[gdluck]

Requested by A

The MyHOME companion of OpenWebNet-HA/OWNd#142 (fix(profiles): drop WHO 16 from the F461 and MyHomeServer1 profiles). Sequencing: merge this → merge OpenWebNet-HA/OWNd#142 → OWNd release → bump the OWNd pin here, and scripts/sync_documentation.py --update regenerates the two docs rows together with the pinned_whos fallback below (one edit in scripts/update_gateway_profiles.py).

Rebase after #794 (2026-10-26)

Rebased on v2-phase1-architecture at a27b32f6 (after #794, #793, #797, #798, #801 and #804); the two commits of the first push are kept and reconciled with #794 in place, three follow-up commits on top. Base SHA a27b32f6, branch head bfc1c270.

What #794 and this PR each bring, and how they are reconciled (the three conflicts: .github/release-notes/2.0.0b16.md, scripts/update_gateway_profiles.py, tests/test_gateway_profiles.py):

• Kept from fix(gateway): skip the startup requests the firmware refuses, poll each zone's temperature on SCS-only gateways #794: the split "Profile Subsystems (advertised)" / "Startup Sweep Requests (sent)" columns, the per-row sweep_omits (the general requests the firmware drops, *#4*0## / *#16*0*5##), its notes per row, the MH200 / MH200N split and the H4684 row, the sweep_omits validator and tests/test_startup_requests_per_firmware.py::test_hardware_compatibility_table_matches_the_sweep (unchanged: it compares the calibrated rows with MyHOMEGatewayHandler._discovery_frames()), plus fix(climate): do not poll humidity (dimension 60) where the emulated firmware did not forward it #793's KNOWN_GATEWAY_MODELS resolver in gateway.py.
• Kept from this PR: the "advertised" column is derived from the installed OWNd profile (discovery_whos_for, the rule of the startup gate), no row carries a static queried_whos; pinned_whos is the no-OWNd fallback (the rows the pinned OWNd 2.0.0b11 yields; the quality-scale, test-coverage and docs CI jobs run the docs check without OWNd); notes_without_who16 on the F461 / MyHomeServer1 rows switches the wording with the profile.
• The join: fix(gateway): skip the startup requests the firmware refuses, poll each zone's temperature on SCS-only gateways #794's validator rule "queried_whos equals the profile" is replaced by the derivation plus the pinned_whos fallback. When a WHO is not advertised, a general request of it is no omission of the sweep (nothing is sent for that WHO at all), so calibrate_discovery_with_ownd() drops that frame from the generated row: with OWNd How to install this extension properly? #142 on PYTHONPATH the F461 row becomes | WHO 1, 2, 4, 18 | WHO 1, 2, 18 (firmware drops the shutters does not work in homekit #40##) | ... Sound is WHO 22 only ... |, exactly what _discovery_frames() sends (*#1*0##, *#2*0##, *#18*51*51##, *#18*51*113##) and what the unchanged fix(gateway): skip the startup requests the firmware refuses, poll each zone's temperature on SCS-only gateways #794 sweep test checks. notes_without_who16 carries fix(gateway): skip the startup requests the firmware refuses, poll each zone's temperature on SCS-only gateways #794's firmware facts (*#4*0## dropped, zones polled with *#4*<WHERE>##, the oracle firmware ids and the live confirmations) with the WHO 22 wording.
• verify_profiles_against_oracle() after the rebase: every row must name a real OWNd profile (not the generic fallback); pinned_whos must be WHO numbers of the startup sweep, in sweep order without repeats, and must equal the installed profile's advertised WHOs except by WHO 16 on a row that declares notes_without_who16 (the one profile change the fallback is meant to absorb between the pinned release and the next, fix(profiles): drop WHO 16 from the F461 and MyHomeServer1 profiles OWNd#142), so the check is green with the pinned OWNd and with the OWNd How to install this extension properly? #142 checkout alike and flags a stale pin at the pin bump; sweep_omits must be *#4*0## / *#16*0*5## of a WHO in pinned_whos (fix(gateway): skip the startup requests the firmware refuses, poll each zone's temperature on SCS-only gateways #794's rule, now against the pin: a leftover *#16*0*5## is flagged once the pin drops WHO 16). A malformed pinned_whos is a violation rather than an exception, and such a row renders none cells with the plain notes.
• Release notes: fix(gateway): skip the startup requests the firmware refuses, poll each zone's temperature on SCS-only gateways #794's bullets kept, this PR's "Docs tooling" bullet added under "Other changes" (reworded for the split columns).

