Files
nickol-knx-mcp/nickol_knx_mcp
Nikolay MiroshnichenkoandClaude Opus 4.8 467772618e fix(#12 residual): recognise a shutter Function by its 1.008 move, not only a role token
Kris1166's real-file re-test of f20b2e4: main bug gone (light:0, all 29 RM → status) but
15/29 position feedbacks stayed lighting/not_mapped. Root cause (reproduced 1:1): those
actuator-level Functions carry blank or raw-GUID roles — no role token anywhere — so the
shutter-Function gate skipped them even though they hold a 1.008 up/down move. The
anomaly (2/3/1 promoting while its step/stop stayed unknown) is explained: its move has a
role; its step/stop is DPT 1.009 'Enable' (the reporter's own modeling slip, not a step).

- A shutter Function is now recognised by a shutter ROLE token OR by an UNCONTESTED 1.008
  move member (category==shutter; a 1.008 whose name pins another domain must not license
  its neutral siblings — gate-1 audit).
- is_step includes DPT 1.017 (trigger), which the cover builder already treats as a stop.
Gate-1: APPROVED; 0 category/kind/platform drift across 5 real .knxproj (1965 GA); the new
fixtures fail on the old code. Gate-2: the LLM council already endorsed the sibling-DPT
(1.008 move) signal as the primary structural evidence — this implements it.

Co-Authored-By: Claude Opus 4.8 <noreply@anthropic.com>
2026-09-04 21:33:25 +02:00
..