feat(consolidation): add sealed scoreboard v0 - #60
Merged
Conversation
Add the fixed 14-row SVG rect/path corpus, hardened Chromium bake, callable chassis frame seam, strict report and baseline contracts, CI shape gate, and the visibly unratified FLIP proposal. Captured-essence ledger — legacy seam: preserve the existing grida_dev SVG render path, white-background and embedded-font assumptions, geometry-fit behavior, and existing lossy unsupported posture; the legacy engine remains context, never the oracle. Captured-essence ledger — chassis seam: rehome render-bin composition as separate explicit Base and Sample entries over the moving latest bounded rect/path proving profile; strict resolver reports remain errors and callers retain allocation, clear color, I/O, cadence, and encoding. Captured-essence ledger — comparator and oracle: retain threshold-zero dify anti-alias classification, integer pixel counts, white background, no mask, exact hashed SVG buffers, JavaScript- and network-disabled Chromium, duplicate byte-identical captures, canonical no-symlink paths, and create-new publication. Captured-essence ledger — rehomes and drops: rehome only callable host composition; drop no model, schema, freeze, or legacy behavior and introduce no general SVG, HTML, .grida, n0 XML, animation-sample bridge, score, baseline, or FLIP decision. Gates: grida_dev tests 19/19; n0 and n0-model suites green; five-crate cargo check; warning-denied grida_dev and n0 clippy; scoreboard check 14/14; Chromium bake 14/14 with unchanged PNG hashes; chassis host sweep 14 fixtures and 56 artifacts byte-identical; sealed run and bless probes produced no output; Rust and TypeScript symlink probes refused; formatting, lint, workflow YAML, local links, diff, and frozen-surface checks green.
Etiology: validation collapsed per-engine entry-point rejection into whole-run failure, leaving UNSUPPORTED schema-only; cooperative deadline polls could not preempt a blocked synchronous stage. The contract now retains typed preflight dispositions and keeps durable report publication outside a terminal watchdog worker. Captured-essence ledger: preserve strict two-engine acceptance for scoreboard check; authorized reports retain every denominator row as scored, unsupported, or error; preflight rejection is unsupported and post-accept failure remains error; the fixed budget starts before corpus, oracle, parser, renderer, comparator, and prior-baseline validation. No renderer, corpus, oracle, score, baseline, rule, model, schema, freeze surface, or existing host behavior changed; deliberate drops: none. Gates: scoreboard tests 18/18; grida_dev tests 22/22; scoreboard check 14/14; five-crate cargo check; warning-denied grida_dev and n0 clippy; cargo fmt; diff and frozen-surface checks; OSS scan clean; seal tests green and report/baseline candidate absent.
Etiology: the seal tests proved the guard helper rejected injected unratified rules, but did not assert the production active_rule configuration itself. An accidental ratified value could therefore enable score-producing paths without failing the CI shape gate. Lock the real configuration to None while FLIP remains unratified. No renderer, corpus, oracle, report, baseline, rule, model, schema, or frozen surface changed; no score was produced or inspected. Gates: scoreboard tests 19/19; scoreboard check 14/14; five-crate check; warning-denied grida_dev and n0 clippy; n0 and n0-model suites; rustfmt and diff hygiene.
|
The latest updates on your projects. Learn more about Vercel for GitHub.
|
|
Caution Review failedThe pull request is closed. ℹ️ Recent review info⚙️ Run configurationConfiguration used: Organization UI Review profile: CHILL Plan: Pro Run ID: ⛔ Files ignored due to path filters (29)
📒 Files selected for processing (22)
WalkthroughChangesConsolidation tooling and gates
Estimated code review effort: 5 (Critical) | ~120 minutes Sequence Diagram(s)sequenceDiagram
participant CI
participant ScoreboardCLI
participant CorpusValidator
participant LegacyEngine
participant ChassisEngine
participant ReportWriter
CI->>ScoreboardCLI: run scoreboard check/preflight
ScoreboardCLI->>CorpusValidator: validate corpus and oracle
CorpusValidator-->>ScoreboardCLI: ValidatedCorpus
ScoreboardCLI->>LegacyEngine: render fixtures twice
ScoreboardCLI->>ChassisEngine: render fixtures twice
LegacyEngine-->>ScoreboardCLI: engine cells and PNGs
ChassisEngine-->>ScoreboardCLI: engine cells and PNGs
ScoreboardCLI->>ReportWriter: validate and write report
ReportWriter-->>CI: scoreboard gate result
Possibly related issues
✨ Finishing Touches📝 Generate docstrings
🧪 Generate unit tests (beta)
Thanks for using CodeRabbit! It's free for OSS, and your support helps us grow. If you like it, consider giving us a shout-out. Comment |
Etiology: the new and amended Markdown tables had not been normalized by the repository formatter, so the Node workflow stopped at pnpm fmt:check before its downstream steps. Format only the three branch-owned Markdown files; prose and contracts are unchanged. Gates: pnpm fmt:check; oxlint --deny-warnings; www type-check; reftest type-check, build, and 51-test suite. The local Next production build stalled without diagnostics during Turbopack optimization and was stopped; remote Ubuntu CI remains the build gate.