Suite against both OWNd checkouts (venv with HA 2026.10.2, Python 3.14, pytest -n 4; the OWNd checkout on PYTHONPATH):

OWNd on PYTHONPATH	Result	Notes
master 87cafc3 (#136 merged; the behaviour of the 2.0.0b11 pin)	4965 passed, 1 skipped, verify_ownd_coverage.py 100 % (16,770 / 16,770 statements, 0 missing)	committed tree
fix/profiles-f461-mhs1-no-who16 95dc8a5 (OpenWebNet-HA/OWNd#142, on f04d8aa)	4964 passed, 1 failed, 1 skipped	committed tree: the one failure is tests/test_sync_documentation.py::test_check_all_documentation_real_tree, "Gateway Discovery table out of sync" (the F461 / MyHomeServer1 rows the pin bump regenerates); sync_documentation.py --check reports the same two rows and nothing else (no oracle violation)
fix/profiles-f461-mhs1-no-who16 95dc8a5, pin-bump rehearsal	4965 passed, 1 skipped	pinned_whos of the two rows set to [1, 2, 4, 18], *#16*0*5## removed from their sweep_omits, sync_documentation.py --update: only the F461 and MyHomeServer1 rows change (docs/getting-started/hardware-compatibility.md: 2 lines, `WHO 1, 2, 4, 18

Checks on the final head with OWNd master: ruff check ., scripts/typing_ratchet.py (0 errors, baseline 0), scripts/verify_ha_standards.py (with OWNd and without), scripts/sync_documentation.py --check (with OWNd and without), scripts/manage_translations.py check. tests/test_gateway_profiles.py now has 25 tests; test_discovery_table_with_ownd_matches_the_committed_docs compares row by row and lets only the rows whose profile differs from pinned_whos differ (and requires them to), so it is right before and after the pin bump while --check keeps reporting the drift until --update is run.

Independent review (focus: the validator still rejects a wrong sweep_omits; the derivation cannot silently drop a WHO the sweep sends; the no-OWNd fallback): agreed on all three for the shown rows; found that pinned_whos itself was not checked against any profile (a mistyped pin would have passed), that a malformed pin raised instead of being flagged, that a repeated or reordered pin passed as a set, and that an empty row rendered an unfinished sent cell with the WHO 22 wording. All four are fixed in the two follow-up commits and pinned by tests. Accepted residual, by design: on the two rows that declare notes_without_who16 the validator cannot tell "pin with WHO 16 against OWNd #142" from "pin without WHO 16 against the pinned release" (both OWNds report __version__ 2.0.0b11); the WHO 16 cell of those rows is pinned by test_discovery_table_without_ownd_matches_the_committed_docs (pin equals the committed docs) and by the sweep_omits rule (*#16*0*5## requires WHO 16 in the pin), not by the profile comparison.

---

Review follow-up (review 5651153472)

• fd20792: negative test test_oracle_check_rejects_a_row_that_resolves_to_the_generic_profile for the GenericGatewayProfile guard of verify_profiles_against_oracle(): a row whose profile_name is Generic or an unknown model string reports resolves to the generic profile for that row and for no other; it fails with the previous is None check. Full suite 4991 passed / 1 skipped at 100 % coverage. Marked ready for review as asked.

