Manual review requested by the owner after the automated reviewer failed.
Material af09d36d: green, no High/Medium findings, one non-blocking Low.
Issue: #765
User-Visible: no
The body of _process.yml is read from dev (@dev, #623). After "Перейти на
ветку задачи" the working copy of job prepare is the task branch, and a
show/ship branch with a clean merge is not rebased before review: its
scripts/ may lag dev or be replaced. #749 fixed job integrate; prepare
still ran four control scripts from the material — the issue-body digest
for the anchor, --reuse of a green verdict (#499), validate-gate (#510)
and the spec-change check (#517). The material decided its own admission:
a branch whose review-doc-guard.mjs prints reuse=true merges without the
model. model_review took model-usage.mjs from the material too: a lagging
branch has none, and the publication silently wrote reason=missing.
Now prepare extracts one snapshot right after setup-node, before the
branch switch: `git rev-parse origin/dev` once and `git archive <sha>
scripts .github/workflows/validate.yml`, so the git fetch of the track and
rebase steps cannot mix versions. Every repo script of the job runs from
it via TOOLS — the track step and the rebase guard lose their own
extractions. The SHA goes out as job output tools_sha; the usage step of
model_review archives the same commit inside itself, so a snapshot failure
is a failure of the reporting step (continue-on-error), not of the stage.
The model session runs on that runner, so the usage line stays untrusted
input parsed strictly (#556, #737).
withMaterialAnchors is idempotent: a repeated call drops the separator the
previous call wrote instead of piling up `---` lines.
Tests: test/process-prepare-tools.test.mjs — the job contract (no step
calls scripts/ from the working copy, one pinned archive before the branch
switch, tools_sha reaches model_review) and the steps as they are, on real
bash and git: a branch behind dev without model-usage.mjs and with a
substituted review-doc-guard.mjs; the anchor digest, reuse and the spec
check come from dev, the usage line is data even after dev moved. Red on
the old workflow: all four. Harnesses of process-track, rebase-generated,
review-doc-guard and model-usage take the prepare snapshot. Three registry
mutants (reuse from the material, usage from moving dev, separators).
Canon: PROCESS.md §10.4 «Скрипты конвейера — из dev» covers prepare and
the usage step; «Расход модели» names it a pipeline step, not the reviewer's.
Issue: #765
User-Visible: no
After the rebase onto dev (#762) the raw dist total is 2 607 854 B against
the 2 605 172 B baseline, +2 682 B, over the 2 000 B band. The growth is the
task's own and deliberate: the floor-geometry key module and its call sites
(src/floor-geometry-key.ts, houseplan-card.ts, clean-floor.ts) plus the
wrapper of the new lazy space-editor chunk that keeps the initial View graph
under its absolute budget (initial View 300 572 B gzip, headroom 494 B).
No other metric moved past its band (portMembers 347 -> 349, band 5).
Issue: #744
User-Visible: no
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_018qZfe7YS4rqEMKoVeS3GKd
The floor-geometry key of #744 (+~143 B gzip) put the initial View graph at
301 097 B gzip, 31 B over the absolute 301 066 B wall in
scripts/bundle-budget.mjs. The comment over that wall says the next growth
is paid by moving code into lazy graphs, not by raising the budget.
`src/space-card.ts` imported its Lovelace GUI config editor
(`src/space-editor.ts`, the `houseplan-space-card-editor` element)
statically, so every View paid for a form only the dashboard editor opens.
`getConfigElement()` now imports it on demand, exactly as
`houseplan-card.getConfigElement()` already does for `./editor`; Home
Assistant awaits the returned promise. The editor's dependencies (lit, i18n,
space-geometry) stay in the shared chunk, so the new `space-editor-*.js`
chunk holds only the element itself (~1.1 KB gzip).
