Blocked Unit Tests

This document tracks tests that cannot be generated due to missing mocks or unmockable dependencies. Only items that are currently blocked are listed here. Previously blocked items that have since been resolved are summarized in the History section at the bottom.

Current Status (2026-03-25)

411 test binaries, ~55 DISABLED individual tests (infrastructure limitations). 90.0% line coverage · 97.4% function coverage · 76.2% branch coverage (gcovr: 54,531/60,501 lines · 39,098/51,303 branches · 4,747/4,876 functions)

Session additions (2026-03-25)

12 new branch coverage test files across 8 modules: nsmPcieGroupBranch, nsmLeakDetectionBranch6, nsmGpuClockControlBranch3, nsmFpgaPortBranch4, nsmSecurityRBPBranch3, nsmEventConfigDeepBranch, nsmApSkuIdBranch, nsmMemoryBranch5, nsmRoTPropertyBranch4, nsmDebugTokenNICBranch3, nsmProcessorFactoryBranch9, nsmGpmOemBranch7.

8 test files renamed for consistent naming convention (commonBranch→common_branch, branches→branch, _branch→Branch).

Coverage delta: +3.8% branches (+1,502), +2.9% lines.

Session additions (2026-03-18)

Production code changes for testability (2 commits before tests):

ed8cf48c fix: remove static from factory functions — removed static from 40 factory coroutines/helpers in 14 files, enabling tests to forward-declare and call them directly.
51e7ef00 fix: dependency injection for testability — made MctpDiscovery public methods virtual (enabling mock subclasses), made NsmDevice::discoveryEvents() virtual with try-catch fallback, guarded dumpNsmDeviceInfoTask() MctpDiscovery call.

350+ new branch coverage tests across 14 test files: nsmProcessorBranch2 (92), nsmPowerLimitBranch (40), nsmMemoryBranch (22), nsmGpmOemBranch2 (25), nsmSwitchBranch (22), nsmPortBranch (27), nsmDebugTokenNICBranch (20), nsmProcessorModulePowerControlBranch (30), nsmEventConfigBranch (21), nsmCommonBranch (10), nsmGpuPresenceAndPowerStatusBranch (15), nsmLeakDetectionBranch (15), nsmSetECCModeBranch (5), nsmSetEgmModeBranch (5).

Session additions (2026-03-11)

Branch 12 second-operand coverage — systematic coverage of the rc == NSM_SW_SUCCESS && cc != NSM_SUCCESS (second operand of ||) pattern across all nsmd source files. Two commits:

Commit 5c6d932d — nsmClockOutputEnableState L65, nsmIRoTResponder L77, nsmSecurityRBP L118, nsmDebugTokenUnified L702, GPUSWInventory L121.

Commit ef5b7cda — nsmLeakDetection L596, nsmGpuClockControl L301 (clearReqClockLimit + setRangeClockLimits), nsmProcessorModulePowerControl L187, nsmLogInfo L92, nsmDebugInfo L110 + L240, nsmRetimerPort L1201.

Coverage delta: +1.0% branches, +2.3% decisions. 15 new tests added.

Recently Fixed DISABLED Tests (2026-03-10)

Three previously-disabled tests have been fixed and re-enabled:

Test	File	Fix
`CoroutineSemaphoreMock.Contended_DeferFails_WaiterNotResumed`	`common/test/sleep_semaphore_mock_test.cpp`	Fixed memory leak in `coroutineSemaphore.hpp` — `delete awaiterPtr` added on `sd_event_add_defer` failure
`NsmEraseTraceObjectTest.EraseDebugInfo_UnsupportedType_SetsInternalFailure`	`nsmd/nsmEvent/test/nsmEvent_dump_test.cpp`	Changed `static_cast<EraseInfoType>(99)` → `EraseInfoType::Other` to avoid sdbusplus enum-to-string throw
`NsmSwitchIsolationModeHandleResp.HandleResponseMsg_ShortLen_ReturnsError`	`nsmd/nsmChassis/test/nsmSwitchBranch_test.cpp`	Fixed buffer size from `sizeof(nsm_msg_hdr)` → `sizeof(nsm_msg_hdr)+sizeof(nsm_common_non_success_resp)`; cc=0 so decode_reason_code_and_cc succeeds → NSM_SW_ERROR_LENGTH correctly returned

