styleproof 3.8.0 → 3.10.0

package/CHANGELOG.md

@@ -7,6 +7,45 @@ this project adheres to [Semantic Versioning](https://semver.org/spec/v2.0.0.htm
 
 ## [Unreleased]
 
+## [3.10.0] - 2026-07-06
+
+### Added
+
+- **Determinism provenance — a green needs a proven-deterministic capture.** Source-of-
+  truth step 2: a clean diff of two _nondeterministic_ captures is meaningless (they
+  might just happen to match). The capture now records its determinism basis in the
+  ledger — `self-checked` (captured twice, styles matched — a drift would have failed the
+  capture), `replayed` (rendered against a recorded HAR, deterministic by construction),
+  or `unproven` (neither) — and `styleproof-diff`:
+  - **blocks (exit 1) when either side is `unproven`**, even on an empty style diff;
+  - prints `✓ determinism proven — base X, head Y`, `✗ determinism NOT proven …`, or
+    `⚠ determinism basis unknown` (an older bundle with no field — degrades, never a
+    false red). The record-then-replay flow (base `self-checked`, head `replayed`) is
+    proven, as it should be.
+
+## [3.9.0] - 2026-07-05
+
+### Added
+
+- **Coverage provenance — a green now states its completeness basis.** Source-of-truth
+  step 1: a green from `styleproof-diff` used to silently imply completeness it couldn't
+  back up (it certified only the surfaces it happened to capture). Now the capture writes
+  a coverage ledger (`styleproof-coverage.json`) into the bundle — the declared
+  `expected` registry, or `null` — and `styleproof-diff` reads it and:
+  - **blocks (exit 1) when a registered surface was never captured**, even if the style
+    diff is empty — the one failure the gate couldn't catch before (a green over a
+    surface it never looked at). This checks the surfaces actually captured, so a
+    declared surface whose capture _failed_ is caught here, where the suite guard (which
+    checks the declared list) cannot see it;
+  - prints the basis on every run: `✓ coverage complete — all N registered surface(s)
+captured`, `✗ coverage INCOMPLETE — …`, or `⚠ completeness NOT asserted` (no
+    registry — certifies only the captured surfaces, so declare `expected` to certify
+    completeness). A crawl records `expected: null` honestly (it captures what the nav
+    links to, not proven-every-route).
+- `isMapFile` / `RESERVED_BUNDLE_FILES` (map-store) — one place that knows which bundle
+  files are surface maps vs metadata sidecars, so a new sidecar can't read as a phantom
+  "new surface".
+
 ## [3.8.0] - 2026-07-05
 
 ### Added

package/bin/styleproof-diff.mjs

@@ -41,6 +41,8 @@ import {
 } from '../dist/cli-errors.js';
 import { loadStyleMap } from '../dist/capture.js';
 import { auditRunInventory } from '../dist/inventory.js';
+import { auditCoverage, auditDeterminism, COVERAGE_LEDGER } from '../dist/coverage.js';
+import { isMapFile } from '../dist/map-store.js';
 
 const COMMAND = path.basename(process.argv[1] ?? 'styleproof-diff').replace(/\.mjs$/, '');
 
@@ -55,7 +57,7 @@ const COMMAND = path.basename(process.argv[1] ?? 'styleproof-diff').replace(/\.m
 function dirInventories(dir) {
   return fs
     .readdirSync(dir)
-    .filter((f) => f !== 'styleproof-manifest.json' && /\.json(\.gz)?$/.test(f))
+    .filter(isMapFile)
     .map((f) => loadStyleMap(path.join(dir, f)))
     .map((m) => (m.inventory ? { inventory: m.inventory } : {}));
 }
@@ -110,6 +112,73 @@ function printInventoryAudit(audit) {
   return unexplained.length;
 }
 
