Sync csa branch with main by silabs-matter-ci-bot[bot] · Pull Request #1148 · SiliconLabsSoftware/matter_sdk

silabs-matter-ci-bot · 2026-08-15T00:23:52Z

This PR syncs the csa branch with the main branch.

PR MUST BE MERGED WITH MERGE COMMIT - ADMIN MUST ENABLE THE OPTION

…nterfaces (#73438) On LwIP-based platforms, network interfaces are commonly identified by their netif name. Name-based lookup is fragile, as interfaces are usually named by vendor specific implementation. These new accessors let a platform report its Thread and external interfaces explicitly, avoiding name comparisons and the need for a possible lookup update when interface names change. These accessors have also a default implementation, so platforms that do not override them are unaffected. Signed-off-by: Cristian Bulacu <cristian.bulacu@nxp.com>

…m-bot) * add ota for stm32wba platform with oemirot * [pre-commit.ci] auto fixes from pre-commit.com hooks for more information, see https://pre-commit.ci * Restyled by gn * add apache 2 license * add c++ balise Co-authored-by: Copilot Autofix powered by AI <175728472+Copilot@users.noreply.github.com> * [pre-commit.ci] auto fixes from pre-commit.com hooks for more information, see https://pre-commit.ci * fix include for hal driver and fix IA comments * [pre-commit.ci] auto fixes from pre-commit.com hooks for more information, see https://pre-commit.ci * update readme for spell check job * [pre-commit.ci] auto fixes from pre-commit.com hooks for more information, see https://pre-commit.ci * update readme for ota * [pre-commit.ci] auto fixes from pre-commit.com hooks for more information, see https://pre-commit.ci * fix include order * update readme with st link * [pre-commit.ci] auto fixes from pre-commit.com hooks for more information, see https://pre-commit.ci * update readme with comments on flag for ota * [pre-commit.ci] auto fixes from pre-commit.com hooks for more information, see https://pre-commit.ci * fix misspell * [pre-commit.ci] auto fixes from pre-commit.com hooks for more information, see https://pre-commit.ci --------- Co-authored-by: pre-commit-ci[bot] <66853113+pre-commit-ci[bot]@users.noreply.github.com> Co-authored-by: Restyled.io <commits@restyled.io> Co-authored-by: Copilot Autofix powered by AI <175728472+Copilot@users.noreply.github.com>

* silabs/zephyr: fix Kconfig indentation Apply the same rules than the Zephyr and the Linux projects: - Indent with tab - Indent help with tab + 2 spaces - Limit on 80 columns (In fact, 84 columns because Matter uses 4 spaces for a tab) - No indentation inside choice/if statements - No trailing spaces * silabs/zephyr: fix promptless symbol Silicon Labs Kconfigs provide default values for plenty of options. However, if the type of the symbol is provided, the symbol is a new instance, which is generally not the intention. This patch is specifically fixing BT_HOST_CRYPTO because Gemini is complaining about it. However, this change should be applied to the others symbols. Signed-off-by: Jérôme Pouiller <jerome.pouiller@silabs.com> * silabs/zephyr: fix DeviceInfoProvider initialization order Server::Init() triggers various callbacks (e.g. attribute reads, fabric enumeration) that can call DeviceLayer::GetDeviceInfoProvider(). If gExampleDeviceInfoProvider is registered *after* Server::Init() returns, those callbacks silently get a NULL provider and fail to retrieve device info (product name, vendor ID, etc.), leading to incorrect behaviour during commissioning. Fix by moving the SetStorageDelegate() / SetDeviceInfoProvider() calls to before Server::Init(), using initParams.persistentStorageDelegate directly (which is the same object that Server::GetPersistentStorage() would return afterwards). Also guard the OperationalKeystore::Init() call under CONFIG_OPERATIONAL_KEYSTORE since persistentStorage is only needed in that context. Co-authored-by: Copilot <223556219+Copilot@users.noreply.github.com> * silabs/zephyr: update overlays for siwx917 ulpuart and sram was already configured with the configuration in Zephyr. So these entries were useless. In addition, the offset of the storage_partition overlapped with ota_swap_partition. So the OTA process would probably overwrite the data. The available flash ends at 0x00400000. So, we slightly reduced the size of the code_partition to allocate the storage_partition at the end of the available area: 0x00400000 -- end -- 0x003e9000 storage 0x003e1000 settings 0x003e0000 factory 0x00202000 code This patch also align the size of the storage partitions on brd4338 and brd4342 (was 68kB, updated to 92kB). Signed-off-by: Jérôme Pouiller <jerome.pouiller@silabs.com> * silabs/zephyr: fix PSA_CRYPTO_SW_ONLY dependency Currently, Zephyr compilation fails with: warning: default on the choice symbol PSA_CRYPTO_SW_ONLY will have no effect, as defaults do not affect choice symbols Indeed PSA_CRYPTO_SW_ONLY should be selected either with choice PSA_CRYPTO_DRIVER_IMPLEMENTATION default PSA_CRYPTO_SW_ONLY endchoice or with: depends on PSA_CRYPTO_SW_ONLY BTW, since v2026.06 PSA_CRYPTO_SW_ONLY is the default for swix91x. So this statement can be removed. Signed-off-by: Jérôme Pouiller <jerome.pouiller@silabs.com> * silabs/zephyr: fix CMSIS dependency Currently, the compilation may fail with: In file included from .../zephyr/arch/arm/asm_inline.h:18, from .../zephyr/arch/arm/arch.h:30, from .../zephyr/arch/cpu.h:19, from .../zephyr/kernel_includes.h:36, from .../zephyr/kernel.h:17, from .../matter/src/system/SystemMutex.h:60, from .../matter/src/system/SystemMutex.cpp:26: .../zephyr/arch/arm/asm_inline_gcc.h:24:10: fatal error: cmsis_core.h: No such file or directory 24 | #include <cmsis_core.h> | ^~~~~~~~~~~~~~ compilation terminated. By specifying cmsis_6 and mbedtls explicitly, we guarantee their CMakeLists.txt will be included first. Thus, INTERFACE_INCLUDE_DIRECTORIES is properly filled when Matter CMakelists.txt is included. Signed-off-by: Jérôme Pouiller <jerome.pouiller@silabs.com> * silabs/zephyr: fix support for WiFi During my tests, the commissioning failed without CONFIG_CHIP_ENABLE_WIFI_STATION=y. In addition, the "select" statement only select the specified symbol without taking the dependencies into account. It is better to reverse the dependency (CHIP_WIFI is default y but depends on WIFI). Signed-off-by: Jérôme Pouiller <jerome.pouiller@silabs.com> * silabs/zephyr: relocate POSIX_AEP_CHOICE_NONE POSIX_AEP_CHOICE_NONE is global to the whole Matter project, is not specific to the lighting sample. So, relocate the option. Signed-off-by: Jérôme Pouiller <jerome.pouiller@silabs.com> * silabs/zephyr: drop useless USE_SEGGER_RTT=n Many board in Matter explicitly disable USE_SEGGER_RTT=n. It seems it is because these board enable RTT by default in their _defconfig. Silabs board don't do that (the user has to explicitly reference rtt-console/rtt-console.conf snippet). So, the current USE_SEGGER_RTT=n is either useless or error prone for the user who really want to use RTT. Signed-off-by: Jérôme Pouiller <jerome.pouiller@silabs.com> * silabs/zephyr: drop explicit dependency on TLS CONFIG_THREAD_LOCAL_STORAGE=y is already the default in Zephyr. Explicit selection in prj.conf does not seem useful. The origin of this setting maybe related to CONFIG_SLOW_FLASH_DATA (which is suggested for siwx91x). Help of SILABS_SIWX91X_SLOW_FLASH_DATA says: With current GCC versions (14.x), this option is incompatible with Thread Local Storage (TLS). User may either: - Use a patched version of gcc (preferred option), or - Rebuild libc (CONFIG_PICOLIBC_USE_MODULE=y) without TLS support (CONFIG_THREAD_LOCAL_STORAGE=n). In this case, access to "errno" will no longer be thread-safe. Signed-off-by: Jérôme Pouiller <jerome.pouiller@silabs.com> * silabs/zephyr: drop useless dependency on UART_ASYNC_API Lighting application does not seems to rely on UART async API. BTW, siwx91x does not support this API. So, CONFIG_UART_ASYNC_API=y seems to be a mistake. Signed-off-by: Jérôme Pouiller <jerome.pouiller@silabs.com> * silabs/zephyr: drop useless dependency on SECURE_STORAGE_..._AES_GCM SECURE_STORAGE_ITS_TRANSFORM_AEAD_SCHEME_AES_GCM is already the default in Zephyr. (BTW, SECURE_STORAGE_ITS_TRANSFORM_AEAD_SCHEME_CHACHA20_POLY1305 is probably a faster alternative) Signed-off-by: Jérôme Pouiller <jerome.pouiller@silabs.com> * silabs/zephyr: group Kconfig options Order Kconfig options: 1. Subsystems. Because of "depends on", many options are automatically selected with these options. 2. Application related options 3. Shell commands. It makes sense to easily add remove shell commands during the development 4. Stack sizes. It makes sense to group them since the user often has to tune them. 5. Debug and temporary options (empty in this sample) Signed-off-by: Jérôme Pouiller <jerome.pouiller@silabs.com> * silabs/zephyr: drop debug features lighting-app/silabs/zephyr/prj.conf contain a bunch of option related to the log and the shell threads. These option are probably not required for the sample. The defaults parameters should be sufficient and we don't expect to see a such level of details in a sample project. Signed-off-by: Jérôme Pouiller <jerome.pouiller@silabs.com> --------- Signed-off-by: Jérôme Pouiller <jerome.pouiller@silabs.com> Co-authored-by: Copilot <223556219+Copilot@users.noreply.github.com> Co-authored-by: Jean-Francois Penven <67962328+jepenven-silabs@users.noreply.github.com>

…in fixes that need backport (#73530)

* [AVAnalysis] Add AV Analysis Node device type (0x0149) to matter-devices.xml Generated with Alchemy from the spec's AVAnalysisNode.adoc * [AVAnalysis] Regenerate code for the AV Analysis Node device type

* Update TCP transport test label in PICS.yaml * Update ci-pics-values * Update TCP_Tests.py * Refactor TCP session handling in TC_SEPR_2_3 * Update TC_SC_4_3.py * Update TC_SC_4_1.py * Fixing review comments Updated test case steps and removed MCORE.SC.S.TCP from PICS definitions.

… API (#73533) * Fixing the wifi ICD behavior * [pre-commit.ci] auto fixes from pre-commit.com hooks for more information, see https://pre-commit.ci --------- Co-authored-by: pre-commit-ci[bot] <66853113+pre-commit-ci[bot]@users.noreply.github.com>

…for Linux and NuttX, and fixing potential use-after-free. (#39392) * Prevent segfaults in MdnsAvahi::Resolve for Linux and NuttX * Fixed commit mistake

* fix(tv-casting-app): scope R8 proguard keeps to the 19 slim TLV decoder clusters * docs(tv-casting-app): add measured SDK size impact and update R8 shrinking notes * fix(tv-casting-app): replace Messages with MediaFileManagement in R8 proguard keeps * docs(tv-casting-app): corrections to doc and comments * docs(tv-casting-app): prettier formatting corrections as mandated by CI * fix: add proguard to spell-check wordlist * fix(tv-casting-app): document sources of truth for chip.devicecontroller R8 keeps * docs(tv-casting-app): prettier formatting corrections by pre-commit

…Release wrapper (#73546) * Convert group cluster CommonIterator consumers to RAII Replace manual Iterate*()/Release() call sites in GroupKeyManagementCluster, GroupcastCluster and GroupsCluster with AutoRelease, so early returns cannot leak pooled iterators (see #73331, #73296). Sites that treat a null iterator as an empty result keep that behavior via an explicit IsNull() check. Delete the file-local AutoReleaseIterator wrapper in GroupsCluster.cpp in favor of the shared AutoRelease. Reference the wrappers from the GroupDataProvider::Iterate*() doc comments and fix the IterateGroupInfo retval typo (EndpointIterator -> GroupInfoIterator). * add copyright header * [pre-commit.ci] auto fixes from pre-commit.com hooks for more information, see https://pre-commit.ci --------- Co-authored-by: pre-commit-ci[bot] <66853113+pre-commit-ci[bot]@users.noreply.github.com>

…#73200) * Refactor ZoneManagement * Refactor BooleanStateConfig and FanControl * [pre-commit.ci] auto fixes from pre-commit.com hooks for more information, see https://pre-commit.ci * Replace calls on AmbientContext, OccupancySensing and OnOff clusters * Change Humidistat, OnOffLightning and WindowCovering clusters * [pre-commit.ci] auto fixes from pre-commit.com hooks for more information, see https://pre-commit.ci * Update Chime, LevelControl and CameraAVStream clusters * Update ResourceMonitoring, TimeFormatLocalization, UnitLocalization and ValveConfiguration clusters * [pre-commit.ci] auto fixes from pre-commit.com hooks for more information, see https://pre-commit.ci --------- Co-authored-by: pre-commit-ci[bot] <66853113+pre-commit-ci[bot]@users.noreply.github.com>

…ng (Auto-merged by platform-bot)

* Removed parameter ota_image_expected_version from test TC_SU_2_7 added method to read the software version from the ota image * Updated TC SU 2.5 to read the software version from the ota image * Typo in message for not found * Fix from code review * Added MAGIC compare to check if is ota IMAGE. Removed path exists * Updated step 1 to ensure the transition from Download to kApplying was instant ( or almost ) * Revert "Updated step 1 to ensure the transition from Download to kApplying was instant ( or almost )" This reverts commit 2bfcb36764f131d0ceea70baa407614088c1e431. * Solving cconflict for TC_SU_2_2 * Removed f strings at log statements. * fix: Added docstring to new methods and fixed typos.

