Test strategy

Layers

LayerEvidenceScope
niseCore focused unitGTest/GMock CMake targetsmatcher, sensor, storage and supporting PAL/VFM pieces
WBF source-derived unitExtracted CBiometricDevice functions + generated customer propscapture, thread, completion, wake and cancellation paths
C++/Rust adapter integrationDedicated C++/Rust workflows and replay-oriented teststested adapter surfaces and parity observations
Virtual specificationsMarkdown test cases and generated harnessesfunction-level preconditions and expected results
Customer buildExact niseCore/WBF/include refs + WBF projectcompile/link/package compatibility
Firmware integrationHost command/update tests plus exact device/firmware versioncommand/status/update behavior across the external firmware boundary
Driver validationHLK, accuracy, package/security checksOS/device/customer acceptance evidence
Release evidenceWorkflow run, signed artifact and exact SHA rowdelivery result

utest branch model

dev is the implementation baseline with 333 files and complete test trees. main has nine files and publication/workflow infrastructure; it is not equivalent coverage. See utest branch differences.

Current Jira-driven test direction

  • HDRFP-12256: hardware-independent captured-behavior replay for C++/Rust adapter parity.
  • HDRFP-12248: virtual testing for Dell telemetry behavior.
  • HDRFP-11807: verify existing PAL/VFM failure cases through virtual tests.
  • HDRFP-12166: restore/setup AutoHLK for 26H2/Rust development; current evidence includes mixed builds.
  • HDRFP-12269: quality-verification planning only; no test system is claimed complete.

Result semantics

  • Passing source-derived tests prove the tested function/variant at the selected revision.
  • Passing C++ and Rust adapter workflows proves those workflow checks; it is not automatically complete parity.
  • A build failure, extraction failure, skipped test or missing artifact is not a pass.
  • Customer build success does not by itself prove runtime, firmware, HLK, security or accuracy acceptance.
  • Host mocks do not prove firmware command semantics; joint evidence requires a real or authoritative firmware counterpart.
  • Branch-specific expectations belong in explicit variants, not hidden conditional assumptions.

See GitHub Actions evidence for the bounded run inventory and exact interpretation rules.