[GreenGrassBlueOcean]

What is the problem

Downstream virtual-device emulation and bus-state injection (the expose: engine in OpenWebNet-HA/MyHOME#750, verified on physical hardware in OpenWebNet-HA/MyHOME#753 and issuecomment-6277050571) require four protocol-layer capabilities that were previously missing or incomplete in OWNd:

1. Missing dimension-response (report_*) and dimension-write (write_* / set_*) frame builders:
   • OWNd previously provided actuator command and request builders (switch_on, set_brightness, raise_shutter, set_temperature, get_total_consumption, request_datetime), but lacked builders for dimension response frames (*#WHO*WHERE*DIM*VAL...##) and WHO=4 Dimension #0 temperature writes (*#4*WHERE*#0*TTTT##) needed when acting as a virtual responder or injecting external state onto the SCS bus across WHO=1 (Lighting), WHO=2 (Automation), WHO=4 (Thermoregulation), WHO=13 (Gateway Clock), and WHO=18 (Energy Management).
2. Missing incoming query / translation classification helpers on OWNMessage and OWN*Command:
   • A virtual responder listening on the event session must distinguish status requests (*#WHO*WHERE##), dimension requests (*#WHO*WHERE*DIM##), dimension writes (*#WHO*WHERE*#DIM*VAL##), dimension replies (*#WHO*WHERE*DIM*VAL##), and pushbutton command-translation frames (*WHO*1000#WHAT*WHERE##) without ad-hoc string regexes in downstream application code.
3. Hardware-specific gateway injection capabilities (Report Form vs. #DIM Write Form):
   • Real BTicino/Legrand gateways treat incoming un-hashed dimension response frames (*#WHO*WHERE*DIM*VAL##) vs. #DIM write frames (*#WHO*WHERE*#DIM*VAL##) differently across subsystems and hardware families:
     • WHO=1, WHO=2, WHO=4, WHO=13: On physical F461 (2.0.11), un-hashed response frames on the TCP command session are either NACKed (*#1*08*1*150*0##, *#2*25*10*10*50*0*0##) or answered as dimension reads (*#4*2*0*0999## returns Zone 2's measured *#4*2*0*0249##; *#13**0*22*51*00*999## returns the gateway's own time *#13**0*22*51*02*999##). The #DIM write frames (*#1*08*#1*150*0##, *#2*25*#11#001*50##, *#4*50*#0*0215##, *#13**#0*…##) are ACKed. Timings and sources are in the trace section below.
     • WHO=18: Report-form frames (*#18*51*113*450##, *#18*51*51*12345##, *#18*51*54*320##) are ACKed on physical F461 in 1.27–1.40 s, with no event after the ACK, so the value is not shown to reach the bus. The plain read *#18*51*113## is ACKed in 0.91 s without a reply too.
     • No other gateway family has a capture or a published oracle verdict for these frames.
4. Infeasible parameter and address bounds on expose builders (extending OpenWebNet-HA/OWNd#128):
   • New dimension-reply and dimension-write builders must enforce the same strict input boundaries established in OWNd#128: rejecting empty or *-polluted WHERE addresses, bool values (True/False disguised as int/float), non-finite floats (NaN, ±inf, raising ValueError rather than OverflowError), finite fractional floats on protocol-integer fields (e.g. 50.9 rather than silently truncating via int()), naive datetime/time inputs lacking an explicit time_zone, and out-of-range hardware parameters before constructing frames.

(Note: The command-session post-write retry guard for is_status_request=True was split into standalone PR OpenWebNet-HA/OWNd#131.)