…#73556) * Convert IterateX/Release calls in the core with AutoRelease wrapper Replace manual Iterate*()/Release() calls in WriteHandler, CommandHandlerImpl and Server::RejoinExistingMulticastGroups with AutoRelease, so early returns cannot leak pooled iterators (#73331). remove WriteHandler's file-local AutoReleaseGroupEndpointIterator wrapper. * [pre-commit.ci] auto fixes from pre-commit.com hooks for more information, see https://pre-commit.ci --------- Co-authored-by: pre-commit-ci[bot] <66853113+pre-commit-ci[bot]@users.noreply.github.com>

…560) #72559 stopped minimal-mDNS from sending into the Thread mesh by skipping wpan* interfaces, but only in AddressPolicy_DefaultImpl. The libnl policy, which a Linux build selects with chip_minmdns_default_policy = "libnl", filters on the link flags and has no equivalent, so it still hands wpan0 to the mDNS server and the traffic #72559 set out to avoid goes out over the mesh. Apply the same rule there. Matching on the interface name is not satisfying and libnl could offer something better, but OpenThread presents wpan0 as a tun device, so the link type does not identify it either -- and keeping the two policies in agreement matters more than the shape of the test. Untested at runtime: this policy is not the default and is not what my build selects. Compiled against libnl-3. Assisted-By: Claude Opus 5 Signed-off-by: Christian Glombek <c.glombek@cosa.systems>

@PARAM

…s, commands, and events to reduce Flash footprint (#73545) * [app] Monomorphic forwarding in DataModel::Encode and Decode for enum types Update DataModel::Encode and DataModel::Decode for C++ enum types (std::is_enum) to forward directly to underlying integer types (std::underlying_type_t<X> / to_underlying(x)) rather than instantiating TLVWriter::Put<EnumType> and TLVReader::Get<EnumType> templates per enum type. This eliminates redundant per-enum template instantiations across server cluster implementations, reducing Flash binary footprint. * [app] Monomorphic forwarding in DataModel::Encode and Decode for BitFlags types Update DataModel::Encode and DataModel::Decode for BitFlags<X> bitmap types to forward directly to raw integer storage (x.Raw() / x.SetRaw()) rather than instantiating TLVWriter::Put<BitFlags<X>> and TLVReader::Get<BitFlags<X>> templates per bitmap type. This eliminates redundant per-bitmap template instantiations, reducing Flash binary footprint. * [app] Monomorphize response command payload encoding using non-virtual callback descriptor Replace the templated virtual class EncodableResponseCommandPayload<CommandData> in CommandHandler with a lightweight, non-virtual EncodableResponsePayload descriptor struct containing a data pointer and a type-erased EncodeFn function pointer. Previously, every response command type in the SDK instantiated a distinct class subclassing DataModel::EncodableToTLV, generating virtual function tables (vtables), RTTI metadata, and virtual destructors for ~200 response command types across cluster implementations. Using EncodableResponsePayload: - Eliminates vtables and RTTI structures for command responses. - Reduces .rodata section footprint by ~1.02 KiB on nRF52840DK (all-clusters-app). - Reduces overall Flash binary footprint by -824 B. - Reduces symbol/string metadata by ~134 KiB. [app] Enhance doc comments detailing virtual destructor and vtable bloat elimination Add explicit header documentation explaining the design rationale behind replacing templated virtual class instantiations with POD descriptors and standalone function pointer callbacks: - Avoids compiler-generated virtual class constructors and destructors (both complete object and deleting destructors). - Eliminates per-type vtable arrays and RTTI/typeinfo metadata across command response and event payload instantiations. * [app] Monomorphize event payload writer using non-templated callback class Replace the templated virtual class SimpleEventPayloadWriter<T> in EventsGenerator.h with a single, non-templated SimpleEventPayloadWriter class that takes a data pointer and a type-erased EncodeFn function pointer callback. Previously, every event type T across all clusters instantiated a distinct class subclassing EventLoggingDelegate, generating virtual function tables (vtables), RTTI metadata, and virtual destructors for dozens of event payload types. Using the non-templated callback wrapper: - Reduces .rodata section footprint by -508 bytes on nRF52840DK (all-clusters-app). - Reduces overall Flash binary footprint by -340 bytes. - Reduces symbol/string metadata by -13.4 KiB. * [app] Add null pointer validation checks for callback descriptors and type_traits header - Add VerifyOrReturnError null check for function pointer callbacks in EncodableResponsePayload::EncodeTo and SimpleEventPayloadWriter::WriteEvent to prevent null pointer dereferences. - Add explicit #include <type_traits> in Decode.h to ensure header self-sufficiency. * [app] Document non-null requirement for encoding callbacks instead of runtime null checks - Remove redundant runtime null checks on mEncodeFn in EncodableResponsePayload::EncodeTo and SimpleEventPayloadWriter::WriteEvent. - Add Doxygen notes documenting that function pointer callbacks must be non-null and are guaranteed non-null when constructed via internal inline helpers (EncodeTypedCommandPayload / EncodeTypedEventPayload). * [app] Remove static internal linkage from EncodeTypedEventPayload function template - Remove static keyword from free-standing function template EncodeTypedEventPayload in EventsGenerator.h to allow COMDAT template deduplication across translation units by the linker. * [app] Add error logging parity to CommandHandlerImpl::AddResponse for EncodableResponsePayload - Log CHIP_ERROR format string on encoding failure in EncodableResponsePayload overload, matching established behavior of EncodableToTLV overload. * [app] Deduplicate TryAddResponseData and MockCommandHandler::AddResponseData overloads - Delegate CommandHandlerImpl::TryAddResponseData and MockCommandHandler::AddResponseData for EncodableResponsePayload to their respective EncodableToTLV overloads using PayloadAdapter. - Eliminates ~50 lines of duplicate framing and session setup logic across production and test double code. * [pre-commit.ci] auto fixes from pre-commit.com hooks for more information, see https://pre-commit.ci * [app] Fix Doxygen comment placement for AddResponseData and EncodableResponsePayload - Relocate AddResponseData Doxygen documentation block to directly precede AddResponseData(..., EncodableToTLV &) declaration. - Prevents Doxygen parser from misassociating AddResponseData @PARAM documentation with EncodableResponsePayload struct. * [app] Restore compile-time abstract class protection for EncodableToTLV - Restore pure virtual (= 0) declaration for EncodeTo(TLV::TLVWriter &, TLV::Tag) in EncodableToTLV. - Default FabricAwareTLVWriter overload delegates to mTLVWriter. - Updates class Doxygen documentation to clearly state abstract interface contract. * [app] Consolidate duplicate PayloadAdapter into nested EncodableResponsePayload::Adapter - Consolidate local PayloadAdapter struct definitions into a single nested EncodableResponsePayload::Adapter struct in CommandHandler.h. - Simplifies call sites across CommandHandler.h, CommandHandlerImpl.cpp, and MockCommandHandler.cpp. - Reduces code repetition and maintains clean API boundaries. * [pre-commit.ci] auto fixes from pre-commit.com hooks for more information, see https://pre-commit.ci * [app] Add Doxygen non-null precondition note to EncodeTypedEventPayload - Document non-null precondition for data parameter in EncodeTypedEventPayload. - Explain rationale for omitting redundant runtime null checks to prevent per-event-type template Flash growth. * [app/tests] Add unit test for EncodableResponsePayload::Adapter - Add TestEncodableResponsePayloadAdapter to TestCommandInteraction.cpp. - Verifies encoding via both FabricAwareTLVWriter and standard TLVWriter overloads. * [app] Return CHIP_ERROR_INCORRECT_STATE for plain TLVWriter in EncodableResponsePayload::Adapter - Return CHIP_ERROR_INCORRECT_STATE for non-fabric-aware TLVWriter overload in EncodableResponsePayload::Adapter. - Prevents accidental encoding of server command responses with an undefined fabric index (kUndefinedFabricIndex) which could mis-encode fabric-scoped response fields. - Updates unit test in TestCommandInteraction.cpp to assert CHIP_ERROR_INCORRECT_STATE. * [pre-commit.ci] auto fixes from pre-commit.com hooks for more information, see https://pre-commit.ci --------- Co-authored-by: pre-commit-ci[bot] <66853113+pre-commit-ci[bot]@users.noreply.github.com>

* [docker] Build the stage images in CI again The stage-1, stage-2 and stage-3 jobs have been disabled with '&& false' since 2023, on the grounds that images either exceed the runner's disk or fail to resolve their parent. Both halves are now addressable. A per-image probe measured what each image actually costs when it is the only thing on a runner. Of 30 images, 26 build inside the 14GB GitHub guarantees, most using under 60% of it. Only efr32 (17.8GB) and nxp (14.6GB) exceed it, and those are commented out individually with their measured cost, matching how imx was already handled. imx is re-enabled. It was excluded for 'no space left on device' and peaks at 8.9GB on its own runner, so the original failure was an artifact of sharing a runner rather than a property of the image. vscode stays disabled. It does not merely fail: on a dedicated runner with 88GB free it killed the runner after ~12 minutes, leaving no logs. Also give the newly enabled jobs the same permissions as the base job, and correct the vscode build step, which pointed at a vscode-3 directory that does not exist. * [docker] Run this workflow when this workflow changes The trigger watched integrations/docker/** but not the workflow file, so a change to which images CI builds could not demonstrate that they build. This PR is the case in point: it enables three stages and triggered none of them. * [docker] Run the image builds on namespace runners A GitHub-hosted runner guarantees 14GB of free space and frequently supplies far more, so an image needing more than the guarantee does not fail cleanly. It fails according to which runner it draws, which is a flake rather than a result. Two images measured above that line, efr32 at 17.8GB and nxp at 14.6GB. A namespace shape has a fixed disk size, so that becomes something to size for rather than to babysit. With the runner changed, efr32 and nxp no longer need excluding: they were only ever over GitHub's guarantee, not over any inherent limit, so both are enabled here. vscode remains disabled. It killed a runner outright with 88GB free, so it needs its own investigation rather than a larger shape. * [docker] Log in to ghcr.io before pushing from the stage jobs Only the base job authenticated. The stage and vscode jobs call build.sh --latest --push, which runs docker push directly, so a dispatch with push_images=true would fail on authentication. The path was unreachable while those jobs were disabled and becomes reachable here. Gated on the same condition as the push step it precedes, so pull requests and ordinary pushes still build without authenticating. * [docker] Add a concurrency group so superseded runs are cancelled Without one, every push starts another full build of every image in parallel with the runs it supersedes. This PR demonstrated it: two runs, roughly twenty jobs each, were in flight at once with only the newer one of any use. Uses the same group expression as the majority of workflows here. * [docker] Correct the vscode note and drop historical comments The note claimed nothing depends on chip-build-vscode. That is wrong: integrations/cloudbuild/ runs its steps in that image, and maintainers use it locally in place of one image per platform. Replaced with what the image is for, how to refresh it by hand, and a TODO for giving it a shape large enough to build on. Also rephrased the remaining comments against the current state rather than how things used to be, and dropped ones that only described how the figures were obtained. --------- Co-authored-by: Andrei Litvin <andy314@gmail.com>

…it (Auto-merged by platform-bot) * [NXP] Upate non-EL2GO factorydataImpl to import the DAC key during init Signed-off-by: Martin Girardot <martin.girardot@nxp.com> * [pre-commit.ci] auto fixes from pre-commit.com hooks for more information, see https://pre-commit.ci --------- Signed-off-by: Martin Girardot <martin.girardot@nxp.com> Co-authored-by: pre-commit-ci[bot] <66853113+pre-commit-ci[bot]@users.noreply.github.com>

… codes (#73461) * eval setup code * address pr comment

* [docker] Do not rebuild an image that already exists These images fetch SDKs and toolchains from vendor servers while building, and those fetches fail for reasons this repository does not control. A build that does not happen cannot fail that way, so skip it when the image already exists at this version, locally or published. When a build is needed, seed the layer cache from the previous version, since usually only one step has changed and the layers before it are still valid. Record inline cache metadata so that published images can serve as a cache source: existing images do not carry it, so this takes effect from the next published version onward. --no-cache still forces a rebuild, and a missing previous version is not an error. * [docker] Read the version in base ten when deriving the previous tag Bash treats a leading zero as octal, so a version of 08 or 09 aborted the build outright with 'value too great for base', and 010 would quietly have looked for tag 8 rather than 9. Force base ten, and skip the predecessor lookup at version zero, which has none.

