[gdluck]

Requested by A

Summary

One narrow parser hardening for WHO 13, documentation fixes, and six golden records from committed MyHOME captures. Every ASCII frame decodes exactly as before.

Fixed (runtime)

• OWNGatewayEvent: value fields are ASCII digits only. str.isdigit and int accept fullwidth digits, so *#13**12*0*３*80*1*214*135## decoded a MAC address, *#13**19*１*2*3*4## an uptime and *#13**15*２３## an Unknown (２３) device type; such frames now stay untyped events with the raw frame as their log, as the Dimension 27 allowlist already did (_is_ascii_digits). A fullwidth digit in a clock field (*#13**1*03*26*10*２０２６##, *#13**0*12*00*00*0０１##, the Dimension 22 form) raised a ValueError out of date() / fromisoformat; it now raises the same ValueError("Invalid gateway date" / "Invalid gateway timezone") as any other malformed clock value, which the sessions catch (fix(gateway): WHO 13 clock echoes with timezone 0530/124 no longer raise #144; _gateway_timezone matches with re.ASCII). The review's empty-token cases (*#13**1**01*01*2026##, *#13**22*...*001**01*...##, *#13**12*0**80*...##, *#13**15*##, *#13**19**0*0*0##) never raised on master (no weekday is decoded; Dimensions 12 / 19 are guarded) and are pinned as such in tests/test_who13_ascii_values.py.

Documentation only

• docs/protocol_conformance_matrix.md: the WHO 13 examples are the published empty-WHERE read form *#13**0## / *#13**22## (Encyclopedia who-13-integration-gateway/dimensions.md "Read form"), the form OWNGatewayCommand.request_time() and the keepalive emit; own-firmware-oracle ownd-audit-frames, who13-clock and who13-clock-readback have all thirteen emulated gateways answering them with ACK and the time, while the old *#13*0*0## is nack, silent on all thirteen (ownd-audit-frames).
• OWNGatewayCommand.request_ip_address / request_netmask / request_mac_address: docstring caveat from who13-identification: answered by 8 of 13, ACKed silently by F453, F453AV and MH200N, NACKed with a worker crash by H4684 020054 and MH200 020102. Not refused at the send layer.
• OWNHeatingCommand.actuators_status: #Z#C is zone Z through a four-zone central unit C (a controlled-zone address, not an actuator; feat(heating): OWNHeatingCommand.actuators_status(where) builder for *#4*WHERE*20## #140 leftover).
• OWNGatewayEvent.device_type / WHO13_DEVICE_TYPES: an unlisted Dimension 15 code stays Unknown (N); the non-unique 200 is a list of sharers, told apart by WHO 1013 Dimension 1.
• tests/golden/frames/who13_gateway.yaml (+6, community-plant-capture): F461 2.0.11 *#13**0*17*45*24*999##, *#13**16*2*0*11##, *#13**23*5*10*35##, *#13**24*2*26*4## (MyHOME#865, f461_live_2026-11-04 sweep), *#13**22*22*59*16*999*01*26*10*2026## (f461_who13_clock_tz_live_2026-10-26, line 309) and H/L4684 6.0.11 *#13**15*23## (issue_466/config_entry-myhome_HL4684_fw6_0_11.json); corpus regenerated (172 frames), golden README count 172.
• tests/test_message_protocol.py: a regression test pinning the "Gateway broadcasting ..." log text MyHOME relies on.

Review follow-up (round 2, 2026-11-10)

#	Item	Action
1	Title / body; 999 attribution; golden frames	Rewritten. The 999 datetime is the physical F461 (f461_who13_clock_tz_live_2026-10-26 line 309), not MH200N (001) or MH202 (002). The six frames are now golden records.
2	master / CHANGELOG (D14)	origin/master (1cb2dc1) merged; the documentation bullet is gone; the ### Fixed entry covers the ASCII hardening only.
3	Matrix: private path, 12 vs 13	Replaced by the suite names and "all thirteen".
4	Dimension 10 / 11 / 12 builders	Docstring caveats added; nothing refused.
5	Fail-closed parser bugs	Cases 1-4 do not reproduce on master (code and probe: no weekday decoding, Dimensions 12 / 19 guarded, *#13**15*## reads Unknown ()); the non-ASCII half of each case does, and is fixed with _is_ascii_digits / re.ASCII. Clock fields keep raising ValueError, by #144.

