Plan 021 workstream A: parity-inventory refresh audit (M4 entry criteria)

[charliek]

What this is

The M4 entry-criteria audit the roadmap itself requested: every row of docs/development/iced-parity-inventory.md re-verified against current behavior (main@166d2d6) with named evidence, plus the one harness repair the audit needed to produce that evidence. Docs + one harness file; no product code.

Changes

Inventory refresh (iced-parity-inventory.md):

• Six rows were stale in iced's favor and closed: the Sidebar-footer P0 and the create / delete / reorder / New-Project-command / workspace-shortcut functional rows all described the pre-plan-010 world.
• Drifted prose corrected in closed rows and the non-table sections (chrome colors, footer deferral note, ordered-slices item 6, palette-feasibility closing).
• Four rows added under the audit's "is there a row for this at all?" question: terminal IME (all three UIs lack it — GTK has no IMContext on the terminal, Swift never adopts NSTextInputClient; iced side is E6, in flight under plan 021), window vibrancy (P2 → 3h), context menus (P2), terminal bell (recorded all-three-UIs absence).
• Dated method-and-evidence block: suite runs (shed GTK 171 passed; iced X11 80 passed on both wgpu and tiny-skia; macOS iced 54 passed), digest-pinned parity captures, file:line refs.

Roadmap M4 entry-criteria rewrite (iced-migration-roadmap.md): the audited open set — open P0: none; open P1 = the user-directed 3h polish cluster + upstream-blocked drops (#302) + terminal IME (E6) + the badge-color one-constant fix (#311, found by this audit: iced #4e9af1 vs the references' deliberate #007aff). #284 recommendation recorded: waive the cross-toolkit golden-image gate (evidence + revisit trigger in the doc and on the issue) — final call stays with Charlie; issue left open.

Harness repair (tools/screenshot/parity.py + unit tests): current shed cairo/pango broke measure_agent_palette's exact pixel matches against visually correct captures (alpha-composite rounding #48484E→#48484D; AA text off its constants). Scoped fix: ±2 tolerance on the alpha-composited selection (both match sites), semantic ink predicates for the two text presence scans. Geometry assertions and all other color matches stay exact.

Verification

• harness-unit: 42 tests OK (locally, same command as CI).
• Shed gate: GTK + iced parity captures green post-fix (run 166d2d6-96c9832705, full measurements.json both targets).
• Evidence suite runs listed above; markdown table structure machine-checked.
• Cursor reviews (Codex usage-limited): harness diff — no real bugs, 2 coverage findings fixed, 1 pre-existing scope skip; docs diff — 4 citation/consistency findings fixed, 2 nits fixed, dispositions in the commit messages.

Impact

No dependency, privacy, or secret changes. No product behavior changes. Accepted risk: the parity-measurement tolerances are deliberately scoped (selection ±2, text-presence predicates only); everything else stays exact-match.

Plan 021 § Workstream A (authoritative spec for this PR)

The audit re-verifies EVERY parity-inventory row with named evidence (in-process screenshots + IPC dumps, both UIs, Mac + shed Linux), closes stale rows, states the honest open P0/P1 set in the roadmap's M4 paragraph, and records an evidence-backed #284 recommendation — the final call is Charlie's, flagged in the morning checklist; subjective polish rows are marked "needs Charlie", not guessed. Evidence discipline: every verdict cites a suite run + commit + counts, a digest-pinned capture, or a file:line ref. Full plan: ~/.claude/plans/roost/021-audit-and-ime.md (session-local).

Sibling PR (workstream B, E6 IME input in iced) follows separately.

🤖 Generated with Claude Code

https://claude.ai/code/session_01SgxiEBQTqgNjPswKqcs12d

Summary by CodeRabbit

• Documentation

  • Updated the visual-parity roadmap and inventory with current audit results, remaining gaps, completed work, and CI gate recommendations.
  • Documented decisions and follow-up items for IME support, context menus, terminal behavior, tab badges, and other polish areas.

• Tests

  • Added coverage for tolerant color matching, anti-aliased text, palette measurement, and visual-bound detection.

• Bug Fixes

  • Improved visual comparisons to recognize slight color variations and anti-aliased text, producing more reliable parity measurements.

[coderabbitai]

📝 Walkthrough

Walkthrough

The change improves screenshot parity detection for rounded colors and anti-aliased text. It adds focused tests and refreshes the Iced parity inventory and migration roadmap with current audit results, remaining gaps, and CI gate recommendations.