Measured: initial View 301 097 -> 300 489 B gzip (-608 B), headroom to the
301 066 B budget 577 B. The beta ceiling (300 142 B, band 2 000 B above it,
#699) and the budget are unchanged; lazy editor/onboarding graphs are not
touched.
Issue: #744
User-Visible: no
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_018qZfe7YS4rqEMKoVeS3GKd
smoke_floor_geometry_cache failed in Validate run 36875756451 on one check,
ac2cPreviewWallStandsWhereAFreshCardDrawsIt, with every area check green.
The check read the moving wall only from the live layer (data-kind
"preview"). A host render during a held drag ends that layer: updated()
commits it, and nothing repaints it until the next accepted move. The
settled scene then draws the preview record itself, and its walls equal
those of a fresh card on that record. In the smoke, the "Room updated"
toast from the AC1 rename expires 3.5 s after the save, about when the drag
starts: locally the rename-to-drag gap is 3.47-3.6 s. On a faster runner
the expiry landed after the last move, so the check found no live path. A
light toggled in Home Assistant mid-drag gives the same red.
The drag now starts once the toast is gone, so the live frame normally
judges the live layer. If a host render still lands, the check judges the
settled union the same way; a stale union has none of the moved faces. A
new check takes the settled frame on purpose: a state change during the
held drag, then the union path and the areas must equal the fresh card's.
The live strips never pass through the floor-geometry caches, so the old
check stayed green with a floor key that ignores the preview. The new
check fails on that key, and a failure prints a diagnostic line naming the
judged layer and whether the settled union is the stored one.
Issue: #744
User-Visible: no
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_018qZfe7YS4rqEMKoVeS3GKd
The physical bodies, the wall union pool, the inner room contours and the
clean floor carried the global config epoch in their keys. Every edit of
any floor bumps it, so after one edit every other floor was cold again:
in large-house the first visit to an untouched floor rebuilt its wall
union and paid ~0.7 s flat / ~0.65 s 2.5D instead of ~40-55 ms.
A floor's geometry reads only its own config record (spaceModels) and
constants, so the key is now a content fingerprint of that record
(src/floor-geometry-key.ts), remembered per epoch and per record object.
The geometry also reads the current floor's config next to the model it
is given; when those records differ the key covers both. The live resize
preview is its own record, so preview frames get their own key; the
editor runtime seeds the pool and re-keys the bodies through the same
reader. The stairs editor no longer clears the clean floors of every
floor: the stairs are part of the floor's record.
The #735 switch-cycle guard now also sees the union pool and the inner
contours (optional members of the large-house card contract, so an older
comparison bundle reads 0). smoke_floor_geometry_cache proves the warm other floor and the
invalidation against an independent card (multi-floor push with shared
walls, a stair, a resize preview and its cancel); two mutants guard it.
Issue: #744
User-Visible: yes
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_018qZfe7YS4rqEMKoVeS3GKd
Validate on the conveyor's rebase eb439e81 failed preflight: the branch
changes visual sources (stairs), so the screenshot check is strict on it,
and the fingerprint was stale against the moved dev (#739, #742, #759).
The branch is rebased onto b84465e5 (inventory counts merged with #742:
lifecycle 89, total 204/200) and `docs:accept --identical` re-captured
all 11 frames: pixel-identical, only the source fingerprint moves.
Issue: #740
User-Visible: no
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_018qZfe7YS4rqEMKoVeS3GKd
A floor with stairs pays for them on every switch to it: the stair layer
is emptied on other floors, so Lit recreates every symbol on each return
and the browser lays out and paints it again. With 250 stairs (the
large-house fixture, the per-floor limit) that was 2,875 SVG elements and
about 40 ms per entry locally; each stair carried 3-7 separate tread lines
with four bound coordinates each.