Review follow-up (round 3, 2026-11-12)

#	Item	Action
1	"please rebase" (the branch conflicted with master after #157, #159 and #160 landed)	origin/master c546a2e merged (fa54e90, one conflict: the golden README fixture count, #159's 166 against this PR's 172); corpus.json regenerated (172, byte-identical to the merge result). #159's test test_golden_corpus_dimension_10_write_is_conflict_and_no_tilt pins that count literally, so the pin moves from 166 to 172 (a8573f2). No other file differs from the round-2 head.

Testing

• ruff check OWNd tests setup.py scripts, mypy OWNd tests/typing, python scripts/verify_library_standards.py, python scripts/sync_documentation.py --check, python tools/golden/validate_corpus.py (172 frames).
• python -m pytest -q -p pytest_cov --cov=OWNd --cov-report=xml on a8573f2: 3348 passed; python scripts/verify_coverage.py coverage.xml 80.0: 28 of 28 OWNd modules at 100 % (tests/test_who13_ascii_values.py: 14).
• MyHOME v2-phase1-architecture (f19adc57) suite with a8573f2 on PYTHONPATH: 5601 passed, 1 deselected in 452.43s (the deselected one is the known test_check_all_documentation_real_tree pin signal).
• Golden README count: fix(automation): WHO 2 builders validate WHERE with require_address_where, ASCII-only priority string; WHO 2 / WHO 9 documentation and golden-corpus errata (Encyclopedia audit, group 2) #159 landed first with 166; this PR moves the README, corpus.json and fix(automation): WHO 2 builders validate WHERE with require_address_where, ASCII-only priority string; WHO 2 / WHO 9 documentation and golden-corpus errata (Encyclopedia audit, group 2) #159's pin in tests/test_who2_audit_fixes.py to 172.

🤖 Generated with Claude Code

[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!

[gdluck]

One documentation commit here from the 2026-11-04 cross-check: gateway.py now records, next to the Dimension 10 / 11 / 12 decoders, that requesting the IP address, netmask and MAC (*#13**10##, *#13**11##, *#13**12##) makes the gateway daemon exit on H4684 020054 and MH200 020102 in the firmware oracle (nack/crash; firmware_observed under QEMU emulation, not confirmed on physical hardware; findings/who13-identification.md), while the other eleven emulated gateways ACK them, eight with a telegram on the bus and three silently (F453, F453AV, MH200N). OWNd builds no request for them and deliberately applies no send-layer guard: the frames are legitimate on most gateways, so a refusal would need the gateway model at send time. CHANGELOG line added.

[gdluck]

Live on the F461 2.0.11 test plant (2026-11-04), supporting the note that only H4684 / MH200 crash on WHO 13 dims 10/11/12: the F461 answers all of them without crashing (network values redacted) —

*#13**10## -> *#13**10*<IP octets>##       (the gateway's own IPv4)
*#13**11## -> *#13**11*<netmask octets>##
*#13**12## -> *#13**12*<MAC octets>##
*#13**15## -> *#13**15*200##
*#13**16## -> *#13**16*2*0*11##            (firmware 2.0.11)
*#13**27## -> *#13**27*<ASCII codes>##      ("F461")

So dims 10/11/12 are safe on this gateway, consistent with the oracle rows; the crash is H4684/MH200-specific. (These dims echo the gateway's live network interface, which is also why a committed oracle TSV had to be aligned with CI's jail interface rather than a capture host's.)

[GreenGrassBlueOcean]

Thanks for catching the *#13*0*0## conformance-matrix error (confirmed NACK (*#*0##), SILENT across all 12/12 emulated gateways in lookup_firmware_verdict) and the #Z#C docstring leftover from #140!

Under our standing rule — if we don't have real, committed on-wire plant traces from physical hardware, we don't change behaviour — the documentation and conformance-matrix fixes here are spot-on and ready to merge, whereas the OWNd/message/gateway.py runtime changes are doc-derived without real plant captures and should be reverted so we can merge the doc fixes cleanly.

