Skip to content

COMPANY_PARITY_AUDIT.md

Latest commit

 

History

History
39 lines (33 loc) · 2.21 KB

COMPANY_PARITY_AUDIT.md

File metadata and controls

39 lines (33 loc) · 2.21 KB
Raw

Company process audit

Current architecture - Python WolfEye / native BeachEye - 2026-10-21

User-directed scope: keep BeachEye's working DLL, remove the WolfEye core DLL. Company majidnow/wolfeye master was verified through GitHub CLI at 7e58f4af439c6fa9c9db51e4558936264401732a.

  • wolfeye_beacheye.dll, ABI 12, retains device transport, settings serializers, point maps, measurements and event decoding. Its source implementation was not changed for this migration.
  • services/wolfeye.py directly reads SIGNAL float32 and ADC uint16 using struct, following FaultViewer.cpp::load_signal() with INFO-selected 32/64-sample channel strides. Multiple ADC cycles remain intact; padded 32-sample SIGNAL responses use only their valid prefix.
  • TCC follows TCCCalculator.cpp and the actual IEC_Def / ANSI_Def field order in company headers, including float32 intermediate rounding, uint32 milliseconds, Ta, Tmin and Tmax.
  • services/ar_rules.py ports SettingsView.cpp stage transitions and disabled-button flags. No algorithm calls a WolfEye DLL.
  • Get signal dispatches independently of pending settings-editor validation.
  • CMake and packaging no longer build or include wolfeye_core.dll. BeachEye remains native and must match the Python/EXE architecture. 32 samples per cycle is independent of the process's bitness.

Validation: 191 existing tests passed, plus four new regression tests. Golden outputs cover 12,000 AR transitions and 6,000 TCC values from the previous unchanged company C++ functions. Signal tests also exercise fragmented TCP responses through BeachEye into both 32/64-sample Qt viewers. Python 3.8 uses an equal-length zip compatibility helper for signal analysis and socket.timeout handling for EmWolf; numerical and scenario rules remain unchanged.

The development WolfEye C++ checkout is now a reference for these Python ports. Editing its algorithms does not change the active implementation. -UpdateWolfEye refreshes native-support source provenance; a new algorithm revision requires an explicit Python port and parity review.

Both x64/Qt6 and x86/Qt5 builds are supported. Qt 5 exec_() calls are adapted centrally; large numeric values use explicit float conversion in double editors.