rungate 0.6.0 → 0.7.0

This diff represents the content of publicly available package versions that have been released to one of the supported registries. The information contained in this diff is provided for informational purposes only and reflects changes between package versions as they appear in their respective public registries.
Files changed (113) hide show
  1. package/lib/affected-tests.ts +398 -0
  2. package/lib/agent-audit.ts +2 -3
  3. package/lib/agent-worktree.ts +287 -0
  4. package/lib/agents-md-sections.ts +81 -16
  5. package/lib/behavioral-cache.ts +150 -2
  6. package/lib/branch-protection.ts +790 -0
  7. package/lib/brief-ac-alignment.ts +104 -0
  8. package/lib/brief-references.ts +205 -0
  9. package/lib/bun-version.ts +93 -14
  10. package/lib/check-remedy.ts +140 -0
  11. package/lib/ci-checkout.ts +17 -0
  12. package/lib/ci-history-depth.ts +159 -0
  13. package/lib/ci-setup-step.ts +96 -0
  14. package/lib/claude-code-action.ts +85 -0
  15. package/lib/claude-tool-vocabulary.ts +188 -0
  16. package/lib/cli-commands.ts +15 -0
  17. package/lib/commit-sha.ts +42 -0
  18. package/lib/commit-subject.ts +90 -0
  19. package/lib/compliance-report.ts +377 -45
  20. package/lib/compliance.ts +191 -17
  21. package/lib/config-loader.ts +39 -1
  22. package/lib/config-trackable.ts +391 -0
  23. package/lib/conformity.ts +1087 -100
  24. package/lib/containment.ts +238 -0
  25. package/lib/content-hash.ts +24 -0
  26. package/lib/create-sc.ts +249 -16
  27. package/lib/dev-liveness.ts +180 -0
  28. package/lib/discovery-branch-points.ts +432 -0
  29. package/lib/discovery-stability.ts +481 -0
  30. package/lib/distribution-currency.ts +186 -0
  31. package/lib/doc-routing.ts +157 -0
  32. package/lib/evidence-prevalidator.ts +224 -2
  33. package/lib/evidence-satisfiability.ts +534 -12
  34. package/lib/failure-intake.ts +317 -0
  35. package/lib/fallow-availability.ts +157 -0
  36. package/lib/fs-write-capabilities.ts +232 -0
  37. package/lib/gate-run-split.ts +147 -0
  38. package/lib/generators/agents-md.ts +140 -32
  39. package/lib/git-isolation.ts +140 -0
  40. package/lib/github.ts +15 -0
  41. package/lib/google-auth-action.ts +93 -0
  42. package/lib/harness-project.ts +48 -0
  43. package/lib/harness-provenance.ts +255 -0
  44. package/lib/hook-evidence.ts +235 -0
  45. package/lib/hook-scopes.ts +386 -0
  46. package/lib/hook-target-tree.ts +398 -0
  47. package/lib/hook-trees.ts +447 -0
  48. package/lib/js-top-level-code.ts +78 -10
  49. package/lib/junit-checks.ts +52 -0
  50. package/lib/launch-base.ts +253 -0
  51. package/lib/line-citation.ts +372 -0
  52. package/lib/orphan-reaper.ts +468 -0
  53. package/lib/parity-binding.ts +205 -0
  54. package/lib/per-test-timeouts.ts +219 -0
  55. package/lib/plan-coverage.ts +151 -0
  56. package/lib/prior-branch.ts +201 -8
  57. package/lib/process-table.ts +146 -0
  58. package/lib/project-state-derivation.ts +219 -0
  59. package/lib/project-state-root.ts +258 -0
  60. package/lib/published-artifact.ts +525 -0
  61. package/lib/readdir-sorted.ts +45 -0
  62. package/lib/review-execution-reading.ts +206 -0
  63. package/lib/review-tool-agreement.ts +433 -0
  64. package/lib/rule-promotion.ts +234 -0
  65. package/lib/rule-retirement.ts +525 -0
  66. package/lib/run-branch.ts +612 -0
  67. package/lib/rungate-schema.ts +19 -0
  68. package/lib/runner-home.ts +274 -0
  69. package/lib/sc-id-cross-tree.ts +19 -7
  70. package/lib/sc-ids.ts +121 -13
  71. package/lib/sc-reservations.ts +289 -0
  72. package/lib/scaffold/ai-review-findings.ts +403 -0
  73. package/lib/scaffold/ai-review-step.ts +4661 -64
  74. package/lib/scaffold/managed-workflow.ts +587 -192
  75. package/lib/scaffold/owned-files.ts +152 -45
  76. package/lib/scaffold/pre-push-hook.ts +49 -5
  77. package/lib/scaffold/steps.ts +1413 -244
  78. package/lib/scaffold/workflow-census.ts +217 -0
  79. package/lib/scaffold/workflow-drift.ts +217 -0
  80. package/lib/scaffold/write-steps.ts +4 -0
  81. package/lib/scan-budget.ts +247 -0
  82. package/lib/scanner.ts +43 -51
  83. package/lib/scratch-roots.ts +50 -5
  84. package/lib/screenshot-discovery.ts +121 -0
  85. package/lib/screenshot-upload.ts +41 -1
  86. package/lib/secret-patterns.ts +511 -8
  87. package/lib/ship-mode.ts +219 -0
  88. package/lib/source-slice.ts +294 -0
  89. package/lib/stale-cleanup.ts +37 -3
  90. package/lib/suite-flake.ts +278 -0
  91. package/lib/suite-measurement.ts +23 -4
  92. package/lib/suite-run-reading.ts +241 -1
  93. package/lib/suite-timeout-margin.ts +193 -0
  94. package/lib/test-case-bodies.ts +883 -0
  95. package/lib/test-case-deletion.ts +653 -0
  96. package/lib/test-census.ts +97 -29
  97. package/lib/test-command-positions.ts +274 -0
  98. package/lib/test-suite-lock.ts +159 -16
  99. package/lib/tool-probe.ts +31 -9
  100. package/lib/top-level-cd.ts +393 -0
  101. package/lib/tree-entries.ts +89 -0
  102. package/lib/ts-ast.ts +200 -0
  103. package/lib/walk-test-files.ts +97 -0
  104. package/lib/workflow-actions.ts +263 -0
  105. package/lib/worktree-collect.ts +177 -5
  106. package/lib/worktree-repo.ts +36 -1
  107. package/package.json +2 -1
  108. package/scripts/ai-review-findings.ts +45 -0
  109. package/scripts/create-sc.ts +21 -1
  110. package/scripts/extract-constraints.ts +107 -7
  111. package/scripts/generate-code-map.ts +23 -30
  112. package/scripts/scaffold-project.ts +8 -1
  113. package/scripts/sync-sc-status.ts +77 -16
@@ -15,12 +15,14 @@
15
15
  */
16
16
 