What can stay (ready to merge once trimmed)

1. docs/protocol_conformance_matrix.md (G5): Updating the WHO 13 read examples from *#13*0*0## / *#13*0*22## (refused 12/12 in the firmware oracle) to *#13**0## / *#13**22## (answered 12/12).
2. OWNd/message/heating.py (#140 leftover): Fixing the OWNHeatingCommand.actuators_status docstring so #Z#C is accurately described as zone Z via 4-zone central unit C.
3. OWNd/message/gateway.py docstrings / comments (G9): Clarifying the device_type docstring and WHO13_DEVICE_TYPES comment.

---

Blockers — Behaviour changes without committed plant traces (please revert)

#	Symbol / File	Behaviour change without real plant trace	Required action
1	_gateway_date (DIMENSION 1 & 22) in OWNd/message/gateway.py	Changes 2-digit years (*#13**1*04*15*10*26##, *#13**22*...*26##, and write frames) from returning a datetime.date to returning None, driven solely by dimensions.md field-width prose with zero real gateway traces showing malformed 2-digit year emissions requiring a decoder change.	Revert _gateway_date runtime behaviour to master.
2	_gateway_timezone ("100") in OWNd/message/gateway.py	Changes timezone code "100" from "-00:00" to "+00:00" purely from dimensions.md (ownkb:claim:c001701 / c001702) with zero on-wire captures showing a gateway emitting "100" — indeed, your own live F461 2.0.11 check showed real hardware emits "999" (unconfigured), not "100". (Also note: please avoid citing uncommitted local reports/data/live_f461_allprs_2026-11-04/ paths; any live capture worth referencing should be committed as a fixture.)	Revert _gateway_timezone "100" decoding to master.
3	DIMENSION 12 8-octet ZigBee log, addressed DIMENSION 16 log, & valid_mac_address / weekday in OWNd/message/gateway.py	Rewrites human_readable_log on 8-value DIMENSION 12 (Gateway's IEEE address is: ...) and addressed DIMENSION 16 (Product X firmware version is: ...), and adds valid_mac_address / weekday properties with zero ZigBee (ownkb:caution:k000009) or addressed-product WHO 13 plant traces.	Revert the gateway.py runtime decoding, log, and property changes (keeping only the G9 docstring/comment clarifications) so this PR merges cleanly as a docs + conformance-matrix fix.

[gdluck]

Applied in 4b5b8a2 (pushed; the description now carries the same follow-up table at the top).

Blocker	Action taken
1. _gateway_date	Reverted: 2-digit years on DIMENSION 1 / 22 (event and command) return datetime.date / datetime as on master.
2. _gateway_timezone	Reverted: code 100 reads -00:00 as on master.
3. Logs and properties	The 8-octet ZigBee IEEE log, the addressed DIMENSION 16 "Product X" log, valid_mac_address, weekday and their helpers removed; DIMENSION 12 is master's unchecked formatting. gateway.py is master plus the G9 device_type docstring and WHO13_DEVICE_TYPES comment; the 8 tests of reverted behaviour pruned from test_message_protocol.py, the one pinning master's unchanged clock log kept.

Kept, as listed under "What can stay": the conformance-matrix WHO 13 read examples *#13**0## / *#13**22## (G5), the OWNHeatingCommand.actuators_status #Z#C docstring fix and the G9 clarifications. The reports/data/... citation that was in this description is gone (it referred to an uncommitted live sweep). CHANGELOG: one "documentation only, no behaviour change" entry. Nothing at runtime differs from master.

Committed plant evidence. The WHO 13 replies physical gateways actually send are already in MyHOME's fixtures and all carry four-digit years, so the reverted 2-digit branch is exercised by no plant: F461 2.0.11 *#13**1*04*15*10*2026## (f461_live_2026-10-15/, -18/, -23/), MH200N / MH202 / MH200 *#13**22*22*59*16*999*01*26*10*2026## with the 999 timezone field (issue_466/myhome_trace_MH200N_all_2026-09-25T19-50-04.json, issue_466/myhome_trace_MH202_all_2026-09-26T16-59-17.json), F461 / MyHomeServer1 *#13**12*0*3*80*0*4*97## (f461_live_2026-10-23/, myhomeserver1_live_2026-10-24/), F454 / MH201 *#13**16*2*0*11## (issue_466/myhome_sweep_F454_all_2026-09-26T16-59-13.json, issue_466/myhome_sweep_MH201_all_2026-10-27T07-28-09.json). If you want them as OWNd golden records (parser shape only, no decoding change), say so and they come as a separate PR.