Test Infrastructure Available

The following mock/test infrastructure is fully functional and available for new tests:

Component	Location	Purpose
`MockNsmDevice`	`nsmd/test/commonMock.hpp`	Mocks all NsmDevice pure virtuals (sensorIO, postPatchIO, updateNsmDevice, doSleep, refreshCapabilitySensor)
`MockSensorManager`	`nsmd/test/mockSensorManager.hpp`	Provides mockManager singleton, `getEid()`, `getObjServer()`
`NullReturnMockSensorManager`	`nsmd/test/mockSensorManager.hpp`	Returns `nullptr` from `getNsmDeviceFromStaticUUID()` to test factory error branches
`SensorManagerTest` fixture	`nsmd/test/mockSensorManager.hpp`	Base class: `mockSensorIO()`, `mockPostPatchIO()`, `cleanupDeviceSensors()`
`MockDbusAsync`	`common/test/mockDBusHandler.hpp`	`propertyMap(objPath, intf)` registers D-Bus property maps; `findPropertyMap()` returns nullptr if absent
`DBusTest` fixture	`common/test/mockDBusHandler.hpp`	Clears D-Bus mock state in destructor
`EXPECT_THROW_COROUTINE`	`nsmd/test/commonMock.hpp`	Works in both `COVERAGE_DISABLE_COROUTINES` mode (direct throw) and normal mode
`MctpTestFixture`	`common/test/mockSocketIo.hpp`	MCTP socket pair infrastructure for socket_handler tests
`COVERAGE_DISABLE_COROUTINES`	build configuration	Makes coroutines execute synchronously; `co_await` → direct call, `co_return` → `return`

Files Genuinely Blocked (Current)

1. `requester/mctp_endpoint_discovery.cpp` — 0% coverage, 606 lines

Why blocked: The MctpDiscovery class requires a fully-initialized singleton (MctpDiscovery::getInstance()). The singleton initialization requires:

A real sdeventplus::Event loop
A real sdbusplus::bus::bus session
A real MCTP socket file descriptor
D-Bus service registration for xyz.openbmc_project.MCTP.Control

In unit tests MctpDiscovery::getInstance() throws std::runtime_error because the singleton is never initialized. There is no way to inject a mock instance.

Uncovered functionality: Endpoint discovery, MCTP UUID resolution, device enumeration, endpoint state change notifications, service ready detection, all 1,200+ lines of discovery logic across mctp_endpoint_discovery.cpp and related headers.

Recommended action: Integration tests with a real event loop. Note: MctpDiscovery methods are now virtual (commit 51e7ef00), enabling mock subclasses. A MockMctpDiscovery test fixture can now override getNsmDeviceFromEid(), discoveryEvents(), etc. However, the singleton initialize() pattern still prevents full injection — a setTestInstance() or constructor-based injection would complete the testability story.

2. `nsmd/nsmDevice.cpp` — 36% coverage, 411 uncovered lines (of 643)

Why blocked: MockNsmDevice overrides all pure virtuals — meaning the real implementations in nsmDevice.cpp are never called. The actual coroutine bodies (which contain the real logic) require:

A real requester::Handler<requester::Request> for SendRecvNsmMsg
A real sdeventplus::Event and InstanceIdDb for async MCTP send/receive
A real mctp_socket::Manager for socket communication

The real SendRecvNsmMsg is the root call for all sensor I/O, and without a real requester handler the entire chain from sensorIO → SendRecvNsmMsg → MCTP is untestable.

Uncovered lines (183–1228):

Lines	Function	Reason
183–265	`SendRecvNsmMsg()`, `sensorIO()`, `postPatchIO()`	Real implementations never called — MockNsmDevice overrides these
359–943	`updateNsmDevice()`, `getSupportedNvidiaMessageType()`, `getSupportedCommandCodes()`, `waitForNsmDeviceUpdate()`, `setOnline()`, `setOffline()`, `updateSensorsForOffline()`, `markSensorsUnrefreshed()`	All are coroutines calling real sensorIO
966–1084	`getFRU()`, `getInventoryInformation()`, `updateFruDeviceIntf()`	Coroutines calling real sensorIO
1114–1228	`refreshCapabilitySensor()`, `refreshCommandMatrix()`, `dumpNsmDeviceInfoTask()`	Coroutines calling real sensorIO

Recommended action: Same as mctp_endpoint_discovery.cpp — requires real requester infrastructure or a full coroutine executor mock. Note: discoveryEvents() is now virtual with a try-catch fallback (commit 51e7ef00), and dumpNsmDeviceInfoTask() guards the MctpDiscovery call. These changes prevent test crashes but don't make the real coroutine bodies testable without real MCTP infrastructure.