+// ── coverage provenance (the completeness basis of a green) ──────────────────────
+// The head bundle carries a coverage ledger (the declared registry). The gate audits
+// the ACTUALLY-captured surfaces against it, so "clean" states its basis — complete vs
+// the registry, or explicitly "not asserted" — instead of silently implying completeness.
+
+// Surface keys captured in a dir (file `<key>@<width>.json[.gz]` → `<key>`, deduped).
+function capturedSurfaceKeys(dir) {
+  return [
+    ...new Set(
+      fs
+        .readdirSync(dir)
+        .filter(isMapFile)
+        .map((f) => f.replace(/@\d+\.json(\.gz)?$/, '')),
+    ),
+  ];
+}
+
+function readLedger(dir) {
+  const p = path.join(dir, COVERAGE_LEDGER);
+  if (!fs.existsSync(p)) return null;
+  try {
+    return JSON.parse(fs.readFileSync(p, 'utf8'));
+  } catch {
+    return null; // a corrupt ledger reads as no registry → "not asserted", never a false green
+  }
+}
+
+// Print the completeness verdict; return true if it BLOCKS (a registered surface is missing).
+function printCoverageVerdict(v) {
+  if (v.basis === 'complete') {
+    console.log(`\n✓ coverage complete — all ${v.registrySize} registered surface(s) captured`);
+    return false;
+  }
+  if (v.basis === 'unasserted') {
+    console.log(
+      '\n⚠ completeness NOT asserted — the spec declared no `expected` registry, so this certifies only the\n' +
+        '  surfaces that were captured, not that they are all of them. Declare `expected` to certify completeness.',
+    );
+    return false;
+  }
+  console.log(
+    `\n✗ coverage INCOMPLETE — ${v.uncovered.length} registered surface(s) not captured (of ${v.registrySize}):`,
+  );
+  for (const k of v.uncovered) console.log(`  ✗ missing: ${k}`);
+  console.log(
+    "  → capture each (or move it to `exclude` with a reason). A green can't certify what was never captured.",
+  );
+  return true;
+}
+
+// Print the determinism verdict; return true if it BLOCKS (a side's capture was unproven).
+function printDeterminismVerdict(v) {
+  if (v.status === 'proven') {
+    console.log(`\n✓ determinism proven — base ${v.base}, head ${v.head}`);
+    return false;
+  }
+  if (v.status === 'unknown') {
+    console.log('\n⚠ determinism basis unknown — a capture predates the determinism ledger; recapture to certify it.');
+    return false;
+  }
+  console.log(
+    `\n✗ determinism NOT proven — base ${v.base}, head ${v.head}. An unproven capture can drift, so a clean\n` +
+      '  diff might be two matching NONDETERMINISTIC reads. Enable selfCheck (default) or replay a recorded HAR.',
+  );
+  return true;
+}
+
 const HELP = `${COMMAND} — certify a CSS refactor by diffing two computed-style map captures
 
 usage: ${COMMAND} [baseRef] [options]
@@ -201,12 +270,18 @@ if (args.length <= 1) {
 
 let result;
 let inventoryAudit = null;
+let coverageVerdict = null;
+let determinismVerdict = null;
 try {
   assertCompatibleMapDirs(dirA, dirB);
   result = diffStyleMapDirs(dirA, dirB);
-  // Read inventory here, while the (possibly cached/restored) dirs still exist —
-  // the finally below deletes them in cached-map mode.
+  // Read inventory + the certification ledgers here, while the (possibly cached/restored)
+  // dirs still exist — the finally below deletes them in cached-map mode. Coverage is the
+  // HEAD bundle's completeness basis; determinism needs both sides.
   inventoryAudit = readInventoryAudit(dirA, dirB);
+  const headLedger = readLedger(dirB);
+  coverageVerdict = auditCoverage(capturedSurfaceKeys(dirB), headLedger);
+  determinismVerdict = auditDeterminism(readLedger(dirA), headLedger);
 } catch (e) {
   console.error(e.message);
   process.exit(2);
@@ -243,8 +318,14 @@ for (const sd of surfaces) {
 }
 
 const invRemovals = printInventoryAudit(inventoryAudit);
+const coverageFails = printCoverageVerdict(coverageVerdict);
+const determinismFails = printDeterminismVerdict(determinismVerdict);
 