---

Update (2026-11-04, 16:07Z): New committed plant trace for WHO 13: MyHOME#865 (tests/fixtures/traces/f461_live_2026-11-04/, full custom_components.myhome DEBUG log and the bus-monitor sweep export of the physical F461 2.0.11, captured on the released OWNd 2.0.0b13, SHA-256 pinned by tests/test_f461_live_2026_11_04_trace.py). The F461 answers dimensions 0, 1, 10, 11, 12, 15, 16, 17, 19, 22, 23, 24, 27 and WHO 1013: *#13**1*03*04*11*2026## and *#13**22*17*42*27*999*03*04*11*2026## (four-digit year with weekday, time zone 999), *#13**0*17*45*24*999##, *#13**17*4*2*0##, *#13**19*0*14*53*9##, *#13**27*70#52#54#49##, *#1013**1*134*15*5*0##. It confirms that no physical F461 reply exercises the 2-digit-year branch reverted here, and that the 999 time zone field is what the plant sends. If golden records for these replies are wanted (parser shape only), they come as a separate PR.

[GreenGrassBlueOcean]

Thanks for reverting the speculative WHO 13 runtime changes in 4b5b8a2. Looking at what remains in 95e99ee, the current diff is purely documentation + one log-regression test, which fixes two real doc errata (*#13**0## / *#13**22## in docs/protocol_conformance_matrix.md and the #Z#C docstring leftover from #140 on OWNHeatingCommand.actuators_status).

Before this can land, please address the following:

1. Rewrite the PR title and PR description so they reflect what the PR is now about

We don't care about the past iterations or what was reverted along the way — the PR title and body (## Summary, ## Changes, ## Testing, ## Precedent check) still describe six reverted runtime changes (valid_mac_address, weekday, 4-digit year enforcement in _gateway_date, 100 -> +00:00 in _gateway_timezone, 8-octet ZigBee Dimension 12 log, and addressed Dimension 16 log). That makes the PR confusing to read and will corrupt the squash-merge commit message on master.

• Please rewrite both the PR title and the entire PR description from scratch so they state only what this PR actually changes today (no changelog of reverted commits/features).
• In the "Committed plant evidence" section, fix the misattributed trace citation: *#13**22*22*59*16*999*01*26*10*2026## (999 timezone) comes from the physical F461 capture f461_who13_clock_tz_live_2026-10-26/myhome_trace_F461_who13_clock_tz_live_2026-10-26.json:309, not MH200N (001, already in who13_gateway.yaml:116) or MH202 (002, *#13**22*20*54*00*002*06*26*09*2026## in issue_466/myhome_trace_MH202_all_2026-09-26T16-59-17.json:73).
• If you cite the new F461 (f461_live_2026-11-04/myhome_sweep_F461_all_2026-11-04T15-45-56.json) and HL4684 (issue_466/config_entry-myhome_HL4684_fw6_0_11.json) captures (*#13**0*17*45*24*999##, *#13**16*2*0*11##, *#13**23*5*10*35##, *#13**24*2*26*4##, *#13**15*23##), please actually add those frames to tests/golden/frames/who13_gateway.yaml.

2. Rebase onto origin/master and fix CHANGELOG.md (D14 / OWNd#143)

• The branch is currently 35 commits behind origin/master.
• Per D14 (OWNd#143 — and as your own commit 3e682a6 noted), CHANGELOG.md only records user-visible runtime changes. Drop the - **Documentation only, no behaviour change**: ... (2026-11-04 audit G5, G9) bullet (or, if you include the fail-closed parser fixes in item 5 below, record only those runtime fixes and strip internal worksheet tags like (2026-11-04 audit G5, G9)).

3. docs/protocol_conformance_matrix.md: remove private local path & fix gateway count (12 -> 13)

On line 39:

(review-session oracle run of 2026-11-03, audit_ency/ORACLE_CONFIRM.md row G5; *#13**0## is answered with ACK and the current time on all twelve)

• audit_ency/ORACLE_CONFIRM.md row G5 is a private local scratch file — cite the public own-firmware-oracle suites (ownd-audit-frames, who13-clock, who13-clock-readback) instead.
• own-firmware-oracle runs those suites across thirteen emulated gateways (3488 030205 included), not twelve: all 13 return NACK (*#*0##), SILENT on *#13*0*0##, and all 13 return ACK (*#*1##) + clock reply on *#13**0## and *#13**22##.