3. `nsmd/sensorManager.cpp` — 29% coverage, 290 uncovered lines (of 411)

Why blocked: Covered lines (121/411) are the static helper methods that were extracted and tested in isolation (isNSMPollReady, dumpReadinessLogs, markEMReady, startPolling). The remaining 71% consists of event-driven handlers that fire only via D-Bus signals and sdeventplus event loop callbacks.

Covered (29%):

isNSMPollReady() (L50-65) — static, tested directly
dumpReadinessLogs() (L386-415) — static, tested directly
markEMReady(), isEMReady() (partial) — tested via SensorManagerTest
startPolling(), pollPrioritySensors(), pollNonPrioritySensors() — tested via deviceTask loop mock in sensorManager_test.cpp

Uncovered (71%) — all require real D-Bus/event-loop infrastructure:

Lines	Function	Blocker
72–358	Constructor body — D-Bus signal matches, `sd_event` setup, timer callbacks, `sdbusplus::asio::object_server`	`sdbusplus::asio::object_server` not mockable
461–545	`scanInventory()`	Requires live EntityManager D-Bus tree
552–635	`interfaceAddedHandler()`, `interfaceAddedTask()`	D-Bus signal callbacks, never fire in tests
639–711	`gpioStatusPropertyChangedHandler()`, `mctpReadinessSigHandler()`	Signal handlers, never fire
771–785	`checkAllDevicesReady()`, `dumpNsmDevicesInfo()`	Called from event handlers

Recommended action: Integration tests. The static helpers are fully covered; the async event handlers cannot be unit-tested without a real D-Bus session.

4. `nsmd/nsmEvent/nsmLongRunningEventHandler.cpp` — 50% coverage, 6 uncovered lines (of 12)

Why blocked: handle() function (lines 35–44) immediately calls MctpDiscovery::getInstance() which throws std::runtime_error in tests. The test Delegate_MctpDiscoveryNotInit_ThrowsRuntimeError in nsmEvent_test.cpp covers the exception path (lines 37–39 via catch). Lines 35, 37, 39–40, 42, 44 (the success path after getInstance() returns) are permanently unreachable.

Recommended action: Refactor handle() to accept MctpDiscovery& by injection instead of using the singleton directly.

5. `requester/mctp_endpoint_prober.cpp` — 47% coverage, 96 uncovered lines (of 183)

Why blocked: McEndpointProber requires a real sendRecvFn callback (actual MCTP socket I/O function pointer). The uncovered half consists of the async send/receive paths triggered by the real sendRecvFn:

probe() body (lines 44–95) — async MCTP probe logic
handleResponse() (lines 131–185) — response parsing per endpoint
Retry/timeout paths (lines 226–301) — require real timer callbacks

The test file requester_test.cpp covers the synchronous paths that don't require real socket I/O.

Recommended action: Inject sendRecvFn as a mock lambda in tests.

6. `requester/handler.hpp` — 50% coverage, 56 uncovered lines (of 112)

Why blocked: Template retry/timeout logic (lines 222–339) is tested only via the mock path in requester_test.cpp. The real paths require:

sdeventplus::Event for async timer callbacks
Real sd_event_source for timeout handling
InstanceIdDb for MCTP instance ID allocation under load

Uncovered paths: Retry-after-timeout (lines 222–260), concurrent request limiting (lines 312–339), request queue draining (lines 449–513).

7. `requester/request.hpp` — 8% coverage, 31 uncovered lines (of 34)

Why blocked: Template class Request<ExecutionContext, Handler> — all paths except construction require a live sdeventplus::Event loop. The timer-based retry/expiry logic (lines 65–234) fires via sd_event callbacks that never trigger in unit tests.