-if (jsonOut) fs.writeFileSync(jsonOut, JSON.stringify({ counts, surfaces, compared }, null, 2));
+if (jsonOut)
+  fs.writeFileSync(
+    jsonOut,
+    JSON.stringify({ counts, surfaces, compared, coverage: coverageVerdict, determinism: determinismVerdict }, null, 2),
+  );
 
 const total = counts.dom + counts.style + counts.state;
 const newSurfaces = surfaces.filter((s) => s.missing).length;
@@ -252,13 +333,17 @@ const newSurfaces = surfaces.filter((s) => s.missing).length;
 const surfaceCount = surfaces.length;
 const newNote = newSurfaces ? ` (+${newSurfaces} new surface(s) with no baseline)` : '';
 const invNote = invRemovals ? ` + ${invRemovals} unacknowledged inventory removal(s)` : '';
+const covNote = coverageFails ? ` + ${coverageVerdict.uncovered.length} uncaptured registered surface(s)` : '';
+const detNote = determinismFails ? ' + determinism unproven' : '';
+const clean = total === 0 && invRemovals === 0 && !coverageFails && !determinismFails;
 console.log(
-  total === 0 && invRemovals === 0
+  clean
     ? newSurfaces === 0
       ? `\n✓ 0 changed surfaces across ${compared} captured surface(s): every computed style, pseudo-element, and hover/focus/active state matches`
       : `\nℹ ${newSurfaces} new surface(s) captured with no baseline to compare — review before baselining`
-    : `\n✗ ${counts.dom} DOM change(s), ${counts.style} computed-style difference(s), ${counts.state} state-delta difference(s) across ${surfaceCount} surfaces${newNote}${invNote}`,
+    : `\n✗ ${counts.dom} DOM change(s), ${counts.style} computed-style difference(s), ${counts.state} state-delta difference(s) across ${surfaceCount} surfaces${newNote}${invNote}${covNote}${detNote}`,
 );
-// 0 = identical, 1 = reviewable differences (incl. unacknowledged inventory
-// removals), 3 = only new surfaces (no baseline). 2 reserved for usage/capture errors.
-process.exit(total > 0 || invRemovals > 0 ? 1 : newSurfaces > 0 ? 3 : 0);
+// 0 = identical, 1 = reviewable differences (incl. unacknowledged inventory removals, an
+// incomplete coverage registry, or an unproven-determinism capture), 3 = only new surfaces
+// (no baseline). 2 = usage/capture error.
+process.exit(total > 0 || invRemovals > 0 || coverageFails || determinismFails ? 1 : newSurfaces > 0 ? 3 : 0);

package/dist/coverage.d.ts

@@ -34,3 +34,45 @@ export type CoverageGaps = {
  * dir), and it's exported so a consumer can assert coverage however it likes.
  */
 export declare function coverageGaps(capturedKeys: Iterable<string>, expected: Iterable<string>, exclude?: Record<string, string>): CoverageGaps;
+/** Bundled next to the maps, so the completeness basis travels with the capture. */
+export declare const COVERAGE_LEDGER = "styleproof-coverage.json";
+/**
+ * How a capture's determinism was established — the second half of a trustworthy green.
+ * `self-checked`: captured twice and the computed styles matched (a drift would have
+ * failed the capture). `replayed`: rendered against a recorded HAR, so deterministic by
+ * construction. `unproven`: neither — the styles could have drifted and no one checked.
+ */
+export type DeterminismBasis = 'self-checked' | 'replayed' | 'unproven';
+export type CoverageLedger = {
+    version: 1;
+    /** The declared surface registry, or null when the spec asserted none. */
+    expected: string[] | null;
+    /** Reviewed opt-outs (`key → reason`). */
+    exclude: Record<string, string>;
+    /** How this capture's determinism was established (3.10.0). Absent on older bundles. */
+    determinism?: DeterminismBasis;
+};
+export type DeterminismVerdict = {
+    /** `proven` — both sides self-checked or replayed; `unproven` — a side was neither, so
+     *  a clean diff might just be two matching NONDETERMINISTIC captures; `unknown` — an
+     *  older bundle with no determinism field (degrade, don't block). */
+    status: 'proven' | 'unproven' | 'unknown';
+    base: DeterminismBasis | 'unknown';
+    head: DeterminismBasis | 'unknown';
+};
+/** The gate's determinism call: a green needs BOTH sides proven (self-checked or replayed). */
+export declare function auditDeterminism(base: CoverageLedger | null, head: CoverageLedger | null): DeterminismVerdict;
+export type CoverageVerdict = {
+    /** `complete` — every registered surface captured; `incomplete` — a registered
+     *  surface is missing (gates); `unasserted` — no registry, so a green can only
+     *  certify the captured surfaces, not that they are all of them. */
+    basis: 'complete' | 'incomplete' | 'unasserted';
+    /** Size of the declared registry, or null when unasserted. */
+    registrySize: number | null;
+    /** Registered surfaces neither captured nor excluded — the coverage hole. */
+    uncovered: string[];
+    /** `exclude` entries no longer in `expected` — a rotted opt-out. */
+    staleExclusions: string[];
+};
+/** The gate's completeness call: audit what was actually captured against the ledger. */
+export declare function auditCoverage(capturedKeys: Iterable<string>, ledger: CoverageLedger | null): CoverageVerdict;