---

How is it solved

1. Base Message Classification (OWNd/message/base.py):
   • Added is_status_request (*#WHO*WHERE##), is_dimension_request (*#WHO*WHERE*DIM##), is_dimension_writing (*#WHO*WHERE*#DIM*...##), is_dimension_reply (*#WHO*WHERE*DIM*VAL...##), and effective_what (unwrapping 1000#WHAT translation frames to integer WHAT) on OWNMessage.
2. Declarative Expose Builders & Query Properties (OWNd/message/):
   • WHO=1 (OWNLightingCommand):
     • is_status_query (*#1*WHERE##) and is_brightness_query (*#1*WHERE*1## or *#1*WHERE*4##).
     • report_brightness(where, level, transition=0) $\to$ *#1*WHERE*1*LEVEL*SPEED## (encoding 0..100% into OpenWebNet 100..200, where 100 = OFF).
   • WHO=2 (OWNAutomationCommand):
     • is_status_query (*#2*WHERE##) and is_shutter_status_query (*#2*WHERE*10##).
     • report_shutter_level(where, level, state=10, priority=0, info=0) $\to$ *#2*WHERE*10*STATE*LEVEL*PRIORITY*INFO## (supporting 0..100% or 255 unknown, states 10..15, and both integer 0..255 and $\le 3$-digit priority bitfield strings such as "001" / "000").
   • WHO=4 (OWNHeatingCommand & OWNHeatingEvent):
     • is_temperature_query (*#4*WHERE*0##).
     • encode_temperature(temperature) $\to$ 4-digit signed tenths (0TTT for $\ge 0^\circ\text{C}$, 1TTT for $< 0^\circ\text{C}$, validated to finite [-99.9, +99.9]°C, using Decimal with ROUND_HALF_UP and avoiding -0.0 sign artifacts).
     • report_temperature(where, temperature) $\to$ *#4*WHERE*0*TTTT## and write_temperature(where, temperature) $\to$ *#4*WHERE*#0*TTTT##.
     • OWNHeatingEvent: adds probe_temperature property (returning the scalar float temperature for 3-digit external probes 101–899 alongside secondary_temperature == [sensor, temp]) while preserving existing unique_id semantics ("4-#0" for WHERE="000" / zone == 0, "4-101" for WHERE="101").
   • WHO=13 (OWNGatewayCommand):
     • is_clock_query (*#13**0##, *#13**1##, *#13**22##).
     • set_datetime(dt, time_zone=None) (*#13**#22*...##), report_datetime(dt, time_zone=None) (*#13**22*...##), set_date(d, time_zone=None) (*#13**#1*...##), report_date(d, time_zone=None) (*#13**1*...##), set_time(t, time_zone=None) (*#13**#0*...##), report_time(t, time_zone=None) (*#13**0*...##), accepting datetime.tzinfo | str | None for time_zone, rejecting naive datetime/time inputs when time_zone is None, and rejecting UTC offsets that are not whole hours (Asia/Kolkata, Asia/Kathmandu, America/St_Johns), which the 0HH/1HH field cannot carry. The report_* forms are for answering on the event side. On the F461 command session they are answered as reads, so injection uses set_*.
   • WHO=18 (OWNEnergyCommand):
     • is_dimension_query (*#18*WHERE*DIM##), is_auto_update_command (*#18*WHERE*#1200#1*TIME##), and stream_duration (0..255 minutes).
     • report_active_power(where, watts) (*#18*WHERE*113*W##), report_total_energy(where, wh) (*#18*WHERE*51*Wh##), report_current_month_energy(where, wh, year=None, month=None) (*#18*WHERE*52[#YY#M]*Wh##), report_partial_month_energy(where, wh, month=None) (*#18*WHERE*53[#M]*Wh##), report_partial_day_energy(where, wh, day=None) (*#18*WHERE*54[#D]*Wh##), and report_streaming_state(where, duration, energy_type=1) (*#18*WHERE*1200#TYPE*TIME##).
3. Gateway Profile Capability Flags (OWNd/profiles.py):
   • Three tri-state flags on GatewayProfile, all bool | None = None. None means unverified: no capture or published oracle verdict exists for that family. Callers decide what to do with None.
     • supports_dimension_response_injection: whether reply-form frames (*#1*W*1*…##, *#2*W*10*…##, *#4*W*0*T##, *#13**D*…##) inject state on the command session.
     • supports_who4_command_injection: whether *#4*W*#0*TTTT## is accepted.
     • supports_who18_response_injection: whether *#18*W*DIM*VAL## is acknowledged. This is not proof that the value reaches the bus.
   • Only F461Profile sets them (False, True, True), from the live captures cited in its docstring. Every other profile, MH201Profile included, stays None.
4. Infeasible Input & Boundary Validation Across All Builders (OWNd/message/{lighting,automation,heating,energy,gateway}.py):
   • Enforces require_non_empty_where(where) across report_brightness, report_shutter_level, report_temperature, write_temperature, and all OWNEnergyCommand.report_* builders.
   • Rejects bool, non-finite floats (NaN, ±inf), and finite fractional floats (isinstance(x, float) and x % 1 != 0, e.g. 50.9) with ValueError rather than silently truncating via int() across all protocol-integer builder parameters, and enforces domain bounds: report_brightness (0 <= level <= 100, 0 <= transition <= 255), set_shutter_level (0 <= level <= 100), report_shutter_level (level in 0..100 or 255, 10 <= state <= 15, 0 <= priority, info <= 255, preserving $\le 3$-digit string priority), encode_temperature (math.isfinite, -99.9 <= temperature <= 99.9, ROUND_HALF_UP), OWNEnergyCommand.report_* (_validate_non_negative_energy, paired integral 2000 <= year <= 2099 and 1 <= month <= 12, integral 1 <= day <= 31, integral 1 <= energy_type <= 4, integral 0 <= duration <= 255), and OWNGatewayCommand (datetime.datetime / datetime.date / datetime.time type checks, explicit time_zone required for naive datetime/time, and 2000 <= year <= 2099).