8. `common/dBusHandler.cpp` — 27% coverage, 202 uncovered lines (of 279)

Why blocked: Real sdbusplus::bus::bus D-Bus methods. The 77 covered lines are the MockDBusHandler dispatch paths. The remaining 73% (lines 26–197 approx) are real D-Bus call() and new_method_call() implementations that require a live D-Bus session bus. MockDBusHandler intercepts all calls in tests and the real implementations are never reached.

9. `nsmd/nsmPort/nsmRetimerPort.cpp` — 83% coverage (up from 66%)

Partially resolved — static factory functions unblocked.

9a. Previously blocked static factory functions — NOW RESOLVED

createNsmPCIeRetimerPorts() and createNsmMultiPCIeRetimerPorts() had static linkage. static was removed, forward declarations added to test files, and 16 new tests were written covering all count() FALSE branches, portType switch paths, CX9 includeInboundCounters, and null-device error paths.

9b. Remaining uncovered lines (gcov B50/B51 false negatives, lines 23–90)

Constructor/destructor bodies and string literal continuation lines in lg2:: calls (nsmRetimerPort.hpp constructors, lines 23–90) — appear uncovered in the HTML report but ARE executed. This is a tooling artifact, not a real coverage gap.

10. `nsmd/nsmMsghandler.hpp` — 51% coverage, 13 uncovered lines (of 27)

Why blocked: The real SendRecvNsmMsg implementation in NsmMsgHandler requires a live MCTP socket session. Lines 107, 120, 124–138 are the actual send/receive coroutine body. In tests, MockNsmMessageHandler is used instead.

Individual Branch Blockers

Remaining `||` second-operand branches (rc==NSM_SW_SUCCESS, cc!=NSM_SUCCESS)

All if (rc != NSM_SW_SUCCESS || cc != NSM_SUCCESS) conditions have been systematically analysed (2026-03-11). All second operands are now covered except the items below, which are permanently blocked:

File	Line	Reason
`nsmd/nsmDevice.cpp`	L623, L688, L870	`MockNsmDevice.nsmMsgHandler` is null → `getSupportedNvidiaMessageType`, `getSupportedCommandCodes`, `getInventoryProperty` never called (see §2)
`nsmd/nsmChassis/nsmPCIeFunction.cpp`	L116	Test binary `nsmPCIeFunction_test` disabled in meson.build (`NsmInterfaceProvider` deleted constructor)
`nsmd/nsmDebugToken/nsmDebugTokenNIC.cpp`	L614	Dead code: second `decode_nsm_query_device_ids_resp` call (full buffer) after first (size query) already succeeded — same response can't fail the second time
`nsmd/nsmDebugToken/nsmDebugTokenUnified.cpp`	L715	Dead code: defensive check after `cc == NSM_SUCCESS` already verified at L702
`nsmd/nsmChassis/nsmIRoTResponder.cpp`	L89	Dead code: second decode after `cc == NSM_SUCCESS` already verified at L77

`nsmd/nsmDumpCollection/nsmLogInfo.cpp` — `default:` in synced_time switch

Lines: 121–123 (default: branch in getLogInfoAsyncHandler())

Pattern:

switch (logInfo.synced_time) {
    case SYNCED_TIME_TYPE_BOOT:   // = 0
    case SYNCED_TIME_TYPE_SYNCED: // = 1
    default: → InternalFailure   // UNREACHABLE
}

Why blocked: logInfo.synced_time is a 1-bit bitfield (uint8_t synced_time : 1). A 1-bit field holds only values 0 or 1. Any attempt to assign logInfo.synced_time = 2 is silently truncated to 0 by C++ standard. The default: branch is structurally impossible to reach without changing the struct definition.

`nsmd/nsmNumericSensor/nsmThresholdFactory.cpp` — error branch in `make()`

Lines: 58–63 (if (result != NSM_SW_SUCCESS) after getThresholdInterfacesAsync())

Why blocked: getThresholdInterfacesAsync() always co_return NSM_SW_SUCCESS. There is no code path that returns a non-success value. coGetDbusProperty<string> failure in the mock infrastructure suspends permanently rather than returning an error code. So result is always NSM_SW_SUCCESS and the error branch is dead code.

