Files
houseplan-card/docs/testing-notes/mutation-browser-guards.md
T
Claudeandclaude[bot] cf7a9cc4fc fix(process): reconcile canon and hints with the pipeline after #707–#730 (#748)
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
2026-10-01 16:49:55 +00:00

13 KiB

Browser guards mutation registry (#659)

This is the reviewed classification of every mutation witness that still needs a browser. The guideline is 200 (#699: a guideline, not a wall); mutation-gate --check, npm run inventory and the unit contract all read this same inventory. A new browser guard must be added deliberately under one reason below, and its mutant because must explain the concrete browser-only invariant.

Converted witnesses are not listed here: their registry guard names an explicit node --test suite plus test/mutation-browser-offload.test.mjs, and every converted mutant is run once to prove that the Node witness actually kills it.

Category Count Why a browser is still required
Performance threshold 4 The witness measures real browser wall-time or frame work; a pure assertion cannot prove the budget.
Browser harness integrity 4 The mutation breaks page-error, round-trip or page-registration observation in the browser harness itself.
Paint, cascade and layer composition 26 The invariant depends on computed CSS, SVG paint, clipping, stacking or pixels produced by Chromium.
Pointer geometry and trusted interaction 45 The invariant depends on hit testing, pointer capture, touch/keyboard dispatch or live DOM geometry.
Responsive DOM layout 36 The invariant depends on measured element boxes, responsive breakpoints, native/HA dialog shells or focusable target size.
Custom-element and HA browser lifecycle 90 The invariant crosses Lit/custom-element lifecycle, browser storage/events, lazy loading or a complete HA-card state transition.
Total 205 / 200 Above the guideline mutation-gate --check warns rather than fails (#699); each guard above it is held by its own reason in this inventory and its because.

Measured effect

The measurements below are GitHub-hosted Linux runs from 2026-09-27. They are recorded separately from the policy limit: the limit proves that browser guards cannot grow unnoticed, while timings show the effect actually observed on the shared runners.

Measurement Before After Observed change
Full-registry wall time 54:18 (1,017 mutants) 48:08 (1,028 mutants) -6:10 (-11.4%), despite 11 additional mutants
Sum of the six Each test catches its breakage steps 4:17:19 4:12:32 -4:47 (-1.9%)
Longest full-registry shard step 50:45 45:13 -5:32 (-10.9%)
Same converted-witness slice, aggregate time 33:43.9 12:07.7 -21:36.2 (-64.0%)
Same converted-witness slice, critical shard 8:05.1 2:19.3 -5:45.8 (-71.3%)

The candidate slice compares the same 83 converted mutant ids in the baseline full-registry log and in the S7 candidate Validate 36316464432. It sums the interval between consecutive per-mutant completion timestamps on each of the six shards. One of the 84 converted ids, active-tab-not-revealed, was the first completed mutant in its shard, so the log cannot isolate its time from clean-guard setup and it is excluded from both sides. The complete candidate run, which also selected registry/tooling mutants outside this converted slice, finished in 10:03.

The original estimate of three to four hours saved on a full-registry run is not supported by this A/B evidence. The directly converted slice is materially faster, but registry growth and shared-runner variance consume most of that gain in the end-to-end total; only the measured reductions above are claimed.

Reviewed per-mutant inventory

Performance threshold

The witness measures real browser wall-time or frame work; a pure assertion cannot prove the budget.

  • junction-limit-p3-quadratic-again
  • junction-limit-p4-bruteforce-again
  • wall-draw-full-preflight-again
  • wall-draw-wall-artifact-discarded

Browser harness integrity

The mutation breaks page-error, round-trip or page-registration observation in the browser harness itself.

  • benchmark-page-verdict-unwatched
  • report-page-errors-skips-round-trip
  • smoke-guard-blind-to-tail
  • smoke-guard-forgets-to-register-pages

Paint, cascade and layer composition

The invariant depends on computed CSS, SVG paint, clipping, stacking or pixels produced by Chromium.