---

State diagram

stateDiagram-v2
    direction TB

    state "Bus Query / Virtual-Device Injection Lifecycle" as ExposeFlow {
        [*] --> EventMonitor: Frame received on Event Session
        EventMonitor --> QueryDetected: is_status_query / is_brightness_query / is_shutter_status_query / is_temperature_query / is_clock_query / is_dimension_query
        EventMonitor --> StreamDetected: is_auto_update_command (*#18*W*#1200#1*T##)
        EventMonitor --> TranslationDetected: is_translation (*WHO*1000#WHAT*WHERE##) -> effective_what
        QueryDetected --> ProfileGate: Check GatewayProfile capabilities
        StreamDetected --> ProfileGate
        ProfileGate --> InjectDimWrite: WHO in {1, 2, 4} & supports_dimension_response_injection is not True\n-> set_brightness (*#1*W*#1*L*S##) / set_shutter_level (*#2*W*#11#001*P##) / write_temperature (*#4*W*#0*TTTT##)
        ProfileGate --> InjectWho18Reply: WHO == 18 & supports_who18_response_injection is not False\n-> report_active_power / report_total_energy / report_streaming_state (*#18*W*DIM*VAL##), ACK only
        ProfileGate --> InjectClockWrite: WHO == 13\n-> set_datetime (*#13**#22*...##) / set_time / set_date (report_* forms are answered as reads on F461)
        ProfileGate --> SkipUnsupported: flag is False (none set today)
        InjectDimWrite --> [*]
        InjectWho18Reply --> [*]
        InjectClockWrite --> [*]
        SkipUnsupported --> [*]
    }

Loading

---

How is it tested

• Unit & Boundary Tests (tests/test_message_protocol.py):
  • test_own_message_helper_properties_and_effective_what: verifies is_status_request, is_dimension_request, is_dimension_writing, is_dimension_reply, is_translation, effective_what, and all WHO=1, WHO=2, WHO=4, WHO=13, and WHO=18 builders and query detectors (including 3-digit priority bitfield strings like priority="001" on OWNAutomationCommand.report_shutter_level, datetime.tzinfo and IANA string timezones, and aware vs naive datetime.time).
  • test_expose_builders_infeasible_inputs: exhaustively tests rejection (ValueError) of bool inputs (True/False), NaN, ±inf, finite fractional floats (50.9, 0.5, 10.5, 2026.5, 18.5, 60.5, 1.5) across all integer-valued builder parameters, out-of-range brightness/shutter levels, out-of-range temperatures (< -99.9 or > 99.9), ROUND_HALF_UP half-way rounding (20.25 -> 0203, 20.35 -> 0204), negative energy values, invalid years/months/days, naive datetime/time without time_zone, invalid time_zone types, half- and quarter-hour UTC offsets, and empty/malformed WHERE addresses.
• Profile Capability Tests (tests/test_profiles.py):
  • test_profile_injection_capability_flags: F461 is False / True / True; every other profile and the generic fallback are None on all three flags.
• Golden Corpus Conformance & Builder Parity (tests/test_golden_conformance.py, tests/golden/frames/*.yaml):
  • Added 32 community-plant-capture fixtures across who01_lighting.yaml, who02_automation.yaml, who04_thermo.yaml, who13_gateway.yaml, and who18_energy.yaml (bringing the corpus to 164 validated frames, including the 2026-10-18 live F461 capture from MyHOME#753 and the 2026-10-19 F461 plant-check table in MyHOME#750 issuecomment-6277050571) with builder: blocks pinning OWNLightingCommand.report_brightness, OWNAutomationCommand.set_shutter_level/report_shutter_level, OWNHeatingCommand.report_temperature/write_temperature, OWNGatewayCommand.report_time/set_time/report_date/set_date/report_datetime/set_datetime, and OWNEnergyCommand.report_active_power/report_total_energy/report_partial_month_energy/report_partial_day_energy/report_streaming_state byte-for-byte against real hardware captures.
• Full Suite & Static Analysis:
  • 1,894 passed, 1 skipped locally on Python 3.14 (CI covers 3.11–3.14) with 100.0% statement and branch coverage on all modified modules (automation.py, base.py, energy.py, gateway.py, heating.py, lighting.py, profiles.py), mypy OWNd tests/typing clean (0 issues), ruff check clean.