4. Dimensions 10 / 11 / 12 request builders (request_ip_address, request_netmask, request_mac_address)

In #issuecomment-6619565077 you wrote that "OWNd builds no request for them", overlooking OWNGatewayCommand.request_ip_address() (*#13**10##), request_netmask() (*#13**11##), and request_mac_address() (*#13**12##) in OWNd/message/gateway.py:621-633.

• Do not refuse them at the send layer (8/13 emulated gateways and physical F461 / MyHomeServer1 answer them, and F453, F453AV, MH200N ACK + SILENT), but please add a one-line docstring caveat on request_ip_address(), request_netmask(), and request_mac_address() noting that H4684 020054 and MH200 020102 NACK and crash their worker process in who13-identification.

5. Real WHO 13 fail-closed parser bugs on master that this audit missed

Because OWNMessage._DIMENSION_REQUEST_REPLY and _DIMENSION_WRITING (OWNd/message/base.py:102-110) match (?P<dimension_value>(?:\*\d*)+) (which permits empty tokens "" between stars as well as non-ASCII Unicode \d) and OWNEvent.parse / OWNCommand.parse do not catch ValueError, _parse_dimension in OWNd/message/gateway.py currently crashes OWNMessage.parse() with an uncaught ValueError on malformed wire frames:

1. Dimensions 1 / #1 (gateway.py:208-210) & 22 / #22 (gateway.py:255-259): _gateway_date() checks _is_ascii_digits and returns None when the weekday token is empty (*#13**1**01*01*2026## or *#13**22*12*00*00*001**01*01*2026##), and then the very next line unconditionally runs weekday = int(self._dimension_values[0]) (or [4]), raising ValueError: invalid literal for int() with base 10: ''.
2. Dimension 12 (gateway.py:224-228): f"{int(val):02x}" for val in self._dimension_values raises ValueError if any token is empty (*#13**12*0**80*1*214*135##) and accepts non-ASCII digits.
3. Dimension 15 (gateway.py:231-233): if len(self._dimension_values) >= 1: dtype = int(self._dimension_values[0]) raises ValueError on *#13**15*## (where self._dimension_values == [""]) and accepts non-ASCII digits.
4. Dimension 19 (gateway.py:240-245): int(self._dimension_values[0..3]) raises ValueError if any of the first four tokens is empty (*#13**19**0*0*0##) and accepts non-ASCII digits.
5. _gateway_timezone (gateway.py:89): re.fullmatch(r"[01]\d{2}", value) omits re.ASCII, so fullwidth Unicode digits like "0\uff10\uff11" match and parse as "+01:00" instead of returning None (unlike _gateway_time and _gateway_date, which enforce _is_ascii_digits).

Hardening those five spots with _is_ascii_digits (leaving existing attributes and log messages on valid frames untouched) would actually fix a real fail-closed parser issue in WHO 13 on par with #155–#160.

[GreenGrassBlueOcean]

Per D14 (OWNd#143), CHANGELOG.md only records user-visible runtime changes. Drop this Documentation only, no behaviour change entry (and avoid internal audit worksheet labels like (2026-11-04 audit G5, G9)). If you fix the fail-closed empty-token / ASCII-digit crashes in OWNd/message/gateway.py, record only that runtime fix under ### Fixed.

[GreenGrassBlueOcean]

1. Replace the private local file reference (review-session oracle run of 2026-11-03, audit_ency/ORACLE_CONFIRM.md row G5; ...) with the public own-firmware-oracle suites (ownd-audit-frames, who13-clock).
2. own-firmware-oracle replays these across thirteen emulated gateways (3488 030205 included), not twelve — all 13 return NACK, SILENT on *#13*0*0## and all 13 return ACK + the current time on *#13**0## and *#13**22##.

[gdluck]

Round 2 applied in e501d6a, with three small follow-up commits for the 100 % coverage gate (6eca707, fb5568d, 4b941a9): the Dimension 22 ASCII check is covered with eight-value frames, the Dimension 19 timedelta guard is kept for absurd magnitudes and covered, and the Dimension 12 except the ASCII gate made unreachable is gone. Local coverage of OWNd/message/gateway.py is 100 %.