`nsmd/nsmEvent.cpp` — blocked decision branches

Line 93 — co_return NSM_SW_SUCCESS (FALSE branch of if (!logSuccess) at L85): In COVERAGE_DISABLE_COROUTINES mode, -Dco_await= replaces co_await with nothing. coLogEvent constructor initializes success = false. await_ready() (which would set success = true) is never called. So operator bool() always returns false, meaning !logSuccess is always true and only the error path is taken.

Line 73 — FALSE branch of if (!cachedLoggingService) inside mutex: Requires two concurrent calls where the second call caches the service between the outer guard and the inner check. Impossible in single-threaded tests.

Lines 158–170 in DelegatingEventHandler::delegate() — both branches of if (!nsmDevice): MctpDiscovery::getInstance() throws at line 156 before line 158 is reached. Neither branch is ever taken.

Lowest-Coverage Files — Root Cause Summary

Deep analysis (2026-03-10, updated 2026-03-11) confirmed that all remaining uncovered lines fall into one of three categories. No new test opportunities were found beyond the two items listed under Category G below.

%	File	Uncovered lines	Root cause
27%	`common/dBusHandler.cpp`	202/279	BLOCKED — real D-Bus `call()` required (see §8)
36%	`nsmd/nsmDevice.cpp`	411/643	BLOCKED — MCTP transport required (see §2)
53%	`nsmd/nsmCommon/nsmPciePortIntf.cpp`	6/13	EH_BRANCH — constructor initializer list (gcov artifact)
63%	`nsmd/nsmChassis/nsmPCIeRetimer.cpp`	33/90	EH_BRANCH — constructor initializer list (gcov artifact)
68%	`nsmd/nsmEvent/nsmRediscoveryEvent.cpp`	33/104	BLOCKED — `MctpDiscovery` singleton required
70%	`nsmd/nsmPort/nsmEndpoint.cpp`	19/65	EH_BRANCH — constructor initializer list (gcov artifact)
73%	`nsmd/nsmChassis/nsmPowerControl.cpp`	38/145	EH_BRANCH (L33–64) + DEAD (L129–156 loop getters: NsmPowerCap objects complex to build; L200 stub always returns 0)
74%	`nsmd/nsmChassis/nsmPowerSubSystem.cpp`	16/62	EH_BRANCH — constructor initializer list (gcov artifact)
75%	`nsmd/nsmChassis/nsmApSkuId.cpp`	20/82	EH_BRANCH — constructor initializer list (gcov artifact)
76%	`nsmd/nsmProcessor/nsmFpgaProcessor.cpp`	25/105	EH_BRANCH — constructor initializer list (gcov artifact)
78%	`nsmd/nsmPort/nsmZone.cpp`	11/50	EH_BRANCH — constructor initializer list (gcov artifact)
79%	`nsmd/nsmChassis/nsmRoTProperty.cpp`	100/498	EH_BRANCH (constructors) + DEAD (2-value enum exhausts switch, hardcoded encode) + BLOCKED (imageCopyAsyncHandler, getActiveSlotComponentInfo need real D-Bus subtree)
80%	`nsmd/nsmFwSwInventory/GPUSWInventory.cpp`	25/128	EH_BRANCH (constructor) + BLOCKED (real `sensorIO` required)
80%	`nsmd/nsmPort/nsmFpgaPort.cpp`	36/186	EH_BRANCH — constructor initializer list (gcov artifact)
81%	`nsmd/nsmNumericSensor/nsmEnergy.cpp`	10/54	EH_BRANCH — constructor + shared-memory macro lines (gcov artifact)
81%	`nsmd/nsmNumericSensor/nsmTemp.cpp`	10/54	EH_BRANCH — constructor + shared-memory macro lines (gcov artifact)
82%	`nsmd/nsmChassis/nsmLeakDetection.cpp`	84/469	EH_BRANCH (L42–118 constructor) + DEAD (L358–362 encode fail unreachable) + small testable remainder
82%	`nsmd/nsmDebugToken/nsmDebugTokenNIC.cpp`	67/389	EH_BRANCH (constructors) + DEAD (protocol: decode strips reasonCode on NSM_SUCCESS → else branches unreachable)
82%	`nsmd/nsmEvent/nsmLongRunningEvent.cpp`	6/35	EH_BRANCH — constructor initializer list (gcov artifact)
83%	`nsmd/nsmChassis/nsmUpdateApSkuId.cpp`	17/102	DEAD — `validateApSkuIdFormat` enforces exactly 10 chars → stoul never overflows, pos always equals length, parsedValue always fits uint32
83%	`nsmd/nsmGPIO/nsmGPIOState.cpp`	10/59	EH_BRANCH — constructor initializer list (gcov artifact)
83%	`nsmd/nsmGPM/nsmGpmOem.cpp`	52/323	EH_BRANCH (constructor multi-line initializers) + TESTABLE: `addGpmIntfProperty(name, dataType)` overload (L447–480) has zero tests — see Category G
83%	`nsmd/nsmPort/nsmRetimerPort.cpp`	199/1180	EH_BRANCH — constructor initializer lists (L23–90 gcov artifact); factory branches now covered (see §9)
83%	`nsmd/nsmNumericSensor/nsmPower.cpp`	12/71	EH_BRANCH — constructor + shared-memory macro lines (gcov artifact)