---

Where are the traces

1. Real Physical Bus Captures by Community Contributors (community-plant-capture)

Every frame shape built or classified in this PR is backed by authentic physical bus captures contributed by the community and pinned with builder: blocks in tests/golden/frames/ (provenance recorded in tests/golden/SOURCE.yaml):

Subsystem	Real Community Capture Files (MyHOME/tests/fixtures/traces/ & OWNd/tests/golden/frames/)	Contributor & Hardware Plant	Captured Wire Frames Verified by This PR
Live F461 (2.0.11) Expose (WHO=1, 2, 4, 13, 18), F411/4 Lights & F411/2 5 s Shutter (2026-10-18 capture, 2026-10-19 plant check)	f461_live_2026-10-18/myhome_trace_F461_live_2026-10-18.json (OpenWebNet-HA/MyHOME#753) Plant-check table in MyHOME#750 issuecomment-6277050571 (text only, no capture file) OWNd/tests/golden/frames/who02_automation.yaml OWNd/tests/golden/frames/who04_thermo.yaml	@gdluck & @GreenGrassBlueOcean (MyHOME#753 & MyHOME#750) — BTicino F461 (FW 2.0.11, WHO=13 code 200, WHO=1013 134.15.5.0) + BT-F411/4 (06, 07), BT-F411/2 5 s shutter (05), KW4691 + F430/2 + F430/4 (Zone 2)	Physical shutter 05 (*2*1000#0*05## $\to$ *2*0*05## $\to$ *2*1000#1/2*05## $\to$ *2*1/2*05## $\to$ 5 s auto-stop *2*0*05##); virtual cover 25 event-session translation bursts before NACK on cmd. 2026-10-18 trace: *#1*08*1*150*0## NACK 0.055 s vs *#1*08*#1*150*0## ACK 0.126 s; *#2*25*10*10*50*0*0## (priority 0) NACK 0.055 s; *#4*2*0*0999## answered as a read with *#4*2*0*0249## (0.22 s); *#4*50*0*0215## NACK after 2.13 s vs *#4*50*#0*0215## ACK 0.155 s; *#13**0*22*51*00*999## answered as a read with *#13**0*22*51*02*999## (0.116 s) vs *#13**#0*22*51*00*999## ACK 0.071 s; *#18*51*113*450## / *#18*51*51*12345## / *#18*51*54*320## ACK 1.27–1.40 s, *#18*51*#1200#1*1/0## ACK 0.14–0.15 s, no events after. 2026-10-19 plant check (comment table): *#1*08*1*150*0## NACK 6–10 ms vs #1 write ACK 83 ms; *#2*25*10*10*50*001*0## NACK 10 ms vs *#2*25*#11#001*50## ACK 83 ms, echoed *#2*25*#11#001#1*50##; *#4*50*0*0215## read, NACK after 4.2 s vs *#4*50*#0*0215## ACK 88 ms; *#18*51*113*450## ACK 1.2–1.7 s; the WHO 13 write frame sets the gateway clock.
WHO=1 Lighting (Dim 1 report_brightness & 1000#WHAT translation)	issue_466/myhome_trace_MH202_all_2026-09-26T16-59-17.json issue_466/myhome_trace_F454_f418u2_dimmer_2026-09-27T08-58-54.json issue_466/myhome_trace_MH202_f418u2_dimmer_2026-09-27T09-43-45.json OWNd/tests/golden/frames/who01_lighting.yaml	@anotherjulien (MyHOME#466 comments 5849027587, 5854469497, 5854805415) — BTicino MH202 (FW 2.6.13) + F454 (FW 2.0.5) + F418U2 2-ch dimmer (WHERE 14, 32)	Pushbutton translations *1*1000#1*14##, *1*1000#0*14##, *1*1000#30*14##, *1*1000#31*14## (effective_what) and Dim 1 replies *#1*32*1*100*0## (OFF), *#1*32*1*130*5##, *#1*14*1*130*5##..*#1*14*1*200*5## (report_brightness)
WHO=2 Automation (Dim 10 report_shutter_level, Dim 11 set_shutter_level & 1000#WHAT translation)	issue_304/myhome_trace_MyHomeServer1_who2_2026-10-08T10-49-50.json issue_304/myhome_trace_MyHomeServer1_who2_2026-10-08T10-49-12.json issue_466/myhome_trace_MH202_all_2026-09-26T16-59-17.json issue_466/myhome_trace_MyHomeServer1_scenarios_cover_diagnostic_2026-09-27T13-47-33.json f461_live_2026-10-18/myhome_trace_F461_live_2026-10-18.json OWNd/tests/golden/frames/who02_automation.yaml	@f18m (MyHOME#304 comment 6060208135 — MyHomeServer1 FW 2.53.16 + LN-4660M2 shutters 91..95), @anotherjulien (MH202 + F454, shutters 15, 22, 31), @TheDarkWizard (MyHOME#466 comment 5857411555), @gdluck & @GreenGrassBlueOcean (F461 FW 2.0.11 + BT-F411/2 05, MyHOME#753 & MyHOME#750)	*#2*95*#11#001*29## $\to$ *2*1000#0*95## $\to$ *#2*95*10*12*40*000*0## (closing) $\to$ *#2*95*10*10*29*000*0## (stopped at 29%); *#2*25*#11#001*50## $\to$ *#2*25*#11#001#1*50## (F461 monitor translation); *#2*31*10*10*40*001*0## (priority="001" preserved); *#2*15*10*11*0*000*0## (opening); *#2*44*10*10*100*000*0##; physical F411/2 05 (*2*1000#0*05##, *2*1000#1*05##, *2*1000#2*05##)
WHO=4 Heating (Dim 0 report_temperature, write_temperature & 3-digit external probes)	issue_466/myhome_sweep_MH200N_all_2026-09-25T14-40-10.json issue_466/myhome_trace_MyHomeServer1_all_2026-09-29T17-57-37.json issue_466/myhome_trace_MyHomeServer1_scenarios_cover_diagnostic_2026-09-27T13-47-33.json f461_live_2026-10-18/myhome_trace_F461_live_2026-10-18.json OWNd/tests/golden/frames/who04_thermo.yaml	@caiosweet (MH200N FW 1.1.8, MyHOME#466 comment 5834435606), @gdluck (MyHomeServer1 FW 3.87.13 & F461 FW 2.0.11, MyHOME#753), @TheDarkWizard (MyHomeServer1 FW 2.38.9)	External 3-digit probes *#4*105*0*0296## (29.6 °C on probe 105) and *#4*169*0*...## (probe 169), zone temperatures *#4*2*0*0271## (27.1 °C) and *#4*2*0*0249## (24.9 °C on F461 KW4691), and Dimension #0 write *#4*50*#0*0215## (write_temperature("50", 21.5) ACKed in 88 ms / 155 ms on F461)
WHO=13 Gateway Clock (Dims 0, 1, 22 set_* & report_*)	issue_466/config_entry-myhome_F455.json issue_466/myhome_trace_MH202_all_2026-09-26T16-59-17.json issue_466/config_entry-myhome_F454_step_commands.json f461_live_2026-10-15/myhome_trace_F461_live_2026-10-15.json f461_live_2026-10-18/myhome_trace_F461_live_2026-10-18.json issue_466/myhome_sweep_MH200N_all_2026-09-25T14-40-10.json OWNd/tests/golden/frames/who13_gateway.yaml	@lionelser (F455), @anotherjulien (F454 + MH202), @caiosweet (MH200N), @gdluck & @GreenGrassBlueOcean (F461 FW 2.0.11 & MH200 FW 2.1.0, MyHOME#753)	*#13**0## $\to$ *#13**0*20*54*00*002##, *#13**#0*11*30*00*002##, *#13**1*04*15*10*2026##, *#13**#1*03*07*10*2026##, *#13**22*16*37*03*001*05*25*09*2026##, *#13**#22*17*33*54*002*05*11*09*2026##
WHO=18 Energy (Dim 1200 stream lifecycle + Dims 113, 51, 53, 54)	issue_466/myhome_trace_F454_who18_2026-10-08T19-30-16.json issue_466/myhome_trace_MyHomeServer1_all_2026-10-01T07-59-11.json issue_466/myhome_sweep_F454_all_2026-09-26T16-59-13.json issue_466/myhome_sweep_MH200N_all_2026-09-25T14-40-10.json f461_live_2026-10-18/myhome_trace_F461_live_2026-10-18.json OWNd/tests/golden/frames/who18_energy.yaml OWNd/tests/test_who18_auto_update_reply.py	@xtimmy86x (MyHOME#661 comment 6067730589, F454 FW 2.0 + F520 51..511), @Interstellar0verdrive (MyHomeServer1 + F520 51, 52), @anotherjulien (F454 + F520 52), @caiosweet (MH200N + F520 51..57), @gdluck & @GreenGrassBlueOcean (F461 FW 2.0.11, MyHOME#753), MyHOME#669 (7X#0)	TX *#18*56*#1200#1*2## $\to$ RX *#18*56*1200#1*2## (~235 ms) $\to$ *#18*56*113*9##..*#18*51*113*572## $\to$ expiry RX *#18*56*1200#1*0## (~126.18 s); *#18*51*113*450##, *#18*51*51*12345##, *#18*51*54*320##, *#18*51*#1200#1*1##, *#18*51*#1200#1*0## ACKed on live F461; *#18*51*51*23791364## (Dim 51); *#18*52*53*1## (Dim 53); *#18*51*54*13165## (Dim 54); *#18*77#0*113*161##

