Recover a consumer's bus_id from its source block

[bdraco]

What does this implement/fix?

The last 3 boards allow-listed by #1755 — kincony_ag8 (uart), kincony_kc868_e8t
(uart), kincony_mb (i2c) — shipped a full-setup config that wouldn't compile
because their bus was never lifted. The allow-list comments claimed "source has no
top-level block", but the live devices.esphome.io pages (and the cached source)
show the bus is there, with explicit ids, in the same code block as the leaf —
so that was a misdiagnosis. The real cause is an importer bug.

Each board has multiple same-platform leaves with no id: on the consumer
block (2 switch.uart, 2 sensor.bl0939, 16 sensor.ina226), each binding a bus
via uart_id:/i2c_id:. That ref is dropped from the finalized entry (the bus
isn't a featured component yet), so _collect_bus_dep_refs re-finds the source
block via _find_consumer_block — which filters by platform, gets N id-less
blocks, can't disambiguate, and returns None. With the ref lost and multiple
buses present, _materialize_bus can't pick one, so the bus never lifts.

Fix: _Candidate already retains the source block and each finalized entry's id
equals its unique local_id, so stash the source block on the entry at finalize,
read it back in _collect_bus_dep_refs (falling back to _find_consumer_block
for entries built outside the extractor, e.g. unit tests), and strip the transient
key in _make_record before serialization. Each consumer's <bus>_id is now
recovered directly: kincony_mb's 16 ina226 split correctly across bus_a/bus_b,
ag8/e8t wire each leaf to its own uart.

This empties the guard's _UNSATISFIED_BUS_ALLOW_LIST — every imported board's
featured bus dependencies now lift, and validate_definitions.py proves it (0
errors). Regenerated against ESPHome 2026.6.3; the diff is exactly the 3 boards
gaining their uart/i2c entries + requires.

Related issue or feature (if applicable):

• Follow-up to Validate that a board's featured bus dependencies are satisfied #1755; finishes its 5 allow-listed boards (Infer one_wire platform gpio for a platform-less source block #1756 did the other 2).

Types of changes

• Bugfix (non-breaking change which fixes an issue) — bugfix

Frontend coordination

• No frontend change needed

Checklist

• The code change is tested and works locally.
• Pre-commit hooks pass (ruff, codespell, yaml/json/python checks).
• Tests have been added or updated under tests/ where applicable.
• components.index.json / definitions/components/*.json have not been hand-edited.
• Architecture-level changes are reflected in docs/ARCHITECTURE.md and/or docs/API.md.

[codspeed-hq]

Merging this PR will not alter performance

✅ 27 untouched benchmarks

---

_{Comparing bus-recover-consumer-id (b5fcf67) with main (9acd745)}

[codecov]

Codecov Report

✅ All modified and coverable lines are covered by tests.
✅ Project coverage is 99.54%. Comparing base (9acd745) to head (b5fcf67).

Additional details and impacted files

@@           Coverage Diff           @@
##             main    #1757   +/-   ##
=======================================
  Coverage   99.54%   99.54%           
=======================================
  Files         228      228           
  Lines       18203    18203           
=======================================
  Hits        18121    18121           
  Misses         82       82           

Flag	Coverage Δ
py3.12	99.52% <ø> (ø)
py3.14	99.54% <ø> (ø)

Flags with carried forward coverage won't be shown. Click here to find out more.

🚀 New features to boost your workflow:

• ❄️ Test Analytics: Detect flaky tests, report on failures, and find test suite problems.
• 📦 JS Bundle Analysis: Save yourself from yourself by tracking and limiting bundle sizes in JS merges.

[Copilot]

Pull request overview

Fixes a bug in the devices.esphome.io importer where bus dependencies (i2c/uart) could not be lifted for id-less, same-platform consumers by recovering the consumer’s <bus>_id directly from its original source block.

Changes:

• Stash each finalized featured entry’s source YAML block transiently and use it to recover <bus>_id during bus dependency collection (then strip before serialization).
• Update validation/test coverage to treat _UNSATISFIED_BUS_ALLOW_LIST as a per-(board,bus) allow-list and confirm the list can be empty.
• Regenerate KinCony board manifests/bodies so required i2c/uart components are featured and wired via requires.

Reviewed changes

Copilot reviewed 10 out of 11 changed files in this pull request and generated no comments.