An image is published under the tag in its version file, and a build is skipped when that tag already exists, so changing an image without bumping the version produces a build that never runs and a green result that proves nothing. This is the first item on #30750, which asked for a robot check rather than relying on reviewers spotting it. Co-authored-by: Andrei Litvin <andy314@gmail.com>

Signed-off-by: Zibran Khan <kali834x@gmail.com>

* [docker] Do not re-push an image the registry already serves The publish step pushed the version tag on every run, including runs that built nothing because the image was already present. Those pushes upload nothing, and on a runner whose image cache persists they happen on every merge. Decide from the registry rather than from whether a build happened. A skipped build does not imply the image is published: a cached image can be left over from an earlier run whose push failed, so inferring "published" from the absence of a build would stop publishing silently while the job stayed green. latest is compared by digest instead of existence, since the version tag can be published while latest still points at an older one. An unknown tag, or no buildx, leaves the answer empty and the push proceeds, so failing to reach the registry errs towards publishing. This also puts the BUILT flag added in #73537 to use; it was set but never read. * Restyled by shellharden

…by platform-bot) * [NXP] Implement interface accessor helpers Make use of interface accessor helpers to set and return OpenThread and external interfaces when lwIP is used. Signed-off-by: Cristian Bulacu <cristian.bulacu@nxp.com> * [NXP] Improve portability of DnssdImplBr implementation Refactor the NXP Border Router DNS-SD implementation to remove direct dependencies on LwIP-specific APIs and types, improving portability across platforms and networking stacks. Replace the previous LwIP-specific handling with the platform-agnostic Inet interface abstraction, so network interfaces are obtained through generic accessors rather than LwIP netif lookups. Track outstanding browse and resolve operations using the SDK's IntrusiveList, removing custom list bookkeeping and reducing coupling to platform-specific data structures. No functional change to DNS-SD behavior is intended; this is purely a portability and maintainability improvement. --------- Signed-off-by: Cristian Bulacu <cristian.bulacu@nxp.com>

…rm-bot) * implemented sysbuild * [pre-commit.ci] auto fixes from pre-commit.com hooks for more information, see https://pre-commit.ci * typo * Gemini suggestion * Code rabbit suggestions --------- Co-authored-by: pre-commit-ci[bot] <66853113+pre-commit-ci[bot]@users.noreply.github.com>

…#73587) The buildx in CI ignores --format and prints the whole human readable record, so remote_digest returned that record rather than a digest. It still compares as non-empty, so skipping the version tag kept working, but comparing two of them never matches because each names its own tag, and latest was therefore pushed on every run. Parse the Digest line, which the unformatted output has carried across versions, so the comparison works whatever buildx is present.

* Upload chip-cert and chip-tool for x64 and arm64 as artifacts * Restrict artifacts uploads only for master pushes * Set retention days to 7 days

…-bot) (#73754) * [Telink] Add multi-endpoint Identify support * [pre-commit.ci] auto fixes from pre-commit.com hooks for more information, see https://pre-commit.ci * [Telink] Fix build docs * [Telink] Refine multi-endpoint Identify configuration --------- Co-authored-by: pre-commit-ci[bot] <66853113+pre-commit-ci[bot]@users.noreply.github.com>

…ter for 1.7 release (#73183) * Add sensor fusion supported attribute for 1.7 release Signed-off-by: Lo,Chin-Ran <chin-ran.lo@nxp.com> * [pre-commit.ci] auto fixes from pre-commit.com hooks for more information, see https://pre-commit.ci * Fix by following Gemini's review suggestions Signed-off-by: Lo,Chin-Ran <chin-ran.lo@nxp.com> * Fix the build error on Darwin Signed-off-by: Lo,Chin-Ran <chin-ran.lo@nxp.com> * Fix the build error on Darwin Signed-off-by: Lo,Chin-Ran <chin-ran.lo@nxp.com> * Add the boundary check in the unit-test module Signed-off-by: Lo,Chin-Ran <chin-ran.lo@nxp.com> * [pre-commit.ci] auto fixes from pre-commit.com hooks for more information, see https://pre-commit.ci * Refine the implementation Signed-off-by: Lo,Chin-Ran <chin-ran.lo@nxp.com> * [pre-commit.ci] auto fixes from pre-commit.com hooks for more information, see https://pre-commit.ci * Fix the reviewed problems: * Accept the empty input FusionSupported string to clear the saved data * Add more checking on the input data * Return null attribute if the delegate function has not been set Signed-off-by: Lo,Chin-Ran <chin-ran.lo@nxp.com> * [pre-commit.ci] auto fixes from pre-commit.com hooks for more information, see https://pre-commit.ci * Add to check the type before really convert the input data Signed-off-by: Lo,Chin-Ran <chin-ran.lo@nxp.com> --------- Signed-off-by: Lo,Chin-Ran <chin-ran.lo@nxp.com> Co-authored-by: pre-commit-ci[bot] <66853113+pre-commit-ci[bot]@users.noreply.github.com> Co-authored-by: Amine Alami <43780877+Alami-Amine@users.noreply.github.com>

* Initial code migration * Add command lists * Add command handlers * Add Codegen files * Regenerate ZAP * Remove function from util * Restyled by whitespace * Restyled by prettier-json * Fix clang errors on Darwin * Add small changes for PR comments * Add more PR fixes * add operator for struct * Use AttributeSet for OptionalCommands * Add PR comment changes * Add Config changes * Remove old implementation file and fix compilation error --------- Co-authored-by: Restyled.io <commits@restyled.io>

…spec conformance (#73663) * [AmbientContextSensing] Fix ObjectCountConfig label not being copied SetObjectCountConfig retained countingObject.label as a CharSpan decoded from the write request payload. TLVReader::Get(CharSpan &) does not copy, so the span referenced a PacketBuffer released when the write transaction completed, and later reads and reports re-encoded from it. Copy the label into a cluster-owned 64 byte buffer and re-point the stored span at it, following the PresetStructWithOwnedMembers convention. Apply the two spec constraints on SemanticTagStruct.Label that were missing: a max length of 64 characters, and presence when MfgCode is not null. Both return ConstraintError. Include the label in change detection so a write that only changes the label is stored and reported. * [AmbientContextSensing] Include mfgCode in ObjectCountConfig change detection The change comparison omitted countingObject.mfgCode, so a write that only changed the manufacturer code returned success without storing or reporting the new value. The same tag under a different manufacturer namespace denotes a different object, so that write has to take effect. Compare the whole SemanticTagStruct through a single IsSemanticTagEqual helper rather than listing fields at the call site, so a later field cannot be omitted the same way. * [pre-commit.ci] auto fixes from pre-commit.com hooks for more information, see https://pre-commit.ci --------- Co-authored-by: pre-commit-ci[bot] <66853113+pre-commit-ci[bot]@users.noreply.github.com>

…merged by platform-bot) (#73800)

@pics

* [tests] Add TC_IDM_1_5 python test for DelayReportData invoke action * [pre-commit.ci] auto fixes from pre-commit.com hooks for more information, see https://pre-commit.ci * [tests] Update TC_IDM_1_5 to use ArmFailSafe on Endpoint 0 and add subscription settle time check * [controller][python] Add DelayReportData support and TC_IDM_1_5 test script * [tests] Update TC_IDM_1_5 to match updated test plan with t_invoke_sent and 3600s MaxInterval * [pre-commit.ci] auto fixes from pre-commit.com hooks for more information, see https://pre-commit.ci * [tests][python] Complete type annotations and docstrings in TC_IDM_1_5 and fix SendBatchCommands delayReportData * [pre-commit.ci] auto fixes from pre-commit.com hooks for more information, see https://pre-commit.ci * [controller][python] Add uint16 validation to DelayReportData and cleanup TC_IDM_1_5 in finally block * [python-testing] Update TC_IDM_1_5 to check EP0 ICD cluster, verify MaxInterval, and calculate window from prime report - Check EP0 Descriptor ServerList for IcdManagement directly and skip if present - Remove unnecessary 1-second settling sleep after subscription establishment - Read and verify negotiated MaxInterval in SubscribeResponseMessage equals requested 3600s - Compute subscription remaining window relative to priming report receipt timestamp (t_prime) * [python-testing] Refactor TC_IDM_1_5 timeouts and step 5 teardown error handling - Calculate subscription headroom in Step 2 with 2x MRP retransmission timeout - Use 1x MRP timeout for report reception queue timeout in Step 4 - Execute Step 5 disarm in normal test flow with assertions - Use finally block purely for defensive teardown without masking earlier step failures * [python-testing] Use @pics decorator and inline step definitions in TC_IDM_1_5 - Add @pics("MCORE.IDM.S") decorator - Remove boilerplate desc_TC_IDM_1_5, pics_TC_IDM_1_5, and steps_TC_IDM_1_5 methods - Pass step descriptions inline directly to self.step() calls --------- Co-authored-by: pre-commit-ci[bot] <66853113+pre-commit-ci[bot]@users.noreply.github.com>

* Migrate ColorControl * [pre-commit.ci] auto fixes from pre-commit.com hooks for more information, see https://pre-commit.ci * Add zap files * Restyled by whitespace * Restyled by prettier-json * Fix linter * Sync zcl.json and zcl-with-test-extensions.json AAI attributes Mirror the Color Control and Commissioning Proxy attributeAccessInterfaceAttributes entries into both files so check_zcl_file_sync.py passes. Co-Authored-By: Claude Opus 4.8 <noreply@anthropic.com> * Fix zap template CI * Regenerate zap callback.h to drop stale DynamicLighting callbacks Co-Authored-By: Claude Opus 4.8 <noreply@anthropic.com> * Fix implicit conversion error * [pre-commit.ci] auto fixes from pre-commit.com hooks for more information, see https://pre-commit.ci * Fix clang-tidy errors * Update example app get/set calls * [pre-commit.ci] auto fixes from pre-commit.com hooks for more information, see https://pre-commit.ci * Update Get/Set calls inside namespace * [pre-commit.ci] auto fixes from pre-commit.com hooks for more information, see https://pre-commit.ci * Fix chef undefied errors * Fix timer destruction;delete old files * Wire default values * Add symetric shutdown * Add SceneHandler wiring * Fix continous vs immidiate hue movement * Fix clang-tidy;Name changes * [pre-commit.ci] auto fixes from pre-commit.com hooks for more information, see https://pre-commit.ci * Make delegate reassignable * [pre-commit.ci] auto fixes from pre-commit.com hooks for more information, see https://pre-commit.ci * Minor tweaks * Remove unused attribute * Push regenerated matter files * Replace std::visits * USe emeber signature accsessors * [pre-commit.ci] auto fixes from pre-commit.com hooks for more information, see https://pre-commit.ci * Update json files * Remove unserved attributes * Abandon backwards compat * [pre-commit.ci] auto fixes from pre-commit.com hooks for more information, see https://pre-commit.ci * Revert old usage * Regen matter files * Regen json * Regen config files * Resolve comments * [pre-commit.ci] auto fixes from pre-commit.com hooks for more information, see https://pre-commit.ci * Fix comments * Resolve comments * Fix minor inconsistencies * [pre-commit.ci] auto fixes from pre-commit.com hooks for more information, see https://pre-commit.ci * Name/Comment changes;Signature fixes * [pre-commit.ci] auto fixes from pre-commit.com hooks for more information, see https://pre-commit.ci * Fix name * Case changes;Comment tweaks;Conditon changes * [pre-commit.ci] auto fixes from pre-commit.com hooks for more information, see https://pre-commit.ci * Fix conversion error * [pre-commit.ci] auto fixes from pre-commit.com hooks for more information, see https://pre-commit.ci * Resolve RabbitAI comments * [pre-commit.ci] auto fixes from pre-commit.com hooks for more information, see https://pre-commit.ci * Fix clamps * Fix clang-tidy error * [pre-commit.ci] auto fixes from pre-commit.com hooks for more information, see https://pre-commit.ci * Switch to pos aggregates * Remove scene invalidation logic * [pre-commit.ci] auto fixes from pre-commit.com hooks for more information, see https://pre-commit.ci * Fix tests * Fix comments;Update Delegate * [pre-commit.ci] auto fixes from pre-commit.com hooks for more information, see https://pre-commit.ci * Use helpers --------- Co-authored-by: pre-commit-ci[bot] <66853113+pre-commit-ci[bot]@users.noreply.github.com> Co-authored-by: Restyled.io <commits@restyled.io> Co-authored-by: Claude Opus 4.8 <noreply@anthropic.com>