Review follow-up (CI red on the first push)

• Cause: the quality-scale, test-coverage and docs jobs run scripts/verify_ha_standards.py / scripts/sync_documentation.py --check without OWNd installed ([OWND] Client library verification skipped (OWNd not in path: No module named 'OWNd') in the same log). The first push removed the static queried_whos and printed unknown (OWNd profile not available) for every row when OWNd is not importable, so the regenerated table differed from the committed one: [RULE_DOCS_ANTI_DRIFT] Gateway Discovery table out of sync in docs/getting-started/hardware-compatibility.md. Locally the check had only been run with OWNd on PYTHONPATH.
• Fix: each discovery row carries pinned_whos, the row the pinned OWNd (OWNd==2.0.0b11 in manifest.json) yields; it is used only when OWNd is not importable, so the generated table is byte-identical with and without OWNd. With OWNd importable the row is still derived from the installed profile. Reproduced and fixed locally: sync_documentation.py --check now passes both in a venv without OWNd (pyyaml + ruff only, as in the CI jobs) and with OWNd master on PYTHONPATH; tests/test_gateway_profiles.py::test_discovery_table_without_ownd_matches_the_committed_docs pins the fallback to the committed table and ::test_discovery_table_with_ownd_matches_the_committed_docs pins the derived table to it, so installed profile == committed table == pinned_whos is asserted explicitly.
• Oracle double-check (the maintainer refreshed own-firmware-oracle main to 72759e8, fix(oracle): write a past date in who13-clock so the DIM 1 verdict does not depend on the day own-firmware-oracle#61 and fix(results): refresh 7 stale sound-who22 TSVs and MCP index (#58) own-firmware-oracle#62): the WHO 16 verdicts for the F461 020011 and the MyHomeServer1 028206 are unchanged, see "What does the Firmware Oracle / Encyclopedia say" below.
• Also: verify_profiles_against_oracle() now flags a discovery row whose profile_name resolves to OWNd's generic profile (get_gateway_profile() never returns None, so the earlier is None check could not fire).

What is the problem

scripts/update_gateway_profiles.py keeps a static queried_whos list per startup discovery row next to the list it derives from the installed OWNd profile, and verify_profiles_against_oracle() compares the two. The derived list is what the generated docs/getting-started/hardware-compatibility.md table shows and what the runtime sends (the startup gate in gateway.py asks profile.supports_who(who) for every discovery WHO; #794 reads the same flag for the refused requests). When an OWNd profile changes, the static copy goes stale: with OpenWebNet-HA/OWNd#142 the F461 and MyHomeServer1 rows become [1, 2, 4, 18], the oracle check reports Discovery row **F461** queried_whos=[1, 2, 4, 16, 18] disagrees with F461 profile [1, 2, 4, 18], sync_documentation.py --check fails and tests/test_gateway_profiles.py asserts the old literal row. Hard-coding the new rows instead would be red against the pinned OWNd 2.0.0b11 today.

Refs OpenWebNet-HA/OWNd#142; #794 (the runtime gate); #424 / #459 (the F461 / MyHomeServer1 plants that speak WHO 22 only, and the MyHomeServer1 "firmware discards 5, 16, 22" note). The combined test build (combined/all-prs-2026-10-24, commit 4a79e8cb) showed which files and tests are involved; its hard-coded rows are not used here.