  • daycycle-outline-not-promoted
  • daycycle-outline-promoted-on-inner-paper
  • daycycle-static-outline-promoted-on-inner-paper
  • decor-restored-below-room-fills
  • device-keyboard-bypasses-click-path
  • device-long-value-ellipsis-restored
  • device-unavailable-hover-restored
  • golden-filled-tunnel-removed
  • golden-lamp-out-of-reach
  • hatch-static-renderer-untouched
  • hatch-stroke-not-scaled
  • hatch-zoom-compensation-back
  • iso-first-frame-reveals-flat-during-lazy-load
  • iso-room-label-44-box-centres-name
  • iso-sun-card-drops-occluders
  • iso-sun-flat-wedges-remain
  • iso-theme-dark-wall-rule-returns
  • stage3-w4-device-target-loses-44px-floor
  • stage3-w5-runtime-nudge-writes-storage
  • stage3-w6-no-borders-keeps-raised-plates
  • stage3-w7-sun-state-enters-structural-key
  • sun-ray-origin-cache-ignored
  • sun-ray-origin-save-forced-inner
  • value-static-icon-keeps-live-vacuum
  • value-static-icon-keeps-route-warning
  • value-static-icon-keeps-vacuum-overlay

Pointer geometry and trusted interaction

The invariant depends on hit testing, pointer capture, touch/keyboard dispatch or live DOM geometry.

  • align-guides-exclude-dead-source
  • align-point-reads-frozen-snapshot
  • decor-default-style-debounce-cut
  • decor-default-style-seed-cut
  • decor-keyboard-nudge-drops-focus-dialog-guards
  • decor-keyboard-nudge-reruns-magnet
  • dense-device-hit-browser-skips-painted-priority
  • double-fit-bypasses-canonical-fit-all
  • furniture-art-editor-adopt-skipped
  • furniture-edge-handles-steal-the-corner
  • furniture-exterior-surface-removed
  • furniture-shift-listeners-not-attached
  • furniture-wall-runtime-drops-drag-side
  • furniture-wall-runtime-drops-raw-intent
  • live-pinch-compositor-demoted-on-active-lit-commit
  • opening-dimension-overlay-hidden
  • opening-search-hides-none
  • opening-search-select-not-wired
  • placement-accepts-any-mouse-button
  • reorder-skips-materialization
  • resize-history-boundary-repair-removed
  • resize-label-uses-old-room-gear-centre
  • resize-labels-hide-narrow-area
  • resize-pointer-capture-removed
  • resize-preview-reject-silent
  • room-fit-html-overlay-jumps-ahead
  • room-fit-pan-release-reaccepted
  • room-fit-persists-zoom
  • safe-resize-commit-preflight-bypassed
  • sections-editor-transition-uses-window-height
  • sections-grid-drops-stage-flex-chain
  • sections-grid-falls-back-to-viewport-height
  • sections-resize-drops-stage-refit-observer
  • space-card-decor-capability-change-not-adopted
  • space-card-decor-capability-downgrade-does-not-clear-assets
  • tab-drag-outlives-the-card
  • tab-drag-survives-release-outside
  • tab-drag-target-follows-captured-source
  • tab-drop-indicator-always-before
  • tab-drop-outside-commits-last-target
  • tab-reorder-not-persisted
  • touch-pinch-marker-hold-rearmed
  • touch-pinch-zoom-persists-per-frame
  • unavailable-toggle-stays-silent
  • wall-draw-rejection-rollback-skipped

Responsive DOM layout

The invariant depends on measured element boxes, responsive breakpoints, native/HA dialog shells or focusable target size.

  • dialog-ha-disconnected-reject-reopen-enabled
  • dialog-ha-rejected-close-reopen-disabled
  • dialog-native-reconnect-recovery-disabled
  • dialog-native-surface-stretches-to-viewport
  • dialog-native-update-recovery-disabled
  • fit-house-hidden-walls-vote
  • grid-scale-imperial-roundtrip-drift
  • grid-scale-opening-hit-unscaled
  • grid-scale-opening-symbol-unscaled
  • grid-scale-plan-chrome-unscaled
  • grid-scale-static-factor-missing
  • header-menu-item-below-44
  • header-menu-outside-tap-reaches-plan
  • hidden-room-names-compact-svg-fallback
  • hidden-room-names-full-svg-fallback
  • hidden-room-names-iso-override
  • hp-dialog-ignores-flex-content
  • panel-readonly-empty-bypasses-write-capability
  • phone-editor-close-hidden-with-mode-buttons
  • phone-header-at-tablet-width
  • phone-header-shows-title
  • phone-header-wraps
  • static-card-descendants-hit-testable
  • summary-dialog-drops-flex-content
  • summary-picker-hot-add-stays-filtered
  • tab-editing-without-write-access
  • toolbar-active-highlight-stops-at-tab
  • toolbar-close-back-inside-tab
  • toolbar-close-hidden-at-medium-width
  • toolbar-close-slot-collapses-outside-editor
  • toolbar-close-slot-idle-focusable
  • toolbar-close-slot-leaves-mode-group
  • toolbar-close-target-below-24
  • toolbar-device-count-returns
  • toolbar-mode-zoom-gap-regresses
  • warm-dialog-drops-transferred-baseline

Custom-element and HA browser lifecycle

The invariant crosses Lit/custom-element lifecycle, browser storage/events, lazy loading or a complete HA-card state transition.