@pics

…tion Alarm Cluster (#72279) * Add TC-EPALM-2.1 Python test for Electrical Protection Alarm cluster Adds src/python_testing/TC_EPALM_2_1.py, the attribute test for the Electrical Protection Alarm cluster (EPALM, 0x00A3). Registered as not_automated in test_metadata.yaml until an example app integrates the EPALM server. Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com> * TC-EPALM-2.1: skip Latch step when the attribute is absent EPALM disallows the RESET feature, so Latch (RESET-gated in Alarm Base) is not part of the cluster and is absent from the generated attribute bindings. The step referenced attributes.Latch unconditionally, raising AttributeError against a conformant EPALM server. Guard the access with hasattr so the optional step is skipped when Latch is not present. Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com> * TC-EPALM-2.1: follow the MaximumDischargeCurrent spelling fix Spec PR CHIP-Specifications/connectedhomeip-spec#13304 corrected the misspelling on SurgeProtectionRatingsStruct field 7, and #73621 carries that through the SDK XML and the generated bindings. This updates the matching attribute references so the script agrees with the generated Python once that lands. The script is listed in test_metadata.yaml and is not run in CI, so this does not depend on merge order. * TC-EPALM-2.1: align the script step list with the test plan The script read the same attributes as the plan but in a different order, so step n on each side described a different thing throughout. It also omitted two steps the plan has, and performed one the plan did not. Reorders to the plan sequence: FeatureMap, then the Alarm Base attributes, then the seven rating attributes, then the write-rejection probe, then the subset invariant. Adds the FeatureMap read and the ArcCause write-rejection probe. The subset check the script already performed is now step 15 in the plan as well. * [TC-EPALM-2.1] Correct the not-automated reason The entry claimed the script self-skips via run_if_endpoint_matches(has_cluster(...)). It does not: gating is @pics('EPALM.S') plus @async_test_body, which declares a top-level PICS for the runner to check rather than performing a runtime endpoint match. Restates the reason as the actual situation, matching the wording already used for TC_EPALM_2_3.py. * Point TC-EPALM-2.1 at electrical-protection-app The EPALM server is being demonstrated on the Electrical Distribution Enclosure endpoint of electrical-protection-app rather than on all-clusters-app. The endpoint number is unchanged, so only the app target moves. Co-Authored-By: Claude Opus 5 (1M context) <noreply@anthropic.com> * Enable TC-EPALM-2.1 in CI The EPALM server is now on master on the enclosure endpoint of electrical-protection-app (#73395), which this script's CI directive targets, so the exclusion is no longer needed. Co-Authored-By: Claude Opus 5 (1M context) <noreply@anthropic.com> * Bound the EPALM measurement fields at the data model maximum The amperage-mA and voltage-mV fields in this cluster are declared min 1, max 0x3FFFFFFFFFFFFFFF. matter_asserts.assert_valid_int64 permits values through 0x7FFFFFFFFFFFFFFF, so a DUT reporting anything in between passed while violating the data model. Adds the upper bound to all 21 int64 measurement checks. The two uint64 fields, ResponseTime and EnergyAbsorptionCapability, declare a minimum and no maximum, so they are left alone. Reported by coderabbitai. Co-Authored-By: Claude Opus 5 (1M context) <noreply@anthropic.com> * Assert the declared minimum on the EPALM bitmap fields Five bitmap fields are declared min 1 in the cluster XML: SupportedArcCauses, and TripMechanism on the OverLoad, OverVoltage, ResidualCurrentFault and SurgeProtection ratings structs. The script checked only their upper bound, so a DUT reporting 0 passed. ShortCircuitRatingsStruct.TripMechanism is deliberately not included: it declares a maximum and no minimum, unlike the other four TripMechanism fields. That asymmetry looks unintentional in the data model but is not this script's to resolve. Reported by coderabbitai. Co-Authored-By: Claude Opus 5 (1M context) <noreply@anthropic.com> * ci: retrigger after a cancelled Chef job The Chef - Linux CI Examples job on the previous run was cancelled at the two-hour mark rather than failing. The same job passes on the sibling PRs 72323, 73730, 73731 and 73732, so this is a retrigger rather than a fix. * Hoist ALARM_BITMAP_MAX to module scope ALARM_BITMAP_MAX was declared twice inside test_TC_EPALM_2_1 with the same value. The second assignment was dead: all four uses precede it. Moves the single declaration to module scope alongside MEASUREMENT_MAX, the other data-model-derived bound in this file. Reported by hasty. Co-Authored-By: Claude Opus 5 (1M context) <noreply@anthropic.com> * ci: retrigger after an unrelated TC_JFADMIN_2_2 flake The TC_[G-M] REPL shard failed on TC_JFADMIN_2_2, which timed out after 60s waiting for '[CTL] Commissioning complete'. This branch touches only TC_EPALM_2_1.py, and TC-EPALM runs in the TC_[A-F] shard, so the failing shard does not execute anything this PR changes. Master's own TC_[G-M] shard failed earlier today on a different test (TC_IDM_9_1, run 33168611711), so the instability is in the shard rather than in this change. Co-Authored-By: Claude Opus 5 (1M context) <noreply@anthropic.com> * Address review: expectation clauses, shared validators, and three real gaps Docstring now uses the test plan's Purpose text. Dropped the test-plan URL, since that repo is private and CSA members should be reading the rendered plan, and dropped the step-numbering note. Every step carries the plan's Expected Outcome. The framework parameter is expectation, not expected, so the suggested edit as written raises TypeError. Repeated validation is now shared. _check_measurement replaces twenty-one open-coded int64 min-1 and MEASUREMENT_MAX checks; _check_bitmap replaces the hand-rolled reserved-bit comparisons; _check_current_trip_mechanism, _check_voltage_trip_mechanism and _check_trip_curve are shared by the structs that use those types. _check_choice_group enforces that at least one field of each O.<group>+ struct is present. The six struct validators and the new helpers are module-level functions rather than methods, since none of them touch self. Three gaps this closes: - FeatureMap bit 0 must be clear. RESET is disallowConform on EPALM, and no other test checks it. IDM-10.5 would, but only once a 1.7 data-model XML exists; the prebuilt directories stop at 1.6.1 and carry no EPALM. - State must be 0 in the no-fault baseline the test setup establishes. - Latch must be absent from AttributeList. Latch is RESET-conditional and EPALM disallows RESET, so it is disallowed rather than optional. The old step was dead code: it guarded on an attribute the bindings never generate. Two checks removed as unreachable rather than merely redundant. The FeatureMap uint32 type check was the wrong helper and duplicated IDM. The NullValue guard on TrippingCharacteristic could never fire: the field is optional and not nullable, so only None is reachable. Step 14, the ArcCause write-reject, is dropped per the test plan review, and the subset-invariant step becomes 14. Reported by jamesharrow. Co-Authored-By: Claude Opus 5 (1M context) <noreply@anthropic.com> * ci: retrigger after an unrelated TC_SC_4_3 flake The TC_[N-S] REPL shard failed on TC_SC_4_3.py, a Secure Channel discovery test, with a pairing-delegate error out of ChipDeviceController-ScriptDevicePairingDelegate.cpp. That shard does not run any EPALM script, and this branch changes only TC_EPALM_2_1.py, which runs in the TC_[A-F] shard. Co-Authored-By: Claude Opus 5 (1M context) <noreply@anthropic.com> * Drop the RESET bit check from step 2 Per review: assume the 1.7 data-model XML will be in place, so the generic conformance test covers a disallowed feature bit, and avoid reworking the script between TE2 and SVE. The FeatureMap read stays because the plan step calls for it, but nothing gates on the value now, so it is logged rather than assigned. Reported by jamesharrow. Co-Authored-By: Claude Opus 5 (1M context) <noreply@anthropic.com> --------- Co-authored-by: Claude Opus 4.8 (1M context) <noreply@anthropic.com>

@jamesharrow

…3730) * Populate the EPALM rating attributes in electrical-protection-app Every rating attribute was published as null, so the attribute tests took the null path and their structure validation never ran. Publishes representative fixed values for a 100 A / 240 V residential panel breaker across all seven rating attributes. Currents are amperage-mA and voltages are voltage-mV per the cluster definition, ResponseTime is in nanoseconds, and every field satisfies its min 1 constraint. These are illustrative, not measurements: the app performs none. Requested by @jamesharrow on #73395, so that the Python scripts exercise the detailed structure validation rather than confirming a null. Co-Authored-By: Claude Opus 5 (1M context) <noreply@anthropic.com> * [pre-commit.ci] auto fixes from pre-commit.com hooks for more information, see https://pre-commit.ci --------- Co-authored-by: Claude Opus 5 (1M context) <noreply@anthropic.com> Co-authored-by: pre-commit-ci[bot] <66853113+pre-commit-ci[bot]@users.noreply.github.com>

@pics

PICS.yaml carries PWRTL.S and F00-F03, A0000 and A0001, but nothing for the ElectricalCircuit feature (bit 4) or the ElectricalCircuitNodes attribute (0x0002), both added to the cluster for 1.7. check_pics reads a missing key as False, so a test gated on PWRTL.S.F04 does not error, it silently never runs. TC_PWRTL_2_2.py already carries @pics('PWRTL.S', 'PWRTL.S.F04') and is affected. Both default to 0 in ci-pics-values, matching the CI apps, which do not enable the feature. Co-authored-by: Claude Opus 5 (1M context) <noreply@anthropic.com>

…tion Alarm Cluster (FeatureMap and Supported consistency) (#72323) * Add TC-EPALM-2.3 Python test for Electrical Protection Alarm cluster Adds src/python_testing/TC_EPALM_2_3.py verifying FeatureMap and Supported attribute consistency for the Electrical Protection Alarm cluster (EPALM, 0x00A3). Registered as not_automated in test_metadata.yaml until an example app integrates the EPALM server. Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com> * TC-EPALM-2.3: define explicit steps_ method for dynamic step numbers The test builds steps 4-10 in a loop with a dynamic self.step(step_num). The step extractor requires constant step numbers when steps are inferred from the test body, and raised ValueError at setup. Add an explicit steps_TC_EPALM_2_3() method enumerating all 11 steps so the extractor no longer parses the dynamic calls; the body now advances with self.step(n). Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com> * [TC-EPALM-2.3] Correct the copyright year and the not-automated reason The copyright header said 2025 while the sibling scripts say 2026, and the not-automated reason named the electrical-protection-app as the unblocker. The script's CI directive targets all-clusters-app on endpoint 1, which is what #73395 provides, so name that instead. * [TC-EPALM-2.3] Make the orphan-bit check effective and skip absent-feature steps Two defects in TC_EPALM_2_3.py. Plan clause 11(a) requires that no bits be set in Supported outside positions 0..6. AlarmBitmap is an IntFlag whose members cover exactly those seven bits, so `~all_alarm_bits` complements only within the class mask and evaluates to 0. `orphan_bits` was therefore unconditionally 0 and the assertion could never fail: a DUT reporting Supported=0x80 passed it. Casting both operands to int restores the full-width complement. Steps 4-10 are specified as skipped when the corresponding PICS feature is false, but the script only logged and left the step recorded as PASS. Adds mark_current_step_skipped() so the certification report matches the plan. * Point TC-EPALM-2.3 at electrical-protection-app The EPALM server is being demonstrated on the Electrical Distribution Enclosure endpoint of electrical-protection-app rather than on all-clusters-app. The endpoint number is unchanged, so only the app target moves. Co-Authored-By: Claude Opus 5 (1M context) <noreply@anthropic.com> * Enable TC-EPALM-2.3 in CI The EPALM server is now on master on the enclosure endpoint of electrical-protection-app (#73395), which this script's CI directive targets, so the exclusion is no longer needed. Co-Authored-By: Claude Opus 5 (1M context) <noreply@anthropic.com> * Inline the step descriptions and drop steps_TC_EPALM_2_3 Addresses review on #72323: - steps_TC_EPALM_2_3 is gone; each self.step() carries its own description and an expectation taken from the TC-EPALM-2.3 plan with the adoc macros expanded. - self.step(1) is no longer bare. - The steps 4-10 loop is unrolled, because self.step(step_num) is not a literal and the AST step extractor cannot read it once steps_ is removed. * [pre-commit.ci] auto fixes from pre-commit.com hooks for more information, see https://pre-commit.ci --------- Co-authored-by: Claude Opus 4.8 (1M context) <noreply@anthropic.com> Co-authored-by: pre-commit-ci[bot] <66853113+pre-commit-ci[bot]@users.noreply.github.com>