---

What is the solution

1. scripts/update_gateway_profiles.py:
   • GATEWAY_DISCOVERY_METADATA rows carry model, profile_name, pinned_whos (the no-OWNd fallback, see below) and notes; no queried_whos.
   • New discovery_whos_for(profile): [w for w in STARTUP_DISCOVERY_WHOS if profile.supports_who(w)], the rule MyHOMEGatewayHandler._profile_supports_who applies at startup; [] for anything without supports_who.
   • calibrate_discovery_with_ownd() fills queried_whos from the installed profile for every row and keeps only the sweep_omits of advertised WHOs (a request of a WHO that is not advertised is no omission: nothing is sent for it); the F461 and MyHomeServer1 rows carry a notes_without_who16 text that replaces notes when the row does not carry WHO 16 ("Sound is WHO 22 only (no bt_difson; every WHO 16 frame refused ...)", with fix(gateway): skip the startup requests the firmware refuses, poll each zone's temperature on SCS-only gateways #794's *#4*0## facts), so the wording follows the profile too.
   • Without an importable OWNd (the quality-scale, test-coverage and docs CI jobs) every row falls back to pinned_whos, the row the pinned OWNd 2.0.0b11 yields, so the generated table is the same either way; nothing else is guessed. At a pin bump that changes a profile, pinned_whos is edited together with sync_documentation.py --update (the tests below fail until both are done).
   • verify_profiles_against_oracle() no longer compares a static list with the profile; it checks that every discovery row names a real OWNd profile (not the generic fallback), that pinned_whos is a sweep-ordered list of startup WHOs equal to the installed profile except by WHO 16 where the row declares notes_without_who16, and (fix(gateway): skip the startup requests the firmware refuses, poll each zone's temperature on SCS-only gateways #794's rule, against the pin) that every sweep_omits entry is *#4*0## / *#16*0*5## of a WHO in pinned_whos; the per-gateway supports_who / audio_measured / auth_measured checks against firmware_oracle.json are unchanged.
2. tests/test_gateway_profiles.py: the F461 and MyHomeServer1 cells (advertised and sent) are asserted against the installed profile (right before and after fix(profiles): drop WHO 16 from the F461 and MyHomeServer1 profiles OWNd#142), the other rows stay literal; discovery_whos_for mirrors the runtime gate for six profiles; a profile double without WHO 16 yields [1, 2, 4, 18], the reduced sweep_omits and the WHO 22 wording for both rows while the F454 / MH202 rows are unchanged; without OWNd every row carries exactly its pinned_whos and the generated block equals the committed one; with OWNd the derived block equals the committed one row by row (only rows whose profile differs from the pin may differ); pinned_whos is ignored when OWNd is importable; the validator rejects a stale *#16*0*5## at the pin bump, a pin that disagrees with the installed profile (other than by WHO 16 where declared), a malformed, repeated or reordered pin; the no-OWNd fallback yields the WHO 22 row once the pin drops WHO 16.
3. docs/getting-started/hardware-compatibility.md: unchanged in this PR (against the pinned OWNd the generated table is byte-identical). After the pin bump the F461 and MyHomeServer1 rows regenerate to WHO 1, 2, 4, 18 with the WHO 22 note.
4. .github/release-notes/2.0.0b16.md: one line under "Other changes".

Not changed here, named as follow-ups: the per-zone *#16*ZONE*5## media-player poll (media_player.py) on a gateway whose profile lacks WHO 16 (one refused frame per configured WHO 16 zone; those entities cannot work on these gateways anyway, and the planned WHO 22 backend is the right gate); and the topology side effect of OpenWebNet-HA/OWNd#142 itself (topology.py delegation: an F461 primary lets a WHO-16-capable follower own WHO 16 and 22, and a same-tier F454 / F461 pair is now ordered by WHO count rather than by MAC), which is OpenWebNet-HA/OWNd#142's to document and which this PR does not alter.

---

Companion PR: OpenWebNet-HA/OWNd#142 (fix(profiles): drop WHO 16 from the F461 and MyHomeServer1 profiles, draft). MyHOME #794 is the runtime half (startup requests gated on profile.supports_who, with a MyHOME fallback table until the OWNd release).

---

State Diagram

stateDiagram-v2
    direction LR
    state "Is OWNd importable" as Check
    state "Installed OWNd profile" as Profile {
        [*] --> Pinned : 2.0.0b11 - F461 and MyHomeServer1 supported_who includes 16
        [*] --> AfterPinBump : OWNd release with 142 - no 16
    }
    state "Discovery row - docs table, same rule as the startup gate" as Row {
        Pinned --> Row16 : discovery_whos_for gives 1 2 4 16 18 - notes sound unmeasured class default
        AfterPinBump --> Row18 : discovery_whos_for gives 1 2 4 18 - notes_without_who16 Sound is WHO 22 only
    }
    state "No OWNd - quality-scale, test-coverage and docs CI jobs" as Fallback {
        [*] --> PinnedRows : pinned_whos per row - the pinned OWNd rows, edited at the pin bump with update
        PinnedRows --> Row16 : today, identical to the committed table
    }
    [*] --> Check
    Check --> Profile : yes
    Check --> Fallback : no