Root cause breakdown of all remaining uncovered lines (estimated):

~60% Compiler-generated EH branches / gcov false negatives (tool limitation)
~25% Dead code (input validation / enum exhaustion / hardcoded instance IDs)
~15% Infrastructure-blocked (real MCTP/D-Bus/sensorIO required)

Branch Coverage Analysis

Current state: 72.4% branches (37,596 / 51,925) — 2026-03-18

The 12,836 not-taken branches fall into these categories:

Category A: gcov B50/B51 compound-condition false negatives

The build generates gcno/gcda in ‘B50’ format but the host gcov expects ‘B51’. This causes the FALSE branch of compound && conditions to be reported as “not-taken” even when tests exercise the error path.

Pattern (across hundreds of handleResponseMsg functions):

if (rc == NSM_SW_SUCCESS && cc == NSM_SUCCESS) { /* update */ }
// ↑ FALSE branch always "not taken" in report, even though
//   tests with NSM_ERROR cc DO execute the error path

Not fixable — tooling limitation.

Category B: Compiler-generated EH cleanup branches

The compiler inserts hidden exception-handling branches around constructor/destructor member initializer lists and RAII cleanup code. These appear as “not-taken” in gcov reports for every C++ project.

Confirmed files: nsmThresholdValue.cpp (all 6 branches at L26/38/50/62/74/86), nsmFabricManager.cpp constructors (L45–82, L135–150), nsmMemory.cpp constructors (91 branches), nsmPciePortIntf.cpp (L18–20), nsmSensorAggregator.cpp (L28/30), nsmPeakPower.cpp, nsmGpuOperationalStatus.cpp. Not fixable — compiler artifact.

Category C: `if constexpr` template branches

nsmInterface.hpp line 152 has 860 not-taken branches — all from a single if constexpr (std::is_invocable_v<Func, std::string, IntfType&>) inside NsmInterfaces<T>::invoke(). Each template instantiation compiles only one branch; gcov counts the dead branch of every other instantiation. Not fixable — compile-time decision.

Category D: LG2 macro flood-control branches

LG2_ERROR_FLT expands to if (SHOULD_LOG(...)) { ... }. The FALSE branch (flood suppression) requires the flood threshold to be reached during a test. Affected: nsmSetECCMode.cpp L65, nsmGPIOState.cpp L48/69, nsmFabricManager.cpp L198, and others. Not fixable — requires timing/count conditions.

Category E: Concurrency-required branches

nsmEvent.cpp L73 — FALSE branch of if (!cachedLoggingService) inside mutex. Requires a second concurrent call to set the cache between the outer check and the inner check. Impossible in single-threaded tests. Not fixable — architectural constraint.

Category F: Infrastructure-blocked success paths

nsmEvent.cpp L93 — co_return NSM_SW_SUCCESS in logEventAsync (see above)
nsmRediscoveryEvent.cpp handle() success path — MctpDiscovery singleton not initialized
nsmLongRunningEventHandler.cpp — MctpDiscovery singleton not initialized (see above) Not fixable without architectural changes.

Category G: Remaining testable opportunities

Two specific items have been identified as genuinely testable. Everything else has been analysed and confirmed as EH_BRANCH, dead code, or infrastructure-blocked.