package/dist/coverage.js

@@ -34,3 +34,37 @@ export function coverageGaps(capturedKeys, expected, exclude = {}) {
     const staleExclusions = Object.keys(exclude).filter((k) => !expectedSet.has(k));
     return { uncovered, staleExclusions };
 }
+// ── coverage provenance (the gate-level completeness assertion) ──────────────────
+// The guard above runs in the app's SUITE. That fails a build when the spec forgets a
+// route, but the GATE (styleproof-diff, reading captured maps) never learns whether
+// coverage was asserted at all — so a green silently implies a completeness it can't
+// back up. The capture writes this ledger into the bundle; the gate reads it and
+// certifies "clean" only against a STATED basis, enforcing the registry against the
+// maps actually captured (a declared surface whose capture FAILED is caught here, where
+// the suite guard — which checks the declared list — cannot see it).
+/** Bundled next to the maps, so the completeness basis travels with the capture. */
+export const COVERAGE_LEDGER = 'styleproof-coverage.json';
+/** The gate's determinism call: a green needs BOTH sides proven (self-checked or replayed). */
+export function auditDeterminism(base, head) {
+    const b = base?.determinism ?? 'unknown';
+    const h = head?.determinism ?? 'unknown';
+    const proven = (d) => d === 'self-checked' || d === 'replayed';
+    if (b === 'unproven' || h === 'unproven')
+        return { status: 'unproven', base: b, head: h };
+    if (proven(b) && proven(h))
+        return { status: 'proven', base: b, head: h };
+    return { status: 'unknown', base: b, head: h };
+}
+/** The gate's completeness call: audit what was actually captured against the ledger. */
+export function auditCoverage(capturedKeys, ledger) {
+    if (!ledger || ledger.expected == null) {
+        return { basis: 'unasserted', registrySize: null, uncovered: [], staleExclusions: [] };
+    }
+    const { uncovered, staleExclusions } = coverageGaps(capturedKeys, ledger.expected, ledger.exclude);
+    return {
+        basis: uncovered.length ? 'incomplete' : 'complete',
+        registrySize: ledger.expected.length,
+        uncovered,
+        staleExclusions,
+    };
+}

package/dist/diff.js

@@ -1,6 +1,7 @@
 import fs from 'node:fs';
 import path from 'node:path';
 import { loadStyleMap, isUnder } from './capture.js';
+import { isMapFile } from './map-store.js';
 function diffProps(propsA, propsB, fallbackA, fallbackB, unsetA, unsetB) {
     const changed = [];
     for (const prop of new Set([...Object.keys(propsA), ...Object.keys(propsB)])) {
@@ -173,7 +174,7 @@ function tallyCounts(findings, counts) {
 function indexDir(dir) {
     return Object.fromEntries(fs
         .readdirSync(dir)
-        .filter((f) => f !== 'styleproof-manifest.json' && /\.json(\.gz)?$/.test(f))
+        .filter(isMapFile)
         .map((f) => [f.replace(/\.json(\.gz)?$/, ''), path.join(dir, f)]));
 }
 /** Diff every same-named capture between two directories. `volatile` is the