#	Item	Action taken
1	Title / body / 999 attribution / golden frames	Rewritten. The 999 datetime is the F461 capture (f461_who13_clock_tz_live_2026-10-26 line 309); MH200N and MH202 carry 001 / 002. The five F461 / H/L4684 frames plus that Dimension 22 reply are golden records now (who13_gateway.yaml +6, community-plant-capture, corpus 172).
2	master / CHANGELOG	Merged; the documentation bullet is gone (D14). The ### Fixed entry records only the ASCII hardening below, no worksheet tags.
3	Matrix	The private path is gone; the sentence cites ownd-audit-frames, who13-clock, who13-clock-readback and says all thirteen (ACK + time on *#13**0## / *#13**22##, nack, silent on *#13*0*0##).
4	request_ip_address / request_netmask / request_mac_address	Docstring caveat: 8 of 13 answer, F453 / F453AV / MH200N ACK silently, H4684 020054 and MH200 020102 NACK and crash their worker (who13-identification). Not refused at the send layer.
5	Parser	Checked each case on master with the code and a parse probe. The empty-token halves (1-4) do not raise: no weekday is decoded on Dimensions 1 / 22, Dimension 12 is in a try, Dimension 19 is guarded by all(...), *#13**15*## reads Unknown () (there is no _gateway_date / _is_ascii_digits / weekday = int(...) on master; those lines were in this PR's reverted commits). The non-ASCII halves are real: fullwidth digits decoded a MAC / uptime / device type and raised out of date() / fromisoformat for clock fields. Fixed: Dimensions 1, 12, 15, 19 read ASCII digits only (12 / 15 / 19 stay untyped with the raw log; a clock field raises the same ValueError("Invalid gateway ...") as any malformed clock value, which the sessions catch, #144), _gateway_timezone uses re.ASCII. tests/test_who13_ascii_values.py (14 cases) pins both halves; every ASCII frame decodes as before.

Checks: ruff, mypy OWNd tests/typing, verify_library_standards.py, sync_documentation.py --check, validate_corpus.py (172); pytest 3119 passed; MyHOME v2-phase1-architecture f19adc57 suite with this branch on PYTHONPATH: 5601 passed, 1 deselected (the known test_check_all_documentation_real_tree pin signal). Note: #159 pins the golden README count at 166 and this PR at 172; whichever lands second needs the number bumped.

[GreenGrassBlueOcean]

@gdluck please rebase

[gdluck]

origin/master (c546a2e, after #157, #159 and #160) is merged into the branch in fa54e90 (merge commit, no history rewrite), with one follow-up commit a8573f2.

#	Item	Action taken
1	Rebase	master merged in; the only conflict was the golden README fixture count (#159's 166 against this PR's 172). tests/golden/README.md and corpus.json say 172 (tools/golden/validate_corpus.py regenerates the same file), and #159's test_golden_corpus_dimension_10_write_is_conflict_and_no_tilt, which pins that count literally, now pins 172. Nothing else differs from the round-2 head 4b941a9: the diff against master is the same nine files plus that one-line test change.

Checks on a8573f2: ruff, mypy OWNd tests/typing, verify_library_standards.py, sync_documentation.py --check, validate_corpus.py (172 frames); pytest -p pytest_cov --cov=OWNd 3348 passed with verify_coverage.py at 100 % for all 28 OWNd modules; MyHOME v2-phase1-architecture f19adc57 suite with this branch on PYTHONPATH: 5601 passed, 1 deselected in 452.43s (the deselected one is the known test_check_all_documentation_real_tree pin signal). CI on a8573f2: all 8 checks green.

---

Generated by Claude Code

[GreenGrassBlueOcean]

LGTM.The ASCII-only hardening for WHO 13 value fields is correct and minimal: fullwidth digits no longer decode as MAC/uptime/device-type, clock fields raise the same ValueError already used for other malformed values, and every pure-ASCII frame still decodes exactly as before. Documentation updates (conformance-matrix read form, Dimension 10/11/12 oracle caveats, #Z#C wording) are carefully qualified. The six new golden records come from already-committed plant captures. Branch is rebased onto current master, CI is green, and the MyHOME suite passes unchanged.Ready to merge.