G1. nsmd/nsmGPM/nsmGpmOem.cpp — addGpmIntfProperty(name, dataType) overload

Lines 447–480 are completely untested. The function has three exercisable paths:

Null gpmIntf (before createGPMIntf() is called) → early-return log at L449–451
DataType::Double switch case → register_property(name, NaN) at L455–458
DataType::VectorDouble switch case → register_property(name, vector{}) at L459–462

Estimated gain: ~15 lines, ~6 branches.

Test pattern:

// null gpmIntf path
NsmGPMInterfaceCreator creator(objServer, "/test/path");
creator.gpmIntf.reset();
EXPECT_NO_THROW(creator.addGpmIntfProperty("Metric", DataType::Double));

// valid gpmIntf + switch cases
NsmGPMInterfaceCreator creator2(objServer, "/test/path2");
creator2.createGPMIntf();
creator2.addGpmIntfProperty("DoubleMetric",  DataType::Double);
creator2.addGpmIntfProperty("VectorMetric",  DataType::VectorDouble);

G2. count() FALSE branches in factory coroutines

Every factory coroutine guards D-Bus property access with:

if (allBaseIfaceProperties.count("Name"))
    name = std::get<std::string>(...);  // FALSE branch: name stays ""

Existing tests always provide complete property maps, so the FALSE branch is never taken. These are testable — exercising them increments the gcov counter.

Estimated remaining: ~150–200 branches across the files below (~0.3–0.4% gain).

Partially fixed: commit d09a534a (nsmFpgaPort: Health, ChasisPowerState, PortType, PortProtocol, Priority absent tests), commit 6208fb16 (nsmGpuPciePort: DeviceIndex, ClearableScalarGroup; nsmChassisPCIeDevice: InventoryObjPath; nsmGpuClockControl: else-if FALSE).

Remaining files (by count() check count in factory):

File	~count() checks	Key absent properties
`nsmd/nsmProcessor/nsmProcessor.cpp`	26	Name, UUID, per-type properties
`nsmd/nsmChassis/nsmIRoTResponder.cpp`	13	Multiple factory guards
`nsmd/nsmDeviceInventory/nsmSwitch.cpp`	11	Various per-type properties
`nsmd/nsmEvent/nsmXIDEvent.cpp`	10	Base property guards
`nsmd/nsmPort/nsmPort.cpp`	9	Health, ChasisPowerState, PortType
`nsmd/nsmChassis/nsmNVSwitchAndNVMgmtNICChassis.cpp`	9	Chassis-specific
`nsmd/nsmNumericSensor/nsmThresholdEvent.cpp`	8	Threshold properties
`nsmd/nsmChassis/nsmChassisAssembly.cpp`	8	Assembly properties
`nsmd/nsmMemory/nsmMemory.cpp`	7	ErrorCorrection, DeviceType, Priority
`nsmd/nsmChassis/nsmChassisLED.cpp`	4	Name, Type, UUID, InventoryObjPath
`nsmd/nsmPort/nsmZone.cpp`	4	Zone-specific properties
`nsmd/nsmEventConfig/nsmEventConfig.cpp`	4	EventConfig properties

Test pattern:

auto& pm = utils::MockDbusAsync::propertyMap(objPath, interface);
pm["UUID"] = uuid;
pm["InventoryObjPath"] = std::string("/xyz/.../device");
// "Name" and "Type" intentionally omitted → count() returns 0 → FALSE branches taken
createNsmXxxSensor(mockManager, interface, objPath);

Compilation-Blocked Tests

`nsmd/nsmPort/nsmRetimerPort.cpp` — RESOLVED

createNsmPCIeRetimerPorts and createNsmMultiPCIeRetimerPorts — static keyword removed (2026-03-10). Tests added in nsmRetimerPort_test.cpp (12 tests) and nsmNullDeviceBranch_test.cpp (2 tests). Coverage: 66% → 83%.

`nsmd/nsmPort/nsmPort.cpp` — RESOLVED

createNsmPortSensorWithNetworkPortAddresses and createNsmPortSensorGeneric — static keyword removed (2026-03-10). Tests added in nsmPort_test.cpp (2 tests).