17
17
  import { existsSync, readFileSync, writeFileSync, readdirSync } from "fs";
18
+
19
+ import { walkTestFiles } from "../lib/walk-test-files";
18
20
  import { commentMask, enclosingComment, stripJsComments } from "../lib/strip-js-comments";
19
21
  import { commentClaimForm, titleCall, type ClaimForm } from "../lib/sc-claim-forms";
20
22
  import { join } from "path";
21
23
  import { execSync } from "child_process";
22
24
  import { matchPattern, isBehavioralSC, type ParsedSC } from "../lib/conformity";
23
- import { readFreshCache } from "../lib/behavioral-cache";
25
+ import { behavioralCachePath, readFreshCache } from "../lib/behavioral-cache";
24
26
  import { detectDrift } from "./detect-sc-drift";
25
27
  import {
26
28
  ambiguityOf,
@@ -32,6 +34,7 @@ import {
32
34
  requiresQualification,
33
35
  scDeclarationPattern,
34
36
  scMentionPattern as scIdMentionPattern,
37
+ unenumeratedScRegistry,
35
38
  type ScRegistry,
36
39
  } from "../lib/sc-ids";
37
40
 
@@ -55,13 +58,28 @@ export interface UncheckedSC {
55
58
  * Built from the same spec scan `findUncheckedSCs` walks, so the ids a test
56
59
  * can be credited for and the ids a spec declares are the same set by
57
60
  * construction rather than by two regexes agreeing.
61
+ *
62
+ * ## An absent `specs/` is UNENUMERATED, not "declares nothing" (#606)
63
+ *
64
+ * This is the one producer feeding `allocateScNumber` in the live CLI
65
+ * (`scripts/create-sc.ts`), and it returned `buildScRegistry([])` for a tree
66
+ * with no `specs/` at all. That registry says "this tree declares nothing",
67
+ * which the allocator then reported as a consulted width contributing zero
68
+ * ids. A fix on the reading side alone cannot see this: the registry leg
69
+ * would faithfully report an enumeration flag nobody had ever set wrong.
58
70
  */
59
71
  export function readScRegistry(specsDir?: string): ScRegistry {
60
72
  const dir = specsDir || SPECS_DIR;
61
- if (!existsSync(dir)) return buildScRegistry([]);
73
+ if (!existsSync(dir)) return unenumeratedScRegistry(`${dir} does not exist`);
62
74
  const specs: Array<{ file: string; content: string }> = [];
63
- for (const file of new Bun.Glob("**/*.md").scanSync({ cwd: dir, absolute: false })) {
64
- specs.push({ file, content: readFileSync(join(dir, file), "utf-8") });
75
+ try {
76
+ for (const file of new Bun.Glob("**/*.md").scanSync({ cwd: dir, absolute: false })) {
77
+ specs.push({ file, content: readFileSync(join(dir, file), "utf-8") });
78
+ }
79
+ } catch (err) {
80
+ // A directory that exists and cannot be listed or read. Same fact as an
81
+ // absent one from the allocator's side, and a different reason.
82
+ return unenumeratedScRegistry(`${dir} could not be enumerated: ${(err as Error).message}`);
65
83
  }
66
84
  return buildScRegistry(specs);
67
85
  }
@@ -313,6 +331,9 @@ export function findTestFilesForSCs(
313
331
  const dir = testDir || TEST_DIR;
314
332
  const out = new Map<string, ScCredit[]>();
315
333
  if (!existsSync(dir)) return out;
334
+ // Tree-wide, and taken before any file is read: whether a criterion has a
335
+ // running test anywhere is not a question a single file can answer (#810).
336
+ const intended = new Set(intendedOnly(dir).keys());
316
337
 
317
338
  const wanted = new Map<string, string[]>();
318
339
  for (const sc of scs) {
@@ -321,7 +342,7 @@ export function findTestFilesForSCs(
321
342
  