The treads of one stair are now a single <path class="hp-stair-tread">
with one `M a L b` subpath per tread, in geometry order and with the
numbers the lines carried. Treads of one stair never overlap (straight:
parallel, >= 20 cm apart; spiral: inner ends >= 6.7 cm apart at the
3.6 cm stroke), so the path paints the same pixels at any opacity. The
outline points and the tread data are built once per cached geometry
object (cachedStairMarkup, weak keys), not on every render. The View and
plan-editor layers share the strings; outline, hit polygon, trapezoid,
arrow, attributes and handlers are unchanged. Floor 1 of the fixture
drops from 4,210 to 2,960 elements.
Witnesses: the unit test for the path data and its cache, and the
smoke_stairs markup checks in View and in the plan editor, are red on
dev. The stairs-view-tread-lines mutant restores the View lines.
Issue: #740
User-Visible: no
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_018qZfe7YS4rqEMKoVeS3GKd
Five places still described the pipeline as it was before code that is
already in dev:
- the S3 hint of the task packet told the author to push the branch, while
the spec lives in the issue body (§2.3, #517) and nothing is pushed
before S5 (§11.8);
- process-gate printed «FAIL п.9 Gates: light» for a trailer nobody writes
or reads, while §10.2 item 9 is the unimplemented release:prerelease
verdict check. The check is removed; a contract test ties every RULES key
to an implemented item of §10.2 and every finding number to a RULES key;
- §10.4 item 4 demanded a heredoc in run:, while #723/#730 and their tests
demand the opposite: commit messages echo line by line into a file,
comment and summary texts come from code;
- the ship merge comment, AUTHOR.md, REVIEWER.md and AGENTS.md named only
the pre-beta document, though since #727 the night reads ship code first;
- the nightly publication committed «docs: ship review for nightly …
перед бетой» with the beta step's Issue: #696. It now has its own
subject (the document name), body and Issue: #727; the beta message is
unchanged.
The browser-guard inventory note still said growth above 200 fails
mutation-gate --check; since #699 it is a guideline and --check warns. Its
counts now match the inventory: 205, lifecycle 90.
Issue: #748
User-Visible: no
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_018qZfe7YS4rqEMKoVeS3GKd
#747 landed while #743 was in flight and set the switchCycleMs hardMaxMs
of the 2.5D family to 1550. The backdrop budget is the historical isometric
budget under its own profile id (#743 test), so after the rebase its 8000
became the only difference; it now carries 1550 like the rest of the family.
Issue: #743
User-Visible: no
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_018qZfe7YS4rqEMKoVeS3GKd
No Full Performance profile walked the backdrop (imagePlan) path: every
large-house fixture has plan_url null. So the #739 K1 double render -- a
warm 2.5D floor switch with a backdrop cleared the ready paper, inserted
the veil, probed the card background and rendered a second time -- was
invisible to CI by time and structurally; a temporary probe found it.
large-house-isometric-backdrop-v1 is the twin of large-house-isometric-v1
with the shipped f1.svg under a URL of its own on every floor
(plan_aspect 1, room geometry unchanged). The variant is derived in the
runner as plan-snap's is, so demo/fixtures and the bundle fingerprint do
not change. A first stable frame without the backdrop image fails the
sample. After the switchCycle window and its #735 guard, before forced
GC and outside every timed window, a probe makes six warm switches and
counts performUpdate passes until updateComplete resolves true: the K1
second pass starts after the first updateComplete resolves, so a count
taken right after the first await reads one on both sides.
evaluate.mjs rejects a candidate of this profile unless every switch took
one pass, or when the probe is missing; the base is reported, not judged
(v1.78.0 and dev before #739 take two). The budget is a copy of the
historical isometric budget under the new profile id. performance.yml
gains the isometric-backdrop matrix entry with exact-SHA comparison.
Witness: a tree with #739 reverted reads perSwitch 2 in every sample and
benchmark:compare against the branch report throws on the pass count;
v1.78.0 reads 2, the branch reads 1.
Issue: #743
User-Visible: no
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_018qZfe7YS4rqEMKoVeS3GKd
claude[bot]