Loading

---

Hardware Verification

• This PR sends nothing: it changes the docs generator and its tests. The bus facts behind fix(profiles): drop WHO 16 from the F461 and MyHomeServer1 profiles OWNd#142 are in test(trace): ingest live F461 2.0.11 2026-10-23 capture and MyHomeServer1 3.87.13 reads behind OWNd #136-#142 and MyHOME #791-#798 #799: the physical F461 2.0.11 NACKs *#16*0*5## in 5 ms and *#16*101*5## in 6 ms (run1_oracle_replay_readonly, step oracle_who16, all 18 WHO 16 request forms refused in under 10 ms); the physical MyHomeServer1 3.87.13 NACKs both in 3 ms (run2_reads_thermo_sound_alarm_lights, step read_who16); the upstream tests/fixtures/traces/issue_466 MyHomeServer1 fixture records the same NACK on *#16*0*5##.
• Live retest 2026-10-24, the fix(profiles): drop WHO 16 from the F461 and MyHomeServer1 profiles OWNd#142 / fix(gateway): skip the startup requests the firmware refuses, poll each zone's temperature on SCS-only gateways #794 gate (report-15 consensus label confirmed for this startup configuration): the combined build sent no *#16*0*5## at startup on the F461 (Skipping WHO=16 discovery: not supported by F461 profile.), the zone and humidity requests were present; other models not tested. This PR's own effect (the regenerated docs row) is not a bus event.

---

How is it tested

• tests/test_gateway_profiles.py::test_generate_gateway_discovery_table_and_verify_oracle (updated), ::test_discovery_whos_follow_the_runtime_gate, ::test_discovery_row_drops_who16_when_the_profile_does, ::test_discovery_table_without_ownd_uses_the_pinned_rows, ::test_discovery_table_without_ownd_matches_the_committed_docs, ::test_discovery_table_with_ownd_matches_the_committed_docs, ::test_pinned_rows_are_ignored_when_ownd_is_importable, ::test_oracle_check_rejects_a_stale_sweep_omission_at_the_pin_bump, ::test_discovery_table_without_ownd_follows_the_pinned_rows_at_the_pin_bump, ::test_oracle_check_rejects_a_pin_that_disagrees_with_the_installed_profile, ::test_discovery_row_with_an_unreadable_pin_and_no_ownd_is_empty (new); fix(gateway): skip the startup requests the firmware refuses, poll each zone's temperature on SCS-only gateways #794's ::test_oracle_check_rejects_a_sweep_omission_the_profile_does_not_advertise kept; 25 tests in the file.
• Fail-before / pass-after: on the base script 4 of the first-push tests fail (no discovery_whos_for, rows hard-code their WHOs); on the first-push script (CI red) 3 fail: the table test (rows carry no pinned_whos), the no-OWNd rows test and the no-OWNd-equals-committed-docs test (unknown (OWNd profile not available) cells). All 18 pass after.
• scripts/sync_documentation.py --check and scripts/verify_ha_standards.py: pass without OWNd (OWNd made unimportable, the CI jobs' shape: [OWND] Client library verification skipped; before the fix: Gateway Discovery table out of sync) and with OWNd master (87cafc3) on PYTHONPATH.
• Full suite against OWNd master (87cafc3, the behaviour of the 2.0.0b11 pin), rebased on v2-phase1-architecture a27b32f: 4965 passed, 1 skipped; scripts/verify_ownd_coverage.py 100 % (16,770 / 16,770 statements, 0 missing) (before the rebase, on 31ea9fd with OWNd f04d8aa: 4514 passed, 1 skipped; first push on ddc6d60: 4421 passed, 1 skipped).
• Against the fix(profiles): drop WHO 16 from the F461 and MyHomeServer1 profiles OWNd#142 branch (gdluck/OWNd fix/profiles-f461-mhs1-no-who16, 95dc8a5) on PYTHONPATH, full suite: 4964 passed, 1 failed, 1 skipped on the committed tree (the failure is tests/test_sync_documentation.py::test_check_all_documentation_real_tree, the docs drift the pin bump resolves) and 4965 passed, 1 skipped with the pin bump rehearsed (see "Rebase after fix(gateway): skip the startup requests the firmware refuses, poll each zone's temperature on SCS-only gateways #794"); verify_profiles_against_oracle() returns no violation, and the generator yields | **F461** | WHO 1, 2, 4, 18 | WHO 1, 2, 18 (firmware drops the shutters does not work in homekit #40##) | Lighting, cover, and energy meter 51 queries. Sound is WHO 22 only (no bt_difson; every WHO 16 frame refused, own-firmware-oracle 020011, confirmed live on 2.0.11). The firmware also drops the shutters does not work in homekit #40##, so the sweep leaves it out; zones are polled with the shutters does not work in homekit #4##. Alarm query omitted. | and the same shape for the MyHomeServer1.
• scripts/typing_ratchet.py 0 errors (baseline 0), ruff check ., scripts/verify_ha_standards.py, scripts/manage_translations.py check pass.
• Note for the sequencing: pypi_standards.yml installs OWNd from git (OWND_REF, the branch named like the MyHOME branch, else master), so once fix(profiles): drop WHO 16 from the F461 and MyHomeServer1 profiles OWNd#142 is on OWNd master that job reports the same docs drift on v2-phase1-architecture until the pin bump lands; this was already the case before this PR (the generator derived the rows from the installed profile when importable).