Show a summary per file

File	Description
script/sync_esphome_devices.py	Stashes _source_block on finalized entries, uses it to recover <bus>_id in _collect_bus_dep_refs, and strips it before writing records.
script/validate_definitions.py	Converts the unsatisfied-bus allow-list to a typed, currently-empty frozenset and keeps per-(board,bus) semantics.
tests/test_sync_esphome_devices_bus_dep.py	Adds a regression test ensuring bus id recovery works when _find_consumer_block can’t disambiguate id-less same-platform consumers.
tests/test_validate_featured.py	Updates tests to monkeypatch the allow-list as (board, bus) pairs (and not rely on real board ids).
esphome_device_builder/definitions/boards/kincony_mb/manifest.yaml	Adds featured i2c buses (bus_a/bus_b) and stamps requires on INA226 sensors; updates occupied pins.
esphome_device_builder/definitions/boards/kincony_kc868_e8t/manifest.yaml	Adds featured UARTs (uart_1/uart_2) and stamps requires on BL0939 sensors; updates occupied pins.
esphome_device_builder/definitions/boards/kincony_ag8/manifest.yaml	Adds featured UART (uart_232) and stamps requires on the UART switch; updates occupied pins.
esphome_device_builder/definitions/board_bodies/kincony_mb.json	Regenerated body including i2c buses and per-sensor requires.
esphome_device_builder/definitions/board_bodies/kincony_kc868_e8t.json	Regenerated body including UARTs and per-sensor requires.
esphome_device_builder/definitions/board_bodies/kincony_ag8.json	Regenerated body including UART and requires.

[esphbot]

Previous review — superseded by a newer review below.

[esphbot]

No blocking issues found.

[bdraco]

Thanks @esphbot — applied your suggestion 1 (read-and-discard) in b5fcf678.

_collect_bus_dep_refs is the sole reader of _source_block, so it now pops the key at the top of its loop (before any continue) and uses the popped value, and the separate strip loop in _make_record is gone. The transient key is removed at the moment it's consumed, for every entry, independent of how the entries are later serialized — closing the 'future serialization path leaks it' risk. Updated the _finalize_entry comment to point at the new consumption site.

Behavior-preserving: the regenerated catalog is byte-identical, no _source_block in any board body, validation stays at 0 errors, full suite green.

[esphbot]

@bdraco: Thanks @esphbot — applied your suggestion 1 (read-and-discard) in b5fcf678.

Verified on PR head. Pop sits above every continue, so all leaf entries strip — infra entries included. _finalize_entry is the sole _source_block producer; both call sites feed _make_record's featured, which always runs _extract_bus_deps. _materialize_bus/hub entries are plain dicts, prepended after the pop, so nothing reintroduces the key. Single-serializer coupling closed. Catalog clean, allow-list empty. LGTM.

[esphbot]

PR Review — Recover a consumer's bus_id from its source block

Correct, well-scoped importer fix; read-and-discard hardening applied; merge-ready.

Strengths:

• Accurate root-cause: _find_consumer_block can't disambiguate N id-less same-platform consumers, so each leaf now recovers its own <bus>_id from the stashed source block, with the finder kept as a fallback for entries built outside the extractor.
• Prior suggestion Add section config API and user preferences #1 resolved in b5fcf678: _collect_bus_dep_refs now pops _source_block at the top of its loop (before any continue), and the separate strip in _make_record is gone — the transient key is removed at the moment it's consumed, independent of serialization path.
• Regression test asserts observable behavior (lifted bus ids + per-consumer requires), not source text.

Verification performed:

• _finalize_entry is the only producer of _source_block; both its call sites flow into _make_record's featured, which unconditionally passes through the _extract_bus_deps lift (sole success return is after the loop).

• The pop runs before the if "." not in component_id: continue, so infra entries are stripped too.

• _materialize_bus / hub entries are built as plain dicts (no _source_block) and prepended after the pop, so nothing re-introduces the key.

• _UNSATISFIED_BUS_ALLOW_LIST correctly empties to a typed frozenset(); regenerated KinCony bodies contain no _source_block.

• No blocking issues.

---

---

Checklist

• Transient key stripped before serialization on all paths (read-and-discard)
• Bus-dep read recovers id-less consumer's bus correctly
• No mutation/aliasing of shared source block
• Generated catalog matches stated scope, no transient key
• Test coverage asserts behavior, not source

---

Automated review by Kōan (Claude) HEAD=b5fcf67 3 min 28s

[esphbot]

No blocking issues found.