…rcuitNodes attribute (#73402) * [PowerTopology] Add ElectricalCircuit (CIRC) feature + ElectricalCircuitNodes Add the Matter 1.7 ElectricalCircuit (CIRC) feature to the code-driven power-topology-server, exposing the ElectricalCircuitNodes attribute (0x0002): a fabric-scoped, writable, non-volatile List<CircuitNodeStruct>. - Read is fabric-filtered automatically (fabric-scoped item type). - Write is fabric-scoped: the accessing fabric is authoritative, ReplaceAll replaces only that fabric's slice with full staging/validation before any mutation, AppendItem appends one node. - Non-volatile persistence via an explicit TLV blob (retaining fabricIndex, which the generated fabric-scoped write codec omits), restored on Startup. - Fabric removal purges the removed fabric's nodes via FabricTable::Delegate; the FabricTable is an optional nullable field on the cluster Config. Enable CIRC on the EVSE example app and thread the FabricTable through PtConfig so fabric-removal cleanup is active. The added FabricTable parameters are defaulted, so the water-heater app (which shares ElectricalSensorManager) is unaffected. Extend TC_PWRTL_2_1 to assert ElectricalCircuitNodes present iff CIRC and to read, write (round-trip), and verify Non-volatile persistence of the attribute. Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com> * [pre-commit.ci] auto fixes from pre-commit.com hooks for more information, see https://pre-commit.ci * [PowerTopology] Scope TC_PWRTL_2_1 CIRC change to attribute presence Keep only the AttributeList presence-iff-CIRC assertion for ElectricalCircuitNodes in TC_PWRTL_2_1 (the attributes test). Functional read/write coverage of the attribute belongs to the dedicated TC_PWRTL_2_2, which this PR's CIRC-enabled DUT unblocks; avoid duplicating it here. Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com> * [PowerTopology] Heap-allocate ElectricalCircuitNodes write staging buffer WriteElectricalCircuitNodes staged the incoming list in a stack array of kMaxCircuitNodes StoredCircuitNode entries (~8 KB), which can overflow the constrained Matter thread stack on embedded platforms. Allocate the staging buffer on the heap via ScopedMemoryBuffer, sized to the incoming count. Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com> * [PowerTopology] Heap-allocate ElectricalCircuitNodes node storage The ElectricalCircuitNodes storage was an inline StoredCircuitNode array of kMaxCircuitNodes entries (~8 KB: char label[128] x 50). As a member this made every stack-allocated PowerTopologyCluster instance ~8 KB, exceeding the -Wstack-usage=8192 limit on 32-bit embedded builds (Zephyr/ESP32) in the unit tests, and inflating the cluster's footprint on constrained platforms generally. Store the nodes in a ScopedMemoryBuffer allocated on Startup, only when the CIRC feature is enabled. Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com> * [PowerTopology] Move ElectricalCircuitNodes storage behind an interface Review feedback: the cluster should not hard-code how ElectricalCircuitNodes is stored, and holding the entries in a heap buffer rules out platforms with stricter memory requirements. Following thread-network-directory-server, which solves the same problem for a writable persisted list: - CircuitNodeStorage is an abstract interface: Capacity(), Count(), CountForFabric(), GetNodeAtIndex(), ReplaceNodesForFabric() for the ReplaceAll write, AppendNode() for AppendItem, and RemoveNodesForFabric() for fabric purge. - DefaultCircuitNodeStorage implements it against the cluster's attribute storage, carrying over the previous TLV format unchanged, so existing behaviour (durable nodes, same location) is preserved with no application work. - The cluster holds only a CircuitNodeStorage pointer and allocates nothing for this attribute. StoredCircuitNode, the ScopedMemoryBuffer member, and SaveCircuitNodes()/LoadCircuitNodes() are gone. A platform without a heap now supplies a fixed-array implementation and reports its real limit through Capacity(). The unit tests do exactly that, so the allocation-free path is exercised rather than merely possible. CircuitNodeStorage::Init() is a no-op by default and exists so an implementation that persists through the cluster's attribute storage can obtain it at cluster Startup, which is the earliest point it exists. Implementations with their own persistence ignore it. PowerTopology::Instance gains an overload taking application-provided storage; the existing constructor keeps using DefaultCircuitNodeStorage, so applications need no change. Load() stops at Capacity() rather than erroring, so reducing a platform's capacity cannot prevent startup on previously persisted data. Testing: unit tests pass (22 cases, up from 21). The added DefaultCircuitNodeStoragePersistsAcrossInstances covers the TLV round trip that previously ran through the cluster, since the cluster-level tests now use the non-persisting fixed-array storage. all-clusters-app builds and links. Co-Authored-By: Claude Opus 5 (1M context) <noreply@anthropic.com> * [PowerTopology] Shrink the fixed-array test storage to fit the Zephyr stack budget The FixedCircuitNodeStorage test double held Node[50], roughly 8 KB, and the tests place it on the stack. That tripped -Werror=stack-usage=8192 on the Zephyr native_posix unit-test build (worst case 12112 bytes), breaking the nRF Connect SDK and ESP32_QEMU checks. Make the capacity a template parameter defaulting to 8 (about 1.3 KB), which brings the worst case to roughly 5.4 KB. No test stores more than four entries, and the resource-exhaustion test writes far past any capacity so it still exercises the limit. Fittingly, the double was failing the very constraint it exists to demonstrate. Testing: unit tests pass (22 cases). Co-Authored-By: Claude Opus 5 (1M context) <noreply@anthropic.com> * [PowerTopology] Require the fabric table for CIRC; bound the ReplaceAll staging write Two review findings on #73403, which carries this file until #73402 merges. 1. The ElectricalCircuit feature needs the fabric table, not "uses it if present". ElectricalCircuitNodes is fabric-scoped, so a removed fabric's entries must be purged, and that only happens through the OnFabricRemoved callback registered via AddFabricDelegate. Startup() previously skipped the registration silently when Config::fabricTable was null, and PowerTopology Instance defaults that parameter to nullptr, so an application could enable CIRC, compile cleanly, and leak a removed fabric's nodes with no diagnostic. Startup() now fails with CHIP_ERROR_INCORRECT_STATE and a log line, matching the Config::circuitNodeStorage check three lines above it. Returning an error is preferred over VerifyOrDie: Startup() already reports this class of misconfiguration through its return value, and a crash would be a harsher contract than the surrounding code uses. Shutdown() keeps its null check, since it can run on a cluster that never started or does not have CIRC. The unit tests construct CIRC clusters, so they now supply a fabric table via FabricTestFixture, as ElectricalCircuitNodesPurgedOnFabricRemovalTest already did. Added ElectricalCircuitWithoutFabricTableTest to pin the new contract. 2. WriteElectricalCircuitNodes staged the decoded ReplaceAll list into a buffer sized from list.ComputeSize() and indexed it with a counter driven by the iterator, without checking the two agree. Added a bounds check. Beyond decoder inconsistency this also covers newCount == 0, where the staging buffer is never allocated at all yet staging[0] would be written if the iterator yielded an element. Testing: TestPowerTopologyCluster 23/23 pass (was 22, plus the new one) and TestPowerTopologyClusterBackwardsCompatibility 1/1 passes, built natively with chip_build_tests=true. Confirmed the new guard is actually exercised rather than merely compiled: its error log appears exactly once across the suite, in the new test. clang-format clean. Co-Authored-By: Claude Opus 5 (1M context) <noreply@anthropic.com> * [PowerTopology] Discard a partially decoded ElectricalCircuitNodes value Load() appends entries as it decodes them, so a failure part way through a stored blob left the decoded prefix in place. Init() logs that error and returns CHIP_NO_ERROR, since a bad stored value must not brick startup, so the cluster went on to serve an arbitrary fraction of a corrupt list as the attribute value. Reset the count in that path so the list reads as empty instead. Also document the persisted blob's worst case size against the per-value limits some platforms impose, and move the anonymous namespace close below the final test so every fixture-derived test sits inside it. * TC-PWRTL-2.1: align the script step list with the test plan The script and the plan had drifted: the script combined the AttributeList read and its composition check into one step, wrote to only one of the two read-only attributes, and split the reboot and the persistence read across two steps. Step n in the script and step n in the plan stopped describing the same thing from step 5 onwards. Both are now 11 steps and match one for one. The reboot step also moves to request_device_reboot() with an app-ready-pattern directive, matching TC-PWRTL-2.2, instead of the is_pics_sdk_ci_only and wait_for_user_input dispatch, so it runs in CI rather than being skipped. * [TC-PWRTL-2.1] Make the reserved-bits check effective Step 4 is specified to validate the CIRC bit and reserved bits 5..31, but the reserved-bits half was inert. PowerTopology.Feature is an IntFlag whose members cover exactly bits 0..4, so `~KNOWN_BITS_MASK` complements only within the class mask and evaluates to 0. `reserved_bits` was therefore unconditionally 0 and a DUT setting any reserved bit passed. Casting both operands to int restores the full-width complement. * Drop the step-numbering remark from the TC-PWRTL-2.1 docstring It describes the file rather than the test, and goes stale the moment the plan renumbers. * Drop the unsupported-write step from TC-PWRTL-2.1 Per James Harrow: unsupported-write checks come out, since the IDM and low-level SDK tests already cover write access. Step 11 becomes step 10; the Status import goes with the last use. * Align TC-PWRTL-2.1 with its test plan Per James Harrow: align the script to the plan, and pre-existing lines are not exempt. The merged TC-PWRTL-2.1 plan has three steps; the script had eleven. It is now the plan's three, each carrying the plan's expected outcome as an expectation. Implementing the plan literally also adds an assertion the script never had: the plan requires both endpoint lists to hold no more than 20 entries, and nothing checked that. Removed: FeatureMap O.a conformance, the CIRC and reserved-bit checks, and the AttributeList composition checks, all covered by TC_IDM_10_2 per cecille, which is why the plan rework (chip-test-plans#6166) and its script alignment (#73732) were closed on 2026-08-28. Also removed the Non-Volatile persistence check, which the plan does not describe. * Drop the plan URL from the TC-PWRTL-2.1 docstring Only 16 of 501 scripts on master embed one, and a hardcoded branch-and-anchor link goes stale on any renumber. --------- Co-authored-by: Claude Opus 4.8 (1M context) <noreply@anthropic.com> Co-authored-by: pre-commit-ci[bot] <66853113+pre-commit-ci[bot]@users.noreply.github.com>