package/dist/map-store.d.ts

@@ -3,6 +3,12 @@ export declare const DEFAULT_MAP_LABEL = "current";
 export declare const DEFAULT_MAP_STORE_BRANCH = "styleproof-maps";
 export declare const DEFAULT_REMOTE = "origin";
 export declare const MAP_MANIFEST = "styleproof-manifest.json";
+/** Bundle files that sit alongside the maps but are NOT surfaces (manifest, coverage
+ *  ledger, and any future sidecar). Every place that enumerates surface maps must skip
+ *  these, or a sidecar reads as a phantom "new surface". */
+export declare const RESERVED_BUNDLE_FILES: ReadonlySet<string>;
+/** True for a captured surface map (`<key>@<width>.json[.gz]`), false for metadata. */
+export declare function isMapFile(name: string): boolean;
 export declare class MapStoreError extends Error {
 }
 export interface MapManifest {

package/dist/map-store.js

@@ -5,12 +5,21 @@ import { createRequire } from 'node:module';
 import os from 'node:os';
 import path from 'node:path';
 import { inferBaseRef } from './gitref.js';
+import { COVERAGE_LEDGER } from './coverage.js';
 export const DEFAULT_MAP_DIR = '.styleproof/maps';
 export const DEFAULT_MAP_LABEL = 'current';
 export const DEFAULT_MAP_STORE_BRANCH = 'styleproof-maps';
 export const DEFAULT_REMOTE = 'origin';
 export const MAP_MANIFEST = 'styleproof-manifest.json';
 const GENERATED_DIRTY_ALLOWLIST = new Set(['next-env.d.ts']);
+/** Bundle files that sit alongside the maps but are NOT surfaces (manifest, coverage
+ *  ledger, and any future sidecar). Every place that enumerates surface maps must skip
+ *  these, or a sidecar reads as a phantom "new surface". */
+export const RESERVED_BUNDLE_FILES = new Set([MAP_MANIFEST, COVERAGE_LEDGER]);
+/** True for a captured surface map (`<key>@<width>.json[.gz]`), false for metadata. */
+export function isMapFile(name) {
+    return !RESERVED_BUNDLE_FILES.has(name) && /\.json(\.gz)?$/.test(name);
+}
 export class MapStoreError extends Error {
 }
 function runGit(cwd, args, maxBuffer = 1 << 28) {

package/dist/report.js

@@ -2,6 +2,7 @@ import fs from 'node:fs';
 import path from 'node:path';
 import { PNG } from 'pngjs';
 import { loadStyleMap } from './capture.js';
+import { isMapFile } from './map-store.js';
 import { diffStyleMapDirs, diffContentDirs, } from './diff.js';
 import { describeChange, tokenIndex, toHex, trackCount } from './describe.js';
 // Re-export the plain-English summariser so consumers (and tests) reach it
@@ -443,9 +444,7 @@ function liveCandidateLabel(candidate) {
     return `${label} (${candidate.reason})`;
 }
 function captureFiles(dir) {
-    return fs.existsSync(dir)
-        ? fs.readdirSync(dir).filter((f) => f !== 'styleproof-manifest.json' && /\.json(\.gz)?$/.test(f))
-        : [];
+    return fs.existsSync(dir) ? fs.readdirSync(dir).filter(isMapFile) : [];
 }
 function collectLiveCandidateLabels(beforeDir, afterDir) {
     const seen = new Set();

package/dist/runner.d.ts

@@ -222,19 +222,6 @@ export declare function resolveBaseDir(baseDir: string | undefined, env?: string
 export declare function resolveScreenshots(screenshots: boolean | undefined, env?: string | undefined): boolean;
 /** The capture settings every capturer shares (everything bar the surface set). */
 type CaptureConfig = Omit<DefineOptions, 'surfaces' | 'expected' | 'exclude'>;
-/**
- * Generate one Playwright test per surface × width that captures the style
- * map to `<baseDir>/<dir>/<key>@<width>.json.gz`. Captures are made
- * deterministic with no per-repo fixtures: the baseline run records each
- * surface's data responses to a HAR, and the comparison run replays them (set
- * STYLEPROOF_REPLAY_FROM=<baseline dir> on the comparison capture), while the
- * clock is frozen so time-derived styling is stable.
- *
- * ```ts
- * // styleproof.spec.ts
- * defineStyleMapCapture({ surfaces: SURFACES, dir: process.env.STYLEMAP_DIR });
- * ```
- */
 export declare function defineStyleMapCapture(options: DefineOptions): void;
 /** Options for {@link defineCrawlCapture}: where to crawl, how to filter/key the
  *  links, and the viewport sweep — plus the shared capture settings. */