Changes

Visual parity validation

Layer / File(s)	Summary
Tolerant pixel classification and tests tools/screenshot/parity.py, tools/roosttest_unit/test_visual_parity.py	Adds RGB tolerance and semantic ink classifiers. Updates agent palette measurement and tests rounded palette colors, anti-aliased text, status text, and trailing ink.
Parity audit and roadmap records docs/development/iced-parity-inventory.md, docs/development/iced-migration-roadmap.md	Records refreshed audit evidence, completed project and interaction work, remaining parity gaps, issue references, and the cross-toolkit CI gate waiver recommendation.

Estimated code review effort: 3 (Moderate) | ~20 minutes

Possibly related PRs

• charliek/roost#269: Extends the same agent-palette visual parity and screenshot-testing area.
• charliek/roost#305: Updates the same Iced parity documentation and pixel-guard tooling.
• charliek/roost#307: Covers related renderer and pixel-identical parity validation.

🚥 Pre-merge checks | ✅ 4 | ❌ 1

❌ Failed checks (1 warning)

Check name	Status	Explanation	Resolution
Docstring Coverage	⚠️ Warning	Docstring coverage is 7.69% which is insufficient. The required threshold is 80.00%.	Write docstrings for the functions missing them to satisfy the coverage threshold.

✅ Passed checks (4 passed)

Check name	Status	Explanation
Description Check	✅ Passed	Check skipped - CodeRabbit’s high-level summary is enabled.
Title check	✅ Passed	The title clearly identifies the Plan 021 parity-inventory refresh and M4 entry-criteria audit, which are the primary changes.
Linked Issues check	✅ Passed	Check skipped because no linked issues were found for this pull request.
Out of Scope Changes check	✅ Passed	Check skipped because no linked issues were found for this pull request.

✨ Finishing Touches 💡 1

📝 Generate docstrings 💡

• Create stacked PR
• Commit on current branch

🧪 Generate unit tests (beta)

• Create PR with unit tests
• Commit unit tests in branch feature/plan-021-parity-audit

---

_{Comment @coderabbitai help to get the list of available commands.}

[charliek]

CodeRabbit cloud reported rate-limited on this PR, so the review ran via the local CLI (coderabbit review --committed --base main, the documented fallback). Three major findings, all addressed in 73e9b5f:

1. is_red_dominant_ink admitted the waiting orange (240,160,64) — real; the predicate gains a near-equal g/b term (abs(g-b) <= 30) separating failed red from waiting orange, with lifecycle-color regression vectors. (Not reachable in the current fixture — waiting ink never appears inside the selected failed row the scan is bounded to — but the classifier is now correct on its own terms. The pre-tightening shed gate's green run stands: every observed red-text pixel in the real capture has g≈b, so the added term cannot flip that result.)
2. Truncated digests pin nothing — full 64-hex SHA-256 digests published with per-run manifest provenance.
3. M4 waiver semantics ambiguous — the roadmap now states explicitly: the audit waives nothing; No CI gate for GTK↔Iced visual parity (capture tooling is human-reviewed only) #284's recommendation covers only the CI-gate question; per-row waivers are owner decisions recorded on the row before M4 can be declared entered.

[coderabbitai]

🧹 Nitpick comments (1)

tools/screenshot/parity.py (1)

61-64: 🎯 Functional Correctness | 🔵 Trivial | ⚡ Quick win

Add strict RGB tuple matching.

The project requires Python 3.12 or newer. Add strict=True to zip(value, target) at tools/screenshot/parity.py:64 to prevent partial comparisons when tuple lengths differ.

🤖 Prompt for AI Agents

Verify each finding against current code. Fix only still-valid issues, skip the
rest with a brief reason, keep changes minimal, and validate.

In `@tools/screenshot/parity.py` around lines 61 - 64, Add strict tuple-length
validation in the near function by passing strict=True to zip(value, target),
ensuring RGB comparisons never silently ignore mismatched tuple elements.

Source: Linters/SAST tools

🤖 Prompt for all review comments with AI agents

Verify each finding against current code. Fix only still-valid issues, skip the
rest with a brief reason, keep changes minimal, and validate.

Nitpick comments:
In `@tools/screenshot/parity.py`:
- Around line 61-64: Add strict tuple-length validation in the near function by
passing strict=True to zip(value, target), ensuring RGB comparisons never
silently ignore mismatched tuple elements.

---

ℹ️ Review info

⚙️ Run configuration

Configuration used: Path: .coderabbit.yaml

Review profile: CHILL

Plan: Pro

Run ID: ff1cd976-acbc-469f-8997-7297a6d76461

📥 Commits

Reviewing files that changed from the base of the PR and between 166d2d6 and 73e9b5f.

📒 Files selected for processing (4)

• docs/development/iced-migration-roadmap.md
• docs/development/iced-parity-inventory.md
• tools/roosttest_unit/test_visual_parity.py
• tools/screenshot/parity.py