* Make Callback easier to use in non-list contexts Union mPrev/mNext with void* mContext{A,B} that callees can use to more easily manage individual callbacks without a CallbackDeque. A common use case is to hold a callee context pointer so the CancelFn can find the owning object. Move mNext = mPrev = this initialization into the sub-classes that inherit from Cancelable to use it as a list anchor (CallbackDeque, GroupedCallbackList), and leave the callee context fields uninitialized in the base class. Make Cancelable a protected instead of public base class of CallbackDeque, and tighten the output arguments of CallbackDeque::DequeueAll and DequeueBy to be a CallbackDeque rather than a base Cancelable. This ensures target objects retain their doubly-linked list initialization, and is also semantically the correct type: a CallbackDeque is exactly a Cancelable where mPrev/mNext form a doubly linked list. Also tweak DequeueAll() so it appends to a non-empty target list rather than leaving any existing elements in an orphaned state, and gracefully handle dequeuing into the same deque as a no-op. DequeueBy() into the source list makes no sense, so add a VerifyOrDie for completeness (the method is not used anywhere in the code base). Adjust OperationalSessionSetup::NotifyRetryHandlers accordingly: use CallbackDeques for the temporary lists and use IsEmpty() / First() methods instead of directly accessing mNext. Also add a test case to TestCHIPCallback that validates the logic used in NotifyRetryHandlers for dealing with persistent callbacks that can deallocate themselves. * Put back invalidation to make clang-tidy happy This could be changed to actually poison the pointers at some point. * Add / reword documentation * Fix comment grammar Co-authored-by: Copilot Autofix powered by AI <175728472+Copilot@users.noreply.github.com> --------- Co-authored-by: Copilot Autofix powered by AI <175728472+Copilot@users.noreply.github.com>

…otiate TCP capabilities (#71876) * Add TCP config fields to the Session parameters to allow nodes to negotiate TCP capabilities during session establishment. - Add `SupportedTransports` and `MaxTCPPayloadSize` to `SessionParameters` - Set local session parameters in `Server::Init` for `CASESessionManager` and `CASEServer`. - Encode and decode new parameters in `PairingSession` and `CASESession`. - Use `MaxTCPPayloadSize` in `SessionManager::PrepareMessage` for message length validation when large payloads are allowed. - Add unit tests in `TestCASESession` and `TestPairingSession` for the new parameters. * [pre-commit.ci] auto fixes from pre-commit.com hooks for more information, see https://pre-commit.ci * Refactor to address review comments. * Address review comments Refactor SessionParameters to reduce #if directives - Refactor `SessionParameters` into `BasicSessionParameters` and a derived `SessionParameters` class to minimize `#if INET_CONFIG_ENABLE_TCP_ENDPOINT` directives. - Remove `#if` guards around `SupportsLargePayload()` in `SecureSession.h` and in `SessionManager::PrepareMessage` by relying on default behaviors of `BasicSessionParameters`. - Fix a potential issue in `SessionManager::PrepareMessage` where a peer-advertised large payload size could override local limits now clamped to `min(maxPayload, kMaxLargeAppMessageLen)`. * [pre-commit.ci] auto fixes from pre-commit.com hooks for more information, see https://pre-commit.ci * Relocate to consolidate #if block. * Apply suggestions from code review Co-authored-by: Copilot Autofix powered by AI <175728472+Copilot@users.noreply.github.com> * Potential fix for pull request finding Co-authored-by: Copilot Autofix powered by AI <175728472+Copilot@users.noreply.github.com> * Adjust Max TCP Payload size by the framing header length of 4 bytes. * Add an enable TCP server flag on the controller side in the init params. * Use kTCPFramingHeaderSize instead of SizeOfMaxTCPPayloadSize. * Address more review comments * [pre-commit.ci] auto fixes from pre-commit.com hooks for more information, see https://pre-commit.ci * Address review comments. Consolidate BasicSessionParameters and SessionParameters into a single SessionParameters class, removing the conditional compilation based on INET_CONFIG_ENABLE_TCP_ENDPOINT. Increase CHIP_SYSTEM_CONFIG_MAX_LARGE_BUFFER_SIZE_BYTES to 64004 to keep the default max TCP payload size at 64000 (subtracting 4 bytes for TCP framing). * [pre-commit.ci] auto fixes from pre-commit.com hooks for more information, see https://pre-commit.ci * Remove INET_CONFIG_ENABLE_TCP_ENDPOINT conditional compile around SetLocalSessionParameters() in CASEServer.h. * Remove INET_CONFIG_ENABLE_TCP_ENDPOINT conditional compile around SetLocalSessionParameters() in CASEClient.h. * Remove #if INET_CONFIG_ENABLE_TCP_ENDPOINT in some of the Test Session logic. * Set local session parameters for CASEServer in controller factory Configure and set the local session parameters for both the CASE client and CASE server instances(server interactions enabled) initialized by the DeviceControllerFactory. Previously, the TCP parameters (supported transports and max payload size) were only applied to the CASE client (outgoing). Now, they are also applied to the CASEServer (incoming) when server interactions are enabled, ensuring consistent capability advertisement. * Address review comments. * [pre-commit.ci] auto fixes from pre-commit.com hooks for more information, see https://pre-commit.ci * Define `kLegacyDefaultMaxLargeAppMessageLen` in MessageHeader.h to represent the maximum message payload size we can safely assume a legacy peer (one that does not negotiate session parameters) can receive over TCP. * Clean up TCP session parameter code and PASE tests - Remove unused dead method `PeerAdvertisedTcpSupport` from `SecureSession`. - Assert `SupportedTransports == 0` in PASE test to verify that TCP parameters are not negotiated over PASE. * [pre-commit.ci] auto fixes from pre-commit.com hooks for more information, see https://pre-commit.ci * Support TCP session parameter negotiation in PASE and refine buffer size constants. - Promote `mLocalSessionParams` and accessors to `PairingSession` to enable symmetrical session parameter negotiation in `PASESession`. - Update `CommissioningWindowManager` and `Server::Init` to configure local session parameters for PASE commissioning sessions. - Add `SecurePairingHandshakeTCPParamsTest` in `TestPASESession.cpp` to verify session parameter negotiation during PASE handshakes. - Make `CHIP_SYSTEM_CONFIG_MAX_LARGE_BUFFER_SIZE_BYTES` explicitly evaluate `(64000 + 4)` with documentation for the 4-byte framing prefix. - Define `kLegacyDefaultMaxLargeAppMessageLen` as `64000 - 256` to provide a conservative header margin for unadvertised legacy TCP peers. * [pre-commit.ci] auto fixes from pre-commit.com hooks for more information, see https://pre-commit.ci * Clamp legacy TCP payload limit to local buffer capacity in SessionManager. --------- Co-authored-by: pre-commit-ci[bot] <66853113+pre-commit-ci[bot]@users.noreply.github.com> Co-authored-by: Copilot Autofix powered by AI <175728472+Copilot@users.noreply.github.com>

…otection Alarm state machine (#73812) * TC-EPALM-2.2: Add Python automation test script for the EPALM alarm state machine TC-EPALM-2.2 ("Primary functionality (alarm state machine) with Server as DUT") has had a merged test plan since 2026-07-16 in chip-test-plans #6157, but no Python automation was ever filed for it. TC-EPALM-2.1 and 2.3 both have scripts; 2.2 is the gap in the set. Implements all 14 steps against the merged plan. Steps 4 to 10 drive each fault bit through the General Diagnostics TestEventTrigger codes that the EPALM server already ships (ElectricalProtectionAlarmTestEventTriggerHandler, 0x00A3...0000 through 0007) and verify the corresponding State bit via a subscription report. Step 11 checks State is a subset of Supported. Step 14 clears the alarms. EPALM disallows the inherited RESET feature, so no Reset command is generated into the Python bindings and step 12 declares it locally to exercise the UNSUPPORTED_COMMAND path. TC_REFALM_2_2.py does the same for Refrigerator Alarm. Step 13 follows chip-test-plans #6364: ModifyEnabledAlarms is optional in Alarm Base rather than disallowed, so the script expects UNSUPPORTED_COMMAND only when the DUT does not list the command, and success when it does. Co-Authored-By: Claude Opus 5 (1M context) <noreply@anthropic.com> * ci: retrigger after an unrelated TC_AccessChecker flake The TC_[A] REPL shard failed on TC_AccessChecker.py, which timed out in setup_class with CHIP Error 0x00000032. That shard does not run TC-EPALM, and this branch adds only TC_EPALM_2_2.py. Co-Authored-By: Claude Opus 5 (1M context) <noreply@anthropic.com> * Do not wait for a State report when the trigger causes no transition ElectricalProtectionAlarmCluster::SetState returns early when the effective State is unchanged, with "no transition: no report, no event". The script waited for a subscription report after every trigger, so two paths could hang against a real DUT rather than fail: - a DUT that already has a fault raised when the test starts, where the set trigger for that alarm changes nothing; - the clear-all trigger in step 14 when nothing was raised, which happens when every feature-gated step was skipped. Both now read State instead of waiting when no transition is expected. The assertions are unchanged. The example app always boots with State clear and raises each alarm in turn, so CI does not reach either path; it matters for real DUTs. Reported by coderabbitai. Co-Authored-By: Claude Opus 5 (1M context) <noreply@anthropic.com> * Inline the step descriptions and drop steps_TC_EPALM_2_2 Moves each step's description and expectation into its self.step() call and removes the steps_ method, matching the current style. The feature-gated triggers were a loop over a table, which the AST step extractor cannot read: it requires constant arguments and points at a steps_ method as the alternative for dynamic ones. Steps 4 to 10 are now written out. Reported by jamesharrow. Co-Authored-By: Claude Opus 5 (1M context) <noreply@anthropic.com> * Move the EPALM TestEventTrigger handling into a shared TestBase The trigger codes were duplicated per script: this one carried all eight, and the seven TC-EPALM-3.x scripts each hard-code their own set and clear pair, so the same fifteen codes are spread across eight files. TC_EPALM_TestBase.py now owns them, keyed by AlarmBitmap rather than by raw value, so callers name the alarm and the mapping lives in one place: await self.send_test_event_trigger_set_alarm(AlarmBitmap.kArcFault) await self.send_test_event_trigger_clear_alarm(AlarmBitmap.kArcFault) await self.send_test_event_trigger_clear_all() It also carries read_state(), which was duplicated the same way. Reported by jamesharrow. Co-Authored-By: Claude Opus 5 (1M context) <noreply@anthropic.com> * Annotate the TestBase trigger helpers and document read_state Reported by coderabbitai. Co-Authored-By: Claude Opus 5 (1M context) <noreply@anthropic.com> * Replace desc_TC_EPALM_2_2 with a docstring matching the test plan The docstring carries the test case title and the plan's Purpose text with the adoc macros expanded, so the script reads the same as the rendered plan. The Purpose text is the wording chip-test-plans #6364 lands: the Reset command is rejected, and ModifyEnabledAlarms is optional in Alarm Base and checked against whichever behavior the DUT declares. That PR is still open, so the docstring is ahead of master until it merges. Reported by jamesharrow. Co-Authored-By: Claude Opus 5 (1M context) <noreply@anthropic.com> * Use the generated command id, the read_state helper, and plan-matching expectations Addresses review on #73812: - accepts_modify now reads Commands.ModifyEnabledAlarms.command_id instead of a hand-declared constant. - The three inline State reads go through the TestBase read_state helper. - Every step expectation is now the TC-EPALM-2.2 plan text with the adoc macros expanded, so the script and PR6364 read the same. * Verify Mask changes on ModifyEnabledAlarms; drop the latch clause Tracks the matching PR6364 plan edits: - New step 13b checks that a DUT accepting ModifyEnabledAlarms actually changes Mask, rather than only returning SUCCESS. - The docstring no longer claims latch verification; no step exercises it. * Bring the CodeRabbit fixes for TC-EPALM-2.2 into its own PR These four landed on the TC-EPALM-3.x branch because that is where CodeRabbit reviewed them, but they belong here: - the whole subscription lifetime is now inside try/finally, so a failure in steps 4 to 13b cannot leave alarms latched or the subscription registered - step 13b drains the State report that masking off an active bit produces, which step 14 would otherwise dequeue instead of the clear-all report - step 14 branches on the live State rather than the step 11 snapshot - Mask is captured before step 13a and restored best effort from finally --------- Co-authored-by: Claude Opus 5 (1M context) <noreply@anthropic.com>

* [EPALM] Add per-alarm clear TestEventTriggers The Electrical Protection Alarm trigger handler can set each of the seven alarm bits individually, but the only way to clear one is kClearAll, which clears them all at once. TC-EPALM-3.1 through 3.7, added by chip-test-plans #6302, each drive one alarm up and then back down to observe the Notify event carrying that bit in Active and then in Inactive. They reference clear codes 0x00A3...0011 through 0017, which do not exist, so none of those seven test cases can be automated today. Adds those seven codes and routes them to the cluster's existing DeactivateAlarms(), leaving kClearAll in place. Co-Authored-By: Claude Opus 5 (1M context) <noreply@anthropic.com> * ci: retrigger the Darwin test suites Both Darwin jobs died about three minutes into 'Build Apps Using Unified Build' with orphaned xcodebuild, ccache and llvm processes and no compile error in either log. The same job passes on master and on five sibling PRs, only two files reference the enum this branch extends and both are updated together, and the app builds clean locally from this branch. Co-Authored-By: Claude Opus 5 (1M context) <noreply@anthropic.com> * Call the cluster directly from each trigger case Drops the activate flag: every case now returns the ActivateAlarms or DeactivateAlarms result, and default: return false; is the only fallthrough. 66 lines to 48. --------- Co-authored-by: Claude Opus 5 (1M context) <noreply@anthropic.com>

…ed by platform-bot) (#73707) WiFi diagnostic functions (GetWiFiChannelNumber, GetWiFiCurrentMaxRate, etc.) access WLAN state without checking if STA is connected, causing crash when a wildcard read is performed before the device connects to an AP. Add IsWiFiStaConnected() guard function and return CHIP_ERROR_INCORRECT_STATE when WiFi STA is not connected before accessing WLAN diagnostic data. ### Testing TC-IDM-13.1, TC-IDM-14.1, TC-DESC-2.3 passed, verified on NXP platform.