  • accepted-marker-rolled-back-by-layout-failure
  • area-relocation-clears-whole-history
  • area-relocation-loses-position-on-refusal
  • backdrop-busy-dismiss-races-decision
  • backdrop-downscale-drops-alpha
  • backdrop-phase2-falls-back-to-original
  • backdrop-probe-always-safe
  • barrier-cache-never-invalidated
  • camera-anchor-from-presented
  • camera-cancel-loses-zoom
  • child-readd-clears-parent-tombstone
  • cold-view-toggle-delegated-to-runtime
  • cold-view-vacuum-mapid-delegated
  • color-picker-confirm-click-through
  • color-picker-invalid-confirm-latch-removed
  • config-updated-event-ignored
  • confirm-dialog-loses-alertdialog
  • current-rejected-physical-write-keeps-optimistic-wall
  • danger-confirm-back-into-the-branch
  • danger-confirm-lost-space-request-guard-removed
  • danger-confirm-lost-space-transition-cancel-removed
  • danger-confirm-uses-last-rendered-language-gate
  • danger-confirm-warm-language-guard-removed
  • danger-confirm-warm-transition-cancel-removed
  • daycycle-programmatic-camera-skips-safe-outline
  • device-echo-keeps-local-noncanonical
  • device-focus-tooltip-blur-cleanup-removed
  • device-focus-tooltip-handler-removed
  • device-focus-tooltip-room-hover-overwrites
  • device-inbox-batch-rollback
  • household-enter-stops-acting
  • household-marker-drops-keyboard-reach
  • device-markers-rendered-without-keys
  • device-pointer-leave-clears-focus-fallback
  • device-position-cancel-routed-to-commit
  • discovery-reset-writes-a-copy
  • editor-neutral-escape-does-not-exit
  • empty-space-cleanup-disabled
  • fixed-floor-transition-guard-bypassed
  • french-locale-wrong-dictionary
  • hp-dialog-escape-does-not-close
  • junction-limit-baseline-cache-stale
  • junction-limit-candidate-fail-open
  • junction-limit-write-gate-removed
  • locale-failure-toast-dropped
  • marker-reject-keeps-optimistic-candidate
  • marker-rollback-keeps-enqueue-time-revision
  • moon-static-clock-tick-off
  • namespace-loader-returns-english
  • near-axis-authoring-snap-bypassed
  • onboarding-loader-skips-namespace-ensure
  • openings-rendered-without-keys
  • plan-only-preview-label-hidden
  • plan-room-area-icon-hidden
  • plan-room-area-icon-navigates
  • plan-upload-client-limit
  • plan-upload-guard-original
  • plan-upload-reduced-over-limit-staged
  • post-write-tail-runs-on-refused-gate
  • readonly-view-syncs-new-devices
  • render-invalidation-renders-irrelevant-ha
  • render-reconcile-restarts-editor-runtime-cycle
  • reopened-room-from-registry-space
  • room-accept-leaves-coincident-partitions
  • room-gear-second-touch-keeps-drag
  • room-settings-click-does-not-open
  • room-tooltip-off-skips-pointer-modality
  • rooms-rendered-without-keys
  • rooms-rendered-without-space-key
  • same-space-room-change-recenters
  • space-card-rooms-rendered-without-keys
  • space-create-hidden-display-override
  • stairs-view-pan-opens-target-floor
  • support-invalid-response-leaks-issued-token
  • support-stale-preview-response-revives-consent
  • support-timeout-claims-success
  • vacuum-overlay-back-to-the-dock-space-filter
  • volumetric-kiosk-ignores-setting
  • view-current-space-aria-removed
  • wall-face-apply-skips-overlap-guard
  • wallthick-hit-narrowed
  • warm-pending-mode-leaves-revive-waiting
  • writer-history-skips-finished-chain-normalization
  • zigbee-topology-endpoint-cleanup-skipped
  • zigbee-topology-endpoint-elevation-removed
  • zigbee-topology-hovered-endpoint-elevation-removed
  • zigbee-topology-overlay-double-live-projection
  • zigbee-topology-overlay-layer-lowered
  • zigbee-topology-unknown-casing-gaps-filled
  • zigbee-topology-unknown-casing-removed
  • zigbee-topology-unrelated-markers-raised