Live retest (2026-10-24)

Not exercised live: docs generator and tests only, no bus effect of its own (the gate it documents was confirmed for this startup configuration, above).

---

Where are the traces

Subsystem	Capture	Plant	Frames verified by this PR
Docs generator	No physical trace applies: this change touches no bus frame (configuration / diagnostics / validation only)	—	none
WHO 16 refused by the gateway (the fact OpenWebNet-HA/OWNd#142 encodes and this table will show)	tests/fixtures/traces/f461_live_2026-10-23/ run run1_oracle_replay_readonly step oracle_who16; tests/fixtures/traces/myhomeserver1_live_2026-10-24/ run run2_reads_thermo_sound_alarm_lights step read_who16 (trace PR: #799, merged); upstream tests/fixtures/traces/issue_466/	F461 2.0.11; MyHomeServer1 3.87.13; MyHomeServer1 (#466)	*#16*0*5## NACK 5 ms / 3 ms, *#16*101*5## NACK 6 ms / 3 ms (note: who16_refused_locally)

---

What does the Firmware Oracle / Encyclopedia say

Re-checked against the refreshed results/mcp_index.json at own-firmware-oracle main 72759e8 (schema 1.2.0, 529 inputs; after OpenWebNet-HA/own-firmware-oracle#61 and OpenWebNet-HA/own-firmware-oracle#62): the F461 020011 and the MyHomeServer1 028206 answer nack / silent to all 86 WHO 16 inputs (77 in sound-who16-source, 9 in trace-corpus), *#16*0*5## at sound-who16-source.tsv line 16 and *#16*101*5## at line 28 on both, no bus frame, no emitted frame; identical to the 53f9514 rows mirrored in tests/golden/firmware_oracle.json (schema 1.1.0, 509 inputs). The refresh added the 20 who13-clock inputs and four sound-who22 rows per gateway and changed nine MH202 lights-area0 / interface-routing / trace-corpus replies (nack → ack); nothing about WHO 16 changed. The golden mirror is left as is: it is refreshed by hand in its own commit together with tests/test_firmware_oracle_conformance.py (which pins schema_version to 1.0.0 / 1.1.0) and tests/test_trace_oracle_validation.py (which cites 53f9514), not something this PR should fold in.