2. Emulated Gateway Firmware Oracle Traces (own-firmware-oracle)

tests/golden/firmware_oracle.json pins only these frames related to this PR: *#2*31*10##, *#2*31*#11#001*40##, *#4*1*0##, *#18*51*113##, *#18*51*51##, *#18*51*53##, *#18*51*54##, *#18*51*#1200#1*255##, *#18*51*#1200#1*0##. These are queries and writes, not report-form injection. No published oracle verdict exists for *#4*W*#0*TTTT##, *#4*W*0*TTTT## or any WHO 18 report frame, so no profile flag rests on the oracle. An earlier revision of this PR set F453/MH200N/MH200 overrides from an unpublished local replay; those overrides were removed in e374f3c.

---

What does the Encyclopedia / Firmware Oracle / Live F461 Plant say

1. Grammar & Meaning (OpenWebNet-Encyclopedia + Community Hardware Captures):
   • WHO=1 Dimension 1 (*#1*WHERE*1*LEVEL*SPEED## vs *#1*WHERE*#1*LEVEL*SPEED##): LEVEL on the wire is 100..200 corresponding to 0%..100% (100 = OFF, 200 = 100%). On physical F461 (2.0.11), *#1*08*1*150*0## (report form) is NACKed (0.055 s in the 2026-10-18 trace, 6–10 ms in the 2026-10-19 plant check), whereas *#1*08*#1*150*0## (set_brightness #1 write form) is ACKed (0.126 s / 83 ms).
   • WHO=2 Dimension 10 (*#2*WHERE*10*STATE*LEVEL*PRIORITY*INFO##), Dimension 11 (*#2*WHERE*#11#001*LEVEL##) & Command Translation (*2*1000#0/1/2*WHERE##): STATE is 10 (stopped), 11 (opening), 12 (closing), 13 (step-up), 14 (step-down), 15 (PUL); LEVEL is 0..100 (%) or 255 (unknown position); PRIORITY is a 3-digit bitfield (001 / 000) on physical actuators or integer 0..255. On physical F461 (2.0.11), the report form is NACKed: *#2*25*10*10*50*0*0## in 0.055 s (2026-10-18 trace) and *#2*25*10*10*50*001*0## in 10 ms (2026-10-19 plant check). By contrast, *#2*25*#11#001*50## (set_shutter_level #11#001 write form) is ACKed in 83 ms and echoed on evt as *#2*25*#11#001#1*50## (@gdluck, MyHOME#750 issuecomment-6277050571).
   • WHO=4 Dimension 0 (*#4*WHERE*0*TTTT## vs *#4*WHERE*#0*TTTT##): 4-digit signed tenths of a degree Celsius (0TTT for $\ge 0^\circ\text{C}$, 1TTT for $< 0^\circ\text{C}$). External probes use WHERE = 000 (mapped to 4-#0) or 3-digit pZZ (101–899). On physical F461 (2.0.11), bt_termo parses *#4*W*0*TTTT## without #0 as a Dimension 0 read request returning Zone 2's *#4*2*0*0249##, while OWNHeatingCommand.write_temperature("50", 21.5) (*#4*50*#0*0215##) is ACKed (0.155 s / 88 ms).
   • WHO=13 Dimensions 0, 1, 22: Time (H*M*S*TZ), Date (0W*D*M*Y), and Combined Datetime (H*M*S*TZ*0W*D*M*Y), where TZ is 0HH ($\text{UTC}+$) or 1HH ($\text{UTC}-$). What the captures show for that field:
     • Filled with 0HH: F453AV (001), F454 (001, 002), MH200 (000, 001, 002), MH200N (001) and MH202 (000, 002).
     • Empty: MH201 (3.6.27 *#13**0*21*31*40*## and *#13**22*21*31*43**03*14*10*2026##, the mh201_physical_plant fixture *#13**22*00*22*43**04*17*09*2026##, and the 3.3.6 issue_466 trace *#13**0*21*07*30*##), and F455.
     • 999 (unconfigured): F461 in the 2026-10-15 and 2026-10-18 traces, and one F454 (issue_297).
     • Not captured: MyHomeServer1 has no clock frame in the fixtures, and no plant capture shows a 1HH value; the only one is the openwebnet4j fixture *#13**22*09*37*30*102*03*01*05*2019##.
     • Command session: on the F461, *#13**0*22*51*00*999## is answered as a read with the gateway's real time, like WHO 4, while *#13**#0*…## is ACKed and sets the clock.
   • WHO=18 Dimensions 51, 52, 53, 54, 113, 1200: On physical F461, report-form frames (*#18*51*113*450##, *#18*51*51*12345##, *#18*51*54*320##) are ACKed in 1.27–1.40 s (1.2–1.7 s in the plant check) with no event after, and the plain read *#18*51*113## is ACKed in 0.91 s without a reply. So the ACK does not show the value reaching the bus (supports_who18_response_injection = True on F461 means acknowledged).
2. Firmware Oracle Verdicts (own-firmware-oracle):
   • None of the three profile flags rests on an oracle verdict. Only F461Profile sets them, from the live captures; every other profile is None (unverified) until a capture or a published verdict exists.