* silabs/zephyr: declare requirements.zephyr.txt in module.yml Declare scripts/setup/requirements.zephyr.txt as a pip requirement file via the package-managers/pip/requirement-files schema, so that 'west packages pip --install' installs it automatically. Without this, none of matter's own Python dependencies (e.g. python-path, needed by scripts/codegen_paths.py) were ever picked up, and had to be installed manually into the west pipx venv. Co-authored-by: Copilot <223556219+Copilot@users.noreply.github.com> * silabs/zephyr: add python-path to requirements.zephyr.txt scripts/codegen_paths.py imports the python-path package (the 'python_path' module), but requirements.zephyr.txt only listed west, intelhex, cbor, cbor2 and jsonschema. python-path was only declared in requirements.build.txt, which is used by Matter's own GN/ninja build flow, not by the Zephyr module path. Co-authored-by: Copilot <223556219+Copilot@users.noreply.github.com> * silabs/zephyr: drop obsolete CONFIG_BT_RECV_WORKQ_BT Zephyr removed the BT_RECV_CONTEXT Kconfig choice (and its BT_RECV_WORKQ_SYS/BT_RECV_WORKQ_BT options) in commit 3b379ec594e ("Bluetooth: Host: Remove the BT_RECV_CONTEXT choice"): low-priority HCI packets now always run on the dedicated Bluetooth RX workqueue unconditionally, so the choice - and CONFIG_BT_RECV_WORKQ_BT with it - no longer exists. Setting the now-undefined CONFIG_BT_RECV_WORKQ_BT makes recent Zephyr abort the build with a Kconfig warning-as-error. Since the option is gone and its former behavior is now the only supported one, simply drop the two (duplicate) occurrences of the setting. Co-authored-by: Copilot <223556219+Copilot@users.noreply.github.com> * silabs/zephyr: rename NET_SOCKETS_POLL_MAX to ZVFS_POLL_MAX CONFIG_NET_SOCKETS_POLL_MAX was deprecated in Zephyr 4.0 in favor of CONFIG_ZVFS_POLL_MAX, and removed entirely by commit b3d4e29e088 ("net: sockets: Remove deprecated CONFIG_NET_SOCKETS_POLL_MAX"). Setting the now-undefined symbol makes recent Zephyr abort the build with a Kconfig warning-as-error. Rename to the new symbol, which has existed since Zephyr 4.0 and carries the same meaning, so this is a safe drop-in replacement for any currently supported Zephyr revision. Co-authored-by: Copilot <223556219+Copilot@users.noreply.github.com> * scripts: fix crash in clang-format version check when pigweed absent getClangFormatBinary() wraps the pigweed clang-format version comparison in a try/except meant to just print a warning and continue if anything goes wrong (e.g. pigweed.json is missing). However it called traceback.print_last(), which requires an exception to have already passed through sys.excepthook; inside a live except block there is no 'last exception' yet, so print_last() itself raises ValueError, which escapes the except and aborts the whole zap codegen step. Use traceback.print_exc() instead, which prints the currently-handled exception. This lets the version check degrade gracefully (falling back to the system clang-format with just a printed warning) instead of hard-failing when third_party/pigweed/repo is not checked out. Co-authored-by: Copilot <223556219+Copilot@users.noreply.github.com> --------- Co-authored-by: Copilot <223556219+Copilot@users.noreply.github.com>

…ion Alarm Notify event (#73816) * TC-EPALM-2.2: Add Python automation test script for the EPALM alarm state machine TC-EPALM-2.2 ("Primary functionality (alarm state machine) with Server as DUT") has had a merged test plan since 2026-07-16 in chip-test-plans #6157, but no Python automation was ever filed for it. TC-EPALM-2.1 and 2.3 both have scripts; 2.2 is the gap in the set. Implements all 14 steps against the merged plan. Steps 4 to 10 drive each fault bit through the General Diagnostics TestEventTrigger codes that the EPALM server already ships (ElectricalProtectionAlarmTestEventTriggerHandler, 0x00A3...0000 through 0007) and verify the corresponding State bit via a subscription report. Step 11 checks State is a subset of Supported. Step 14 clears the alarms. EPALM disallows the inherited RESET feature, so no Reset command is generated into the Python bindings and step 12 declares it locally to exercise the UNSUPPORTED_COMMAND path. TC_REFALM_2_2.py does the same for Refrigerator Alarm. Step 13 follows chip-test-plans #6364: ModifyEnabledAlarms is optional in Alarm Base rather than disallowed, so the script expects UNSUPPORTED_COMMAND only when the DUT does not list the command, and success when it does. Co-Authored-By: Claude Opus 5 (1M context) <noreply@anthropic.com> * [EPALM] Add per-alarm clear TestEventTriggers The Electrical Protection Alarm trigger handler can set each of the seven alarm bits individually, but the only way to clear one is kClearAll, which clears them all at once. TC-EPALM-3.1 through 3.7, added by chip-test-plans #6302, each drive one alarm up and then back down to observe the Notify event carrying that bit in Active and then in Inactive. They reference clear codes 0x00A3...0011 through 0017, which do not exist, so none of those seven test cases can be automated today. Adds those seven codes and routes them to the cluster's existing DeactivateAlarms(), leaving kClearAll in place. Co-Authored-By: Claude Opus 5 (1M context) <noreply@anthropic.com> * TC-EPALM-3.1 to 3.7: Add Python automation for the EPALM Notify event chip-test-plans #6302 adds seven Notify event test cases, one per alarm bit, sharing a parameterized test procedure. This adds the seven scripts. Each raises its alarm through the TestEventTrigger that sets it, checks the Notify event carries that bit in Active with State updated, then lowers it through the matching clear trigger and checks the event carries the bit in Inactive with State cleared. Step 3a doubles as cleanup, so the seven are order-independent. Each script is gated on the feature that owns its alarm bit, so a DUT that implements only some alarms runs only the applicable cases. Depends on the per-alarm clear triggers, which this branch carries. Co-Authored-By: Claude Opus 5 (1M context) <noreply@anthropic.com> * ci: retrigger the Darwin test suites Both Darwin jobs died about three minutes into 'Build Apps Using Unified Build' with orphaned xcodebuild, ccache and llvm processes and no compile error in either log. The same job passes on master and on five sibling PRs, only two files reference the enum this branch extends and both are updated together, and the app builds clean locally from this branch. Co-Authored-By: Claude Opus 5 (1M context) <noreply@anthropic.com> * ci: retrigger after an unrelated TC_AccessChecker flake The TC_[A] REPL shard failed on TC_AccessChecker.py, which timed out in setup_class with CHIP Error 0x00000032. That shard does not run TC-EPALM, and this branch adds only TC_EPALM_2_2.py. Co-Authored-By: Claude Opus 5 (1M context) <noreply@anthropic.com> * Do not wait for a State report when the trigger causes no transition ElectricalProtectionAlarmCluster::SetState returns early when the effective State is unchanged, with "no transition: no report, no event". The script waited for a subscription report after every trigger, so two paths could hang against a real DUT rather than fail: - a DUT that already has a fault raised when the test starts, where the set trigger for that alarm changes nothing; - the clear-all trigger in step 14 when nothing was raised, which happens when every feature-gated step was skipped. Both now read State instead of waiting when no transition is expected. The assertions are unchanged. The example app always boots with State clear and raises each alarm in turn, so CI does not reach either path; it matters for real DUTs. Reported by coderabbitai. Co-Authored-By: Claude Opus 5 (1M context) <noreply@anthropic.com> * Clear the alarm in a finally block so a failure cannot latch the DUT Step 3a doubles as cleanup, which the test plan relies on to keep the seven cases order-independent. That only holds when 3a is reached: an assertion or a timeout anywhere after the set trigger skipped both the clear and the subscription teardown, leaving the DUT latched into a fault for whatever ran next. Steps 2a to 3b now run inside a try, with the clear trigger and sub.cancel() in a finally. The cleanup trigger is itself guarded so a failure there cannot replace the original exception, and a second clear after a successful 3a is a no-op because no transition means no report. Reported by coderabbitai. Co-Authored-By: Claude Opus 5 (1M context) <noreply@anthropic.com> * Consume the baseline Notify before flushing, and prove the alarm starts clear flush_events() drains the handler's queue but does not wait for callback delivery. Step 1e fired the clear trigger and flushed immediately, so on a DUT that already had this alarm raised the resulting inactive Notify could arrive after the flush and still be queued when step 2a waits. Step 2a would then read the baseline event instead of the set-transition and fail its Active assertion. Step 1e now reads State first. If the alarm was already raised it consumes the inactive Notify deterministically, and either way it asserts the alarm is clear before the set trigger, so the precondition the rest of the case depends on is verified rather than assumed. Reported by coderabbitai. Co-Authored-By: Claude Opus 5 (1M context) <noreply@anthropic.com> * Inline the step descriptions and drop steps_TC_EPALM_2_2 Moves each step's description and expectation into its self.step() call and removes the steps_ method, matching the current style. The feature-gated triggers were a loop over a table, which the AST step extractor cannot read: it requires constant arguments and points at a steps_ method as the alternative for dynamic ones. Steps 4 to 10 are now written out. Reported by jamesharrow. Co-Authored-By: Claude Opus 5 (1M context) <noreply@anthropic.com> * Move the EPALM TestEventTrigger handling into a shared TestBase The trigger codes were duplicated per script: this one carried all eight, and the seven TC-EPALM-3.x scripts each hard-code their own set and clear pair, so the same fifteen codes are spread across eight files. TC_EPALM_TestBase.py now owns them, keyed by AlarmBitmap rather than by raw value, so callers name the alarm and the mapping lives in one place: await self.send_test_event_trigger_set_alarm(AlarmBitmap.kArcFault) await self.send_test_event_trigger_clear_alarm(AlarmBitmap.kArcFault) await self.send_test_event_trigger_clear_all() It also carries read_state(), which was duplicated the same way. Reported by jamesharrow. Co-Authored-By: Claude Opus 5 (1M context) <noreply@anthropic.com> * Use the shared TestBase in the TC-EPALM-3.x scripts and inline the steps Each of the seven scripts carried its own SET_TRIGGER and CLEAR_TRIGGER pair, an _await_notify and a _read_state. They now inherit all of that from TC_EPALM_TestBase and name the alarm rather than a raw trigger code. await_notify moves into the base as well, since all seven use it and 2.2 does not, and the ValueDecodeFailure guard it carries is worth stating once. Also drops steps_TC_EPALM_3_x in favour of the description and expectation inline at each self.step() call. Reported by jamesharrow. Co-Authored-By: Claude Opus 5 (1M context) <noreply@anthropic.com> * Annotate the TestBase trigger helpers and document read_state Reported by coderabbitai. Co-Authored-By: Claude Opus 5 (1M context) <noreply@anthropic.com> * Replace desc_TC_EPALM_2_2 with a docstring matching the test plan The docstring carries the test case title and the plan's Purpose text with the adoc macros expanded, so the script reads the same as the rendered plan. The Purpose text is the wording chip-test-plans #6364 lands: the Reset command is rejected, and ModifyEnabledAlarms is optional in Alarm Base and checked against whichever behavior the DUT declares. That PR is still open, so the docstring is ahead of master until it merges. Reported by jamesharrow. Co-Authored-By: Claude Opus 5 (1M context) <noreply@anthropic.com> * Use the generated command id, the read_state helper, and plan-matching expectations Addresses review on #73812: - accepts_modify now reads Commands.ModifyEnabledAlarms.command_id instead of a hand-declared constant. - The three inline State reads go through the TestBase read_state helper. - Every step expectation is now the TC-EPALM-2.2 plan text with the adoc macros expanded, so the script and PR6364 read the same. * Verify Mask changes on ModifyEnabledAlarms; drop the latch clause Tracks the matching PR6364 plan edits: - New step 13b checks that a DUT accepting ModifyEnabledAlarms actually changes Mask, rather than only returning SUCCESS. - The docstring no longer claims latch verification; no step exercises it. * Replace desc_TC_EPALM_3_x with docstrings matching the test plan Same treatment as TC-EPALM-2.2: the desc_ method goes and the test method carries the PR6302 Purpose for its alarm, with the adoc macros expanded. * Call the cluster directly from each trigger case Drops the activate flag: every case now returns the ActivateAlarms or DeactivateAlarms result, and default: return false; is the only fallthrough. 66 lines to 48. * Match the step expectations to the TC-EPALM-3.x plan text Same treatment as TC-EPALM-2.2: each expectation is now the PR6302 epalm_alarm_testcase.adoc text with the macros expanded, rather than a paraphrase. 49 expectations across the seven scripts. * [pre-commit.ci] auto fixes from pre-commit.com hooks for more information, see https://pre-commit.ci * Cover the whole subscription lifetime with try/finally Addresses CodeRabbit on #73816: - TC_EPALM_2_2 had no try/finally at all, so a failure in any of steps 4 to 13b left the DUT with alarms latched and the subscription registered. - TC_EPALM_3_3 and 3_7 opened their try after the baseline clear, so a failure in steps 1c to 1e skipped cleanup. The try now starts at the subscription. - Annotate await_notify with its return type, matching the other TestBase helpers. * Restore Mask and drain its State report in step 13b Alarm Base SetMask clears any active State bit the new mask no longer covers and reports the transition. Step 13b masks off a bit that steps 4-10 raised, so it produced a State report that step 14 would dequeue instead of the one from the clear-all trigger, and it left the reduced Mask on the DUT. Step 13b now saves the original Mask, drains that report when one is expected, and restores the Mask. Step 14 branches on the live State rather than the step 11 snapshot, which would otherwise be stale once 13b changes it. * Branch step 14 on the live State, not the step 11 snapshot This belongs with the previous commit, whose message already described it: the edit silently failed to apply and I pushed without noticing. Step 14 tested 'state', read back at step 11, which goes stale as soon as 13b masks off an active bit. If 13b cleared the last active bit, clear-all becomes a no-op that reports nothing and step 14 would block until the report timeout. * Restore Mask from finally, and capture it before step 13a The success-path restore left the DUT with a modified Mask whenever the command, the read-back, the assertion or the report drain failed. Capture now happens before 13a, which is the first write, and the restore is best effort in finally alongside the alarm cleanup. * Factor the TC-EPALM-3.x body into one parameterized base function Per James Harrow: seven near-identical 148-line scripts are hard to review, and the plan already expresses this as one parameterized include. The scripts now mirror that structure. TC_EPALM_TestBase gains alarm_notify_steps(), which builds the nine TestSteps with the epalm_alarm_testcase.adoc text and the alarm name and bit substituted, and run_alarm_notify_test(), which holds the body once. Each script drops from 148 lines to 79, of which 41 are the licence and CI arguments header. What remains is the alarm parameters, the PICS list, a two-line steps_ method and a one-line test body. steps_ comes back deliberately. step_extractor reads self.step() calls from the test method's own source and does not follow calls into a base class, so a delegating body would yield an empty step list with no error. The extractor documents this case: tests with dynamic steps should define an explicit steps_ method. Also fixes an inconsistency: 3_1, 3_2, 3_4, 3_5 and 3_6 still opened their try block after the baseline clear, so a failure in steps 1c to 1e skipped cleanup. Only 3_3 and 3_7 had been corrected. There is now one body and one placement. * [pre-commit.ci] auto fixes from pre-commit.com hooks for more information, see https://pre-commit.ci --------- Co-authored-by: Claude Opus 5 (1M context) <noreply@anthropic.com> Co-authored-by: pre-commit-ci[bot] <66853113+pre-commit-ci[bot]@users.noreply.github.com>