Published rows (results/<P>/<V>/oracle/full/sound-who16-source.tsv, index 53f9514 and 72759e8 alike), *#16*0*5##: nack / silent on F450 020010, F459 020105, F460 020012, F461 020011, MH202 010024, MyHomeServer1 028206; ack / out (bus 24 30 34 42 35 30 30 38 30 38 46 0d) on F453 010100, F453AV 030014, F454 020051, H4684 020054, MH200 020102, MH200N 010108. OpenWebNet-HA/OWNd#142 drops WHO 16 from the F461 and MyHomeServer1 profiles on the strength of all 86 WHO 16 inputs being refused there; it keeps the MH202 (a live MH202 relays WHO 16 state frames from an F441M matrix, MyHOME #669), so the generated MH202 row does not change either. This PR adds no verdict of its own: verify_profiles_against_oracle() keeps checking audio_measured against this row and supports_who against the WHO 1 / 2 / 3 / 4 / 25 probe rows, unchanged.

---

Checklist

• PR Description: Follows the structured sections above (What is the problem, What is the solution, State Diagram, How is it tested, Where are the traces, What does the Firmware Oracle / Encyclopedia say).
• Trace-Backed (if protocol/device change): N/A, docs generator and tests; the WHO 16 refusals it will document are in test(trace): ingest live F461 2.0.11 2026-10-23 capture and MyHomeServer1 3.87.13 reads behind OWNd #136-#142 and MyHOME #791-#798 #799 and the oracle rows above.
• V2 Architecture: N/A, no entity changes.
• Tests & Linters: pytest, ruff, and scripts/verify_ha_standards.py pass.

🤖 Generated with Claude Code

[GreenGrassBlueOcean]

@gdluck i double check the firmware oracle and made a fix for the stale who16 please double check with the oracle and fix failing ci

[gdluck]

Checked with the refreshed oracle and fixed the CI (head 88ed3fa, rebased on the current base, still draft).

CI cause and fix. The quality-scale, test-coverage and docs jobs run the docs check without OWNd installed; this PR's first push derived the discovery rows from the installed profile and, with no OWNd, printed "unknown" rows, so the regenerated table no longer matched docs/getting-started/hardware-compatibility.md. Each discovery row now carries pinned_whos (the rows the pinned OWNd 2.0.0b11 yields: F461 / MyHomeServer1 still WHO 1, 2, 4, 16, 18), used only when OWNd is not importable; with OWNd the row is still derived from profile.supports_who. Tests pin installed-profile table == committed table == pinned_whos, and verify_profiles_against_oracle() now flags a profile name that resolves to the Generic profile (the previous is None check was vacuous). Both shapes pass sync_documentation.py --check locally (with and without OWNd).

Oracle re-check (main 72759e8, after #62). F461 020011 and MyHomeServer1 028206 still answer nack / silent to all 86 WHO 16 inputs (77 in sound-who16-source, 9 in trace-corpus; *#16*0*5## line 16, *#16*101*5## line 28), no bus frames: nothing changed for WHO 16, so OpenWebNet-HA/OWNd#142 stands. The refresh added the who13-clock inputs, 4 sound-who22 rows per gateway and flipped 9 MH202 replies in other suites. tests/golden/firmware_oracle.json (1.1.0 / 509) was not synced here: it is refreshed by hand in its own commit together with the conformance test (which pins the schema version), so that belongs in a separate PR if you want 72759e8 in.

Suite 4514 passed at 100 % coverage against the pinned OWNd; against the OWNd #142 branch the profile, conformance and trace-oracle tests pass (711) and only the docs-equality checks report the expected pin-bump drift, resolved by --update at that time. CI on 88ed3fa was still queued when I wrote this.

[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!