DISABLED Tests (~50 total)

Tests prefixed with DISABLED_ are compiled but not run. Common reasons:

Reason	Example
Sync validation not in production code	DISABLED_SetPowerCap_AboveMax in some fixtures
MockSensorManager always creates device (null branch unreachable)	Some factory null-device tests
Valgrind false positive isolation	nsmRawCommandHandler valgrind sdbusplus false positive
Test infrastructure mismatch	GetBlobHandlesEmptyFile behavior changed
sdbusplus property getter/setter disconnect	`DISABLED_RequestImageCopy_AlreadyProcessing_Throws` — `imageCopyRequestStatus(Processing)` setter not reflected by `ImageCopyRequestStatus()` getter in test D-Bus environment

History — Previously Blocked, Now Solved

The following were documented as blocked in earlier sessions but have since been resolved with tests:

File	Old Status	Current Coverage	Solution
`nsmd/socket_handler.cpp`	0% — “socket operations require real system calls”	98% (246/249 lines)	`socket_handler_test.cpp` using `socketpair()` + `MctpTestFixture`; 3 uncovered lines (303–305) are unreachable dead code in a `[[noreturn]]` error path
`nsmd/sensorQueueMap.hpp`	“compilation fails — excess elements in struct initializer”	97% (33/34 lines)	`sensorQueueMap_test.cpp` fixed initialization; line 33 is a closing brace (gcov artifact)
`nsmd/nsmDebugToken/nsmDebugTokenAggregation.cpp`	“file I/O, TLV parsing, async operations unblocked”	92% (190/206 lines)	Tests in `nsmKeyMgmtAndDebugToken_test.cpp` using mockSensorIO; remaining 16 lines are constructor EH branches (gcov artifacts)
`nsmd/nsmFirmwareUtils/nsmKeyMgmt.cpp`	“coroutine-based key management, NsmDevice mock needed”	89% (148/165 lines)	Tests in `nsmKeyMgmtAndDebugToken_test.cpp`; uncovered: coroutine infrastructure-blocked paths
`nsmd/nsmPort/nsmGpuPciePort.cpp`	“coroutine async, NsmDevice mock”	86% (283/328 lines)	Tests in `nsmGpuPciePort_test.cpp`; uncovered: gcov B50/B51 artifacts (lines 39–119)
`nsmd/nsmChassis/nsmRoTProperty.cpp`	“similar to nsmKeyMgmt, likely coroutine-based”	79% (398/498 lines)	Tests in `nsmChassis/test/`; uncovered: coroutine bodies and EH branches
`nsmd/nsmDumpCollection/nsmLogInfo.cpp`	“1-bit bitfield blocks default: branch”	86% (88/102 lines)	Tests in `nsmLogInfo_test.cpp`; lines 121–123 remain (1-bit bitfield, see above)
`requester/mctp.c` (libnsm)	“requires real sendRecvFn”	100%	Socket mocking infrastructure (MctpTestFixture)
`nsmd/sensorManager.cpp` (static helpers)	“extensive unmockable dependencies, 0%”	29% (static helpers covered)	Static helper methods extracted and tested directly; event-driven portion remains blocked
Factory `if (!nsmDevice)` error paths	“mockManager always creates devices”	Covered	`NullReturnMockSensorManager` fixture; 19 null-device tests across 8 subsystems

Blocked Unit Tests

Current Status (2026-03-25)

Session additions (2026-03-25)

Session additions (2026-03-18)

Session additions (2026-03-11)

Recently Fixed DISABLED Tests (2026-03-10)

Test Infrastructure Available

Files Genuinely Blocked (Current)

1. requester/mctp_endpoint_discovery.cpp — 0% coverage, 606 lines

2. nsmd/nsmDevice.cpp — 36% coverage, 411 uncovered lines (of 643)

3. nsmd/sensorManager.cpp — 29% coverage, 290 uncovered lines (of 411)

4. nsmd/nsmEvent/nsmLongRunningEventHandler.cpp — 50% coverage, 6 uncovered lines (of 12)

5. requester/mctp_endpoint_prober.cpp — 47% coverage, 96 uncovered lines (of 183)

6. requester/handler.hpp — 50% coverage, 56 uncovered lines (of 112)

7. requester/request.hpp — 8% coverage, 31 uncovered lines (of 34)

8. common/dBusHandler.cpp — 27% coverage, 202 uncovered lines (of 279)

9. nsmd/nsmPort/nsmRetimerPort.cpp — 83% coverage (up from 66%)

10. nsmd/nsmMsghandler.hpp — 51% coverage, 13 uncovered lines (of 27)