* Test scripts for AVANLY 2.1 and 2.2 * [pre-commit.ci] auto fixes from pre-commit.com hooks for more information, see https://pre-commit.ci * Exclude the helper class from CI * Address Code Comments * [pre-commit.ci] auto fixes from pre-commit.com hooks for more information, see https://pre-commit.ci * Address code comments, fix bad cut-n-paste * [pre-commit.ci] auto fixes from pre-commit.com hooks for more information, see https://pre-commit.ci * Address code comments * [pre-commit.ci] auto fixes from pre-commit.com hooks for more information, see https://pre-commit.ci * Address code comment re: finally block * [pre-commit.ci] auto fixes from pre-commit.com hooks for more information, see https://pre-commit.ci --------- Co-authored-by: pre-commit-ci[bot] <66853113+pre-commit-ci[bot]@users.noreply.github.com>

…(#73251) * Added code to JNI to manage ThreadCredentialsNeeded and WifiCredentialsNeeded events. Added Listeners that can be notified when those events happen. * [pre-commit.ci] auto fixes from pre-commit.com hooks for more information, see https://pre-commit.ci * Fixes * [pre-commit.ci] auto fixes from pre-commit.com hooks for more information, see https://pre-commit.ci * Potential fix for pull request finding Co-authored-by: Copilot Autofix powered by AI <175728472+Copilot@users.noreply.github.com> * [pre-commit.ci] auto fixes from pre-commit.com hooks for more information, see https://pre-commit.ci * Fixes from copilot. * Fixes suggested by copilot * Fixes * Fix * Replaced Wifi by WiFi. * Fixes * Potential fix for pull request finding Co-authored-by: Copilot Autofix powered by AI <175728472+Copilot@users.noreply.github.com> * Fix from Copilot * Use JniGlobalReference instead of jobject * [pre-commit.ci] auto fixes from pre-commit.com hooks for more information, see https://pre-commit.ci * Potential fix for pull request finding Co-authored-by: Copilot Autofix powered by AI <175728472+Copilot@users.noreply.github.com> * [pre-commit.ci] auto fixes from pre-commit.com hooks for more information, see https://pre-commit.ci * Notify listeners asynchronously. * [pre-commit.ci] auto fixes from pre-commit.com hooks for more information, see https://pre-commit.ci * Potential fix for pull request finding Co-authored-by: Copilot Autofix powered by AI <175728472+Copilot@users.noreply.github.com> * Fix problem reported during AI review. * [pre-commit.ci] auto fixes from pre-commit.com hooks for more information, see https://pre-commit.ci * Fix * Potential fix for pull request finding Co-authored-by: Copilot Autofix powered by AI <175728472+Copilot@users.noreply.github.com> * Fix following Andrei's review. --------- Co-authored-by: pre-commit-ci[bot] <66853113+pre-commit-ci[bot]@users.noreply.github.com> Co-authored-by: Copilot Autofix powered by AI <175728472+Copilot@users.noreply.github.com>

* TC_MOD_1_2: accept null StartUpMode per spec * Update src/python_testing/TC_MOD_1_2.py Co-authored-by: Juan Erasmo Trejo <jtrejoespinoza@griddynamics.com> * [pre-commit.ci] auto fixes from pre-commit.com hooks for more information, see https://pre-commit.ci * TC_MOD_1_2: add missing matter_asserts import and fix tuple bug --------- Co-authored-by: Juan Erasmo Trejo <jtrejoespinoza@griddynamics.com> Co-authored-by: pre-commit-ci[bot] <66853113+pre-commit-ci[bot]@users.noreply.github.com>

* Reject out-of-range SmokeSensitivityLevel writes SensitivityEnum is 0..2, but DataModel::Decode maps any other value onto kUnknownEnumValue and returns success, so a write of 3-255 was answered Success and the sentinel was stored. Reject those with CONSTRAINT_ERROR, matching the existing convention in FanControlCluster. * [pre-commit.ci] auto fixes from pre-commit.com hooks for more information, see https://pre-commit.ci --------- Co-authored-by: pre-commit-ci[bot] <66853113+pre-commit-ci[bot]@users.noreply.github.com>

…(#73728) * Add the mandatory Power Topology server to electrical-protection-app Endpoint 1 declares the Electrical Distribution Enclosure device type (0x0517), which lists Power Topology (0x009C) as a mandatory server cluster, but the app did not implement it. Of the NODE/TREE/SET choice, TreeTopology is the one that describes an enclosure: it provides power to itself and its child endpoints. TREE gates none of the cluster's attributes, so the delegate reports empty endpoint lists rather than inventing topology this app does not model. Also gives the app a TC_DeviceBasicComposition run, which it did not have. That test checks composition and attribute-list consistency, not device-type mandatory clusters, so it would not by itself have caught this. Co-Authored-By: Claude Opus 5 (1M context) <noreply@anthropic.com> * Trim the doc comments to what the code needs The comments restated this PR's rationale, which the description already carries verbatim. Keeps what a reader of the code needs and drops the device-type conformance argument and the feature-gating enumeration. Reported by malbert-silabs. --------- Co-authored-by: Claude Opus 5 (1M context) <noreply@anthropic.com>

… (#73835) * Add build time selection of default device behavior * [pre-commit.ci] auto fixes from pre-commit.com hooks for more information, see https://pre-commit.ci * apply comments --------- Co-authored-by: pre-commit-ci[bot] <66853113+pre-commit-ci[bot]@users.noreply.github.com>

* Fix RHT sensor silabs * fix comments

Cristian Bulacu (Cristib05) and others added 27 commits August 13, 2026 09:07

Update mve-sve to to also include 1.7: we want latest to always conta…

0ea5eb6

…in fixes that need backport (#73530)

Add AV Analysis Node device type (#73542)

5e65145

* [AVAnalysis] Add AV Analysis Node device type (0x0149) to matter-devices.xml Generated with Alchemy from the spec's AVAnalysisNode.adoc * [AVAnalysis] Regenerate code for the AV Analysis Node device type

Adding null check after AllocateResolveContext in MdnsAvahi::Resolve …

d8d0659

…for Linux and NuttX, and fixing potential use-after-free. (#39392) * Prevent segfaults in MdnsAvahi::Resolve for Linux and NuttX * Fixed commit mistake

Bump to docker 206 for cloudbuild (#73554)

9463786

[Silabs] Remove Wi-Fi ICD high-performance request during commissioni…

efc2cee

…ng (Auto-merged by platform-bot)

TC-DA-1.1 - Fix post-factory-reset PASE discovery with manual pairing…

0601790

… codes (#73461) * eval setup code * address pr comment

reject out-of-range keyset key count in KeySetData::Deserialize (#73045)

324c58a

Signed-off-by: Zibran Khan <kali834x@gmail.com>

silabs-matter-ci-bot Bot requested a review from a team as a code owner August 15, 2026 00:23

silabs-matter-ci-bot Bot added changing-git-submodules-on-purpose sl-require-admin-action This TAG indicates that an admin must take action. labels Aug 15, 2026

Arkadiusz Bokowy (arkq) and others added 29 commits August 27, 2026 18:59

Upload chip-cert and chip-tool for x64 and arm64 as artifacts (#73220)

bcbc684

* Upload chip-cert and chip-tool for x64 and arm64 as artifacts * Restrict artifacts uploads only for master pushes * Set retention days to 7 days

[Telink] Enable DeviceInfoProvider by default for all examples (Auto-…

d18d8cb

…merged by platform-bot) (#73800)

[Telink] Update Docker image (Zephyr update) (#73824)

49103c8

[Cirque] Update Bazel installation in Docker image (#73844)

15e052f

[silabs] Fix RHT sensor silabs (Auto-merged by platform-bot) (#73834)

d326795

* Fix RHT sensor silabs * fix comments

Junior Martinez (jmartinez-silabs) approved these changes Sep 1, 2026

View reviewed changes

Navigation Menu

Uh oh!

Sync csa branch with main - #1148

Sync csa branch with main#1148
silabs-matter-ci-bot[bot] wants to merge 142 commits into
mainfrom
automation/update_main

silabs-matter-ci-bot Bot commented Aug 15, 2026

Uh oh!

Reviewers

Assignees

Labels

Projects

Milestone

Development

Uh oh!

20 participants

Uh oh!

Conversation

silabs-matter-ci-bot Bot commented Aug 15, 2026

Uh oh!

Reviewers

Assignees

Labels

Projects

Milestone

Development

Uh oh!

20 participants