain=true` and the originator's
default rule chain is the same as the rule chain containing this node, the message enters
an infinite loop: the node forwards to the default rule chain, which routes back to the
same node, which forwards again, causing unbounded recursion and 100% CPU on rule-engine.
Fix: detect the loop in DefaultTbContext.input() by checking whether the calling rule node
is already present in the message's return stack (TbMsgProcessingCtx). On the second+
iteration the stack already contains the (ruleChainId, ruleNodeId) pair of the node,
so the call is a cycle. In that case tellFailure() is called with a descriptive message
and a WARN log is emitted instead of re-enqueuing the message.
Co-Authored-By: Claude Sonnet 4.6 <noreply@anthropic.com>
When agg=NONE (the default), findAllAsyncWithLimit() executed a
synchronous JPA query on the calling thread and wrapped the result in
Futures.immediateFuture(). Since this runs inside the access-validator
single-thread executor callback chain, a slow query blocks all other
telemetry and attribute requests.
Offload findAllAsyncWithLimit() to the JpaExecutorService work-stealing
pool via service.submit(), matching the pattern already used by
aggregated queries (findAndAggregateAsync).
Closes#15095
Co-Authored-By: Claude Opus 4.6 <noreply@anthropic.com>
Delay and deduplication rule nodes were creating brand new TbMsg objects
instead of copying the original, which reset the ruleNodeExecCounter to 0.
This allowed bypassing the maxRuleNodeExecutionsPerMessage limit.
Co-Authored-By: Claude Opus 4.6 <noreply@anthropic.com>
Viacheslav Klimov
Sergey Matvienko
Vladyslav_Prykhodko
Maksym Tsymbarov
Ekaterina Chantsova
Dmytro Khylko
Dmytro Skarzhynets
Dmytro Skarzhynets