[codecov-commenter]

⚠️ Please install the to ensure uploads and comments are reliably processed by Codecov.

Codecov Report

✅ All modified and coverable lines are covered by tests.

📢 Thoughts on this report? Let us know!

[fedem95]

I found one validation gap: finite fractional values are silently truncated by int() instead of being rejected. For example, report_brightness(..., 50.9) builds a frame with level 50; the same issue affects integer-valued parameters in the other builders. Since these fields represent protocol integers, please reject non-integral values and add regression tests before approval.

[GreenGrassBlueOcean]

Thanks for catching that, @fedem95!

Fixed in d478495:

• All integer-valued builder parameters across OWNLightingCommand.report_brightness (level, transition), OWNAutomationCommand.set_shutter_level (level), OWNAutomationCommand.report_shutter_level (level, state, priority, info), OWNEnergyCommand.report_current_month_energy (year, month), OWNEnergyCommand.report_partial_month_energy (month), OWNEnergyCommand.report_partial_day_energy (day), and OWNEnergyCommand.report_streaming_state (duration, energy_type) now guard against isinstance(x, float) and x % 1 != 0, rejecting both non-finite floats (NaN, ±inf) and finite fractional floats (e.g. 50.9) with ValueError instead of silently truncating via int().
• Added regression tests covering finite fractional float inputs across all affected builders in tests/test_message_protocol.py::test_expose_builders_infeasible_inputs.

[GreenGrassBlueOcean]

Follow-up update (@fedem95):

Also pushed 42280af on this branch to address a few additional review findings:

• Split out the OWNCommandSession retry-guard fix into standalone PR fix(connection): only retry idempotent queries after mid-send transport loss #131 (ix/non-query-no-retry) so this PR stays focused on the message builders and profile capability flag.
• Timezone & naive datetime/ ime handling (OWNd/message/gateway.py): OWNGatewayCommand clock builders (set_datetime,
  eport_datetime, set_date,
  eport_date, set_time,
  eport_time, set_datetime_to_now, set_date_to_today, set_time_to_now) now accept datetime.tzinfo | str | None (avoiding synchronous ZoneInfo(str) disk loads inside the event loop when a zinfo is passed) and reject naive datetime/ ime inputs with ValueError when ime_zone is None instead of silently encoding UTC

[fedem95]

Thanks for d478495 and the rebase: report_brightness('21', 50.9) and True now raise, 50.0 still builds *#1*21*1*150*0##. Tested 30410c6: 1893 passed, CI green. The only change from c68ffd5 is the gateway.py conflict resolution, which keeps master's _resolve_tzinfo, and the set_*_to_now builders work with both str and tzinfo. The F461 2026-10-18 trace backs the main points: *#1*08*#1*150*0## ACK in 0.126 s, *#4*2*0*0999## answered as a read with *#4*2*0*0249##, *#4*50*#0*0215## ACK in 0.155 s, WHO 18 reports ACKed in 0.14-1.40 s. Before approving:

1. The F453/MH200N supports_who4_command_injection=False and MH200/MH200N supports_who18_response_injection=False flags are credited to the oracle, but I can't find the evidence. *#4*1*#0*0215## isn't in tests/golden/firmware_oracle.json, and neither is any *#4*W*#0*TTTT##, *#4*W*0*TTTT## or WHO 18 report frame. The 88 F453/MH200/MH200N/F461 TSVs on oracle main have none either (the only 0215 is *#4*#23#1*#14*0215*1##). Where are those verdicts? If they aren't published yet, I'd keep the defaults and leave the overrides out until they are.
2. Figures that don't match the 2026-10-18 trace: *#1*08*1*150*0## is NACKed in 0.055 s (not 5 ms), and the WHO 2 report there is *#2*25*10*10*50*0*0## (priority 0, NACK in 0.055 s), not …*001*0## in 82 ms. If those come from the 2026-10-20 capture, could it be added to the corpus?
3. On the F461, WHO 13 report frames are answered as reads too (*#13**0*22*51*00*999## → ACK plus the real time), like WHO 4; the diagram lists report_datetime as a valid injection.
4. MH201Profile inherits supports_who4_command_injection=True and supports_who18_response_injection=True with no MH201 evidence. Maybe say "unverified" for profiles without data rather than defaulting to True?

Non-blocking: a half-hour offset is still truncated silently (Asia/Kolkata → TZ 005); rejecting it might be safer.

[GreenGrassBlueOcean]

Thanks, all five points checked against the F461 trace and the oracle checkout; fixed in e374f3c and in the description.

1. You're right: there is no published verdict. The F453/MH200N WHO 4 and MH200/MH200N WHO 18 overrides came from a local replay that never made it into own-firmware-oracle or firmware_oracle.json. Both overrides are gone.
2. Confirmed. The 2026-10-18 trace has *#1*08*1*150*0## NACK in 0.055 s and *#2*25*10*10*50*0*0## (priority 0) NACK in 0.055 s. The 5 ms / …*001*0## / 82 ms figures were a mix-up with the 2026-10-19 plant-check table in MyHOME#750 (issuecomment-6277050571), which reports 6–10 ms and *001*0## at 10 ms. That table is text only, with no capture file, so there is nothing to add to the corpus. The description now cites each figure with its source, and the "2026-10-20 capture" reference is gone.
3. Confirmed: *#13**0*22*51*00*999## → ACK plus *#13**0*22*51*02*999##. The diagram now injects WHO 13 through set_* only, and the description says report_* is answered as a read on the F461. I also noted that the WHO 18 report ACKs (1.27–1.40 s) come with no event after, and that the plain *#18*51*113## read is ACKed in 0.91 s too, so supports_who18_response_injection means "acknowledged", not "reaches the bus".
4. The three flags are now bool | None, defaulting to None (unverified). Only F461Profile sets them (False / True / True), with the trace frames in its docstring. MH201 and every other profile stay None.

Non-blocking: done. _encode_tz_offset now raises ValueError for offsets that are not whole hours (Asia/Kolkata, Asia/Kathmandu, America/St_Johns), including through set_*_to_now.

1894 passed locally; ruff and mypy clean.

[fedem95]

Thanks, all five points are addressed in e374f3c. Merged with current master (a2ce7fb): no conflicts, 1893 passed, CI green. The flags now default to None, only F461Profile sets them, and its docstring matches the 2026-10-18 trace (*#4*50*#0*0215## ACK in 0.155 s, the WHO 18 reports ACKed in 1.27-1.40 s with no event after). MH201Profile stays None on all three. Non-whole-hour offsets now raise, also through set_time_to_now. Approving.

One last fix to the description, no code change needed: it says the 0HH/1HH TZ field is "confirmed across physical … MH201 … captures", but every MH201 capture we have leaves it empty: my 3.6.27 answers *#13**0*21*31*40*## and *#13**22*21*31*43**03*14*10*2026##, the mh201_physical_plant fixture has *#13**22*00*22*43**04*17*09*2026##, and the 3.3.6 trace in issue_466 has *#13**0*21*07*30*##. The F461 in the same list answers 999 (unconfigured) in the 2026-10-18 trace.