sindoring
ba81098797
Work-Package: M00-QAR-INTEGRATOR-CLOSEOUT-001 Agent: sindo-main-codex Tests: post-merge render baseline dry-run; comparator self-test; coordination and documentation gates Fidelity: records seven build-12340 reference categories and measured gaps without claiming parity
3.6 KiB
3.6 KiB
M00-QAR-VISUAL-DIFF-001 — Renderer checkpoint visual diff
Ownership
- Target: M00
- Program: QAR
- Owner/Agent ID: sindo-main-codex
- Branch:
work/sindo-main-codex/m00-checkpoint-diff - Lease expires UTC: 2026-07-13
- Integrator: milestone integrator
Outcome
Provide a deterministic, headless perceptual comparison for paired renderer checkpoint PNGs, with machine-readable results and synthetic regression coverage.
Non-goals
- Supplying or committing original-client screenshots.
- Claiming visual parity or replacing human approval.
- Image registration, camera correction, or temporal animation matching.
Paths
- Exclusive:
src/tools/compare_render_checkpoints.gd - Shared/hotspots:
src/tools/render_baseline_manifest.json,tools/run_render_baseline.ps1,docs/RENDER_BASELINE.md,docs/modules/world-renderer.md,RENDER.md - Generated/ignored: comparison reports, diff PNGs, original-client captures
Contracts and data
- Public API/events: headless CLI arguments and JSON report documented in renderer baseline docs
- Schema/format version: comparison report schema 1; baseline manifest schema remains 1
- Migration/compatibility: additive manifest budget fields
- Consumers: developers and CI
Dependencies
- Requires: Godot
ImageAPI and M00 checkpoint naming contract - Blocks: automated portion of original-client paired comparison
- External state: approved build 12340 screenshots remain unavailable
Verification
- Commands: headless self-test; manifest verification; documentation and coordination gates
- Fixtures: tool-generated identical and changed 2x2 PNGs
- Fidelity evidence: algorithm is deterministic; real-client evidence remains a manual/external input
- Performance budget: offline checkpoint operation, linear in pixel count
Documentation deliverables
- Inline public API docs: CLI usage in script header
- Module specification: update renderer verification and source map
- Data-flow diagram: update baseline documentation
- Sequence/state/dependency diagrams: not stateful or asynchronous; not applicable
- Source map/status updates: renderer module and RENDER notes
Simplicity and naming
- Important names introduced:
mean_perceptual_error,changed_pixel_ratio - Simplest considered solution: direct Godot Image iteration with no dependency
- Rejected complexity/abstractions: external image library, SSIM window pipeline, image registration
- Unavoidable complexity and justification: sRGB linearization avoids comparing encoded channel values directly
- Measured optimization evidence: not required for offline seven-checkpoint comparison
Status
- State: ready
- Done: comparator, runner integration, manifest tolerance contract, synthetic regression, five local build 12340 references, three viewpoint calibrations, corrected visual capture and first paired report
- Next: integrator review; coordinate/placement gap should be resolved before tolerance calibration
- Blocked by: real paired screenshots only for final human fidelity approval
Handoff
- Commit: branch HEAD
- Results: ten pairs compared with no missing candidates; mean error 0.0707..0.1746 and changed ratio 0.5504..0.8187; human review identified under-terrain/inside-geometry camera mismatches
- Remaining risks: coordinate/placement mismatch blocks meaningful tolerance calibration; synthetic animation, dusk and full human approval remain incomplete
- Documentation updated:
docs/RENDER_BASELINE.md,docs/modules/world-renderer.md,RENDER.md