package/dist/runner.js

@@ -3,7 +3,7 @@ import fs from 'node:fs';
 import path from 'node:path';
 import { captureStyleMap, saveStyleMap, trackInflightRequests, } from './capture.js';
 import { diffStyleMaps } from './diff.js';
-import { coverageGaps } from './coverage.js';
+import { coverageGaps, COVERAGE_LEDGER } from './coverage.js';
 import { detectViewportWidths } from './breakpoints.js';
 import { selectCrawlLinks } from './crawl.js';
 /** One-line description of the first drift finding, for the self-check error. */
@@ -452,6 +452,27 @@ function resolveSettings(c) {
  * defineStyleMapCapture({ surfaces: SURFACES, dir: process.env.STYLEMAP_DIR });
  * ```
  */
+/**
+ * Emit a test that records the coverage ledger into the capture bundle, so the GATE
+ * (styleproof-diff) can state its completeness basis — not just the app's own suite.
+ * `expected: null` records that the spec declared no registry, so a green can only
+ * certify the captured surfaces. Runs on a capture run (dir set) only.
+ */
+function writeCoverageLedgerTest(settings, dir, expected, exclude) {
+    test('styleproof coverage ledger', () => {
+        const outDir = path.join(settings.baseDir, dir);
+        fs.mkdirSync(outDir, { recursive: true });
+        // Determinism basis: self-check ON proves it (a drift would have failed the capture);
+        // else a replay run is deterministic by construction; else it's unproven.
+        const determinism = settings.selfCheck
+            ? 'self-checked'
+            : settings.replayFrom
+                ? 'replayed'
+                : 'unproven';
+        const ledger = { version: 1, expected, exclude, determinism };
+        fs.writeFileSync(path.join(outDir, COVERAGE_LEDGER), JSON.stringify(ledger, null, 2));
+    });
+}
 export function defineStyleMapCapture(options) {
     const { surfaces, expected, exclude = {}, dir } = options;
     const settings = resolveSettings(options);
@@ -475,6 +496,8 @@ export function defineStyleMapCapture(options) {
     }
     test.describe('styleproof capture', () => {
         test.skip(!dir, 'set STYLEMAP_DIR=<label> to capture computed-style maps');
+        if (dir)
+            writeCoverageLedgerTest(settings, dir, expected ?? null, exclude);
         for (const surface of captureSurfaces) {
             if (surface.widths && surface.widths.length > 0) {
                 // Explicit widths: one parallelizable test per surface × width.
@@ -525,6 +548,11 @@ export function defineCrawlCapture(options) {
     // that styleproof-map uses to select capture tests picks up crawl specs too.
     test.describe('styleproof capture (crawl)', () => {
         test.skip(!dir, 'set STYLEMAP_DIR=<label> to capture computed-style maps');
+        // A crawl DISCOVERS its surfaces from the nav — it has no registry to check against,
+        // so it records `expected: null` (completeness honestly "not asserted": the crawl
+        // captures what the nav links to, and can't prove that's every route in the app).
+        if (dir)
+            writeCoverageLedgerTest(settings, dir, null, {});
         test('discover surfaces by crawling links, then capture each', async ({ page }) => {
             // 1. Load the root and wait for its nav links to hydrate — an SPA renders them
             //    client-side, so they aren't in the initial HTML.

package/package.json

@@ -1,6 +1,6 @@
 {
   "name": "styleproof",
-  "version": "3.8.0",
+  "version": "3.10.0",
   "description": "Catch every CSS change before it ships — review PRs and certify refactors by the browser's computed styles, not pixels. Works with any styling system.",
   "keywords": [
     "playwright",