[GreenGrassBlueOcean]

@gdluck please rebase

[gdluck]

Rebased onto v2-phase1-architecture a27b32f (head bfc1c27, still draft) and reconciled with #794, which had reworked the same table generator:

• fix(gateway): skip the startup requests the firmware refuses, poll each zone's temperature on SCS-only gateways #794's split advertised / sent columns, sweep_omits, its validator and its sweep test are kept as they are.
• This PR's derivation of the advertised WHOs from the installed OWNd profile, the pinned_whos fallback for the no-OWNd docs job and the notes_without_who16 wording on the F461 / MyHomeServer1 rows are kept, now carrying fix(gateway): skip the startup requests the firmware refuses, poll each zone's temperature on SCS-only gateways #794's *#4*0## facts.
• Join rule: a general request of a WHO the installed profile no longer advertises is no longer listed as a sweep omission (nothing is sent for that WHO), so with fix(profiles): drop WHO 16 from the F461 and MyHomeServer1 profiles OWNd#142 the F461 row becomes WHO 1, 2, 4, 18 | WHO 1, 2, 18 (firmware drops *#4*0##) and still matches _discovery_frames().
• The validator now checks pinned_whos against the installed profile (tolerating only the declared WHO 16 switch, in sweep order, no repeats) and sweep_omits against the pin; an independent second-opinion review found the pin itself was unchecked and that malformed or repeated pins slipped through, both fixed on top (3aff66c, c1405d4).

Suite with OWNd master 87cafc3: 4965 passed / 1 skipped at 100 % coverage; with the OpenWebNet-HA/OWNd#142 checkout: 4964 passed and the single expected docs-drift failure on the committed tree (the two rows a pin bump regenerates), 4965 passed / 1 skipped with the pin bump rehearsed (pinned_whos, sweep_omits, sync_documentation.py --update: only those two docs rows change). ruff, typing ratchet, HA standards, docs sync with and without OWNd, and translations clean; hardware-compatibility.md is byte-identical on this base. Details under "Rebase after #794 (2026-10-26)" in the description.

[gdluck]

Rebased once more onto v2-phase1-architecture e649bc8 (after #818): only the H4684 / H/L4684 row name in the table test changed; 4989 passed / 1 skipped at 100 % coverage against OWNd master, and against the OpenWebNet-HA/OWNd#142 checkout the single expected docs-drift row remains. Head 1e346c3, mergeable, still draft.

[GreenGrassBlueOcean]

@gdluck Clean rebase and reconciliation with #794 and #818 — verified the discovery_whos_for derivation, the pinned_whos no-OWNd CI fallback, the sweep_omits / notes_without_who16 transition rules, and the refreshed own-firmware-oracle (72759e8) WHO 16 verdicts on F461 (020011) and MyHomeServer1 (028206). All 17 CI checks are green.

One final item before merge (inline below):

• Add a negative unit test in tests/test_gateway_profiles.py for the new GenericGatewayProfile guard in verify_profiles_against_oracle() (scripts/update_gateway_profiles.py:642-646), asserting that a discovery row whose profile_name is "Generic" or an unknown model string (resolving to GenericGatewayProfile) emits the "resolves to the generic profile" violation.

Once that test is added (and you mark the PR Ready for review), I will merge right away.

[GreenGrassBlueOcean]

This new isinstance(prof, GenericGatewayProfile) guard fixes the earlier vacuous is None check, but none of the negative tests in tests/test_gateway_profiles.py exercise it yet. Please add a quick negative test case (e.g. inside test_oracle_check_rejects_a_pin_that_disagrees_with_the_installed_profile or a dedicated test) setting a row's profile_name to "Generic" or a nonexistent model name and asserting "resolves to the generic profile" is reported in violations.

[gdluck]

Added in fd20792: test_oracle_check_rejects_a_row_that_resolves_to_the_generic_profile, parametrised over Generic and an unknown model string, asserts the resolves to the generic profile violation for that row and for no other; it fails with the previous is None check and passes with the GenericGatewayProfile guard. Full suite 4991 passed / 1 skipped at 100 % coverage, ruff and typing ratchet clean. Marked ready for review.