Etiology: the scoreboard-shape job compiled the graphics stack under Cargo test, then cargo run selected the dev profile and began a second cold native build. The 30-minute job envelope cancelled that build before the corpus preflight could execute. Run the CLI preflight under the already-built test profile. This preserves both required checks and the hard job timeout; it changes no renderer, corpus, rule, report, baseline, or score behavior. Captured essence: synthetic contract tests and the real scoreboard check command both remain mandatory. Deliberate drops: none. Gate evidence: pnpm fmt:check; cargo test --locked -p grida_dev scoreboard (19 passed); cargo run --locked --profile test -p grida_dev -- scoreboard check (14/14 preflight); git diff --check.
softmarshmallow
marked this pull request as ready for review
July 21, 2026 08:28
This file contains hidden or bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
1 participant
Add this suggestion to a batch that can be applied as a single commit.This suggestion is invalid because no changes were made to the code.Suggestions cannot be applied while the pull request is closed.Suggestions cannot be applied while viewing a subset of changes.Only one suggestion per line can be applied in a batch.Add this suggestion to a batch that can be applied as a single commit.Applying suggestions on deleted lines is not supported.You must change the existing code in this line in order to create a valid suggestion.Outdated suggestions cannot be applied.This suggestion has been applied or marked resolved.Suggestions cannot be applied from pending reviews.Suggestions cannot be applied on multi-line comments.Suggestions cannot be applied while the pull request is queued to merge.Suggestion cannot be applied right now. Please check back later.
Outcome
Installs the consolidation program's Phase 2 measurement instrument before any score exists. This gives both engines one Chromium-anchored, CI-checked contract for future capability grants while keeping scoring owner-sealed.
This PR:
The follow-up commit makes rejected preflight rows typed
UNSUPPORTEDevidence while post-accept failures remain errors, and moves every future score-producing stage behind a terminal watchdog. The final test locks the production rule configuration itself to sealed.Governance and seal
FLIP remains open at #49.
scoreboard runandscoreboard blessrefuse before corpus, renderer, report, or baseline access because the production active rule is absent; CI locks that state and invokes only synthetic contract tests plusscoreboard check.No scoreboard score was produced or inspected. No report, committed baseline, or baseline candidate exists. The committed PNGs are Chromium oracle inputs, not scores or renderer outputs.
This PR installs the instrument but does not ratify FLIP, exit Phase 2, grant general SVG capability, or flip any conformance bar. After owner ratification or revision of FLIP, the first authorized report and baseline are a separate gated step.
Captured-essence ledger
Scope inventory: the existing legacy SVG render seam; the chassis's bounded static rect/path Base/Sample composition; the fixed SVG sources; Chromium oracle inputs and bake provenance; comparison, coverage, report, baseline, regression, wall-clock budget, and CI-shape contracts.
Retained caveats: legacy rendering keeps its white-background, embedded-font, geometry-fit, and existing unsupported posture. Chassis resolution stays strict and callers retain allocation, clear color, I/O, cadence, and encoding. The comparison retains threshold-zero dify antialias classification, integer pixel counts, white background, no mask, exact source hashes, JavaScript- and network-disabled Chromium, duplicate byte-identical captures, canonical no-symlink paths, and create-new publication.
Re-home: only host composition becomes callable. Entry-point rejection is retained as typed unsupported coverage; accepted-entry failures remain errors. Corpus/oracle/parser/render/comparator/prior-baseline work is enclosed by the hard budget, while report publication remains outside the worker so partial or over-budget work cannot publish.
Deliberate drops: none. No model, schema, frozen surface, legacy behavior, general SVG entry, HTML entry, n0 XML entry, format bridge, animation-sample bridge, score, baseline, or registry decision is introduced.
The corpus is deliberately a bounded proving shell. Its exclusion ledger names local-only corpora, general SVG import, HTML, n0 XML, the frozen format, and legacy exact-time animation rather than hiding those gaps.
Gate evidence
All validation below is score-incapable:
cargo test --locked -p grida_dev scoreboard— 19 contract/seal testscargo run --locked -p grida_dev -- scoreboard check— 14/14 source/oracle hashes, dimensions, and both preflights; no rasterization or comparisoncargo test --locked -p n0 -p n0-modelcargo checkgrida_devandn0clippyThe original instrument commits also record deterministic Chromium bake and chassis host-artifact sweeps. Those are input/host equivalence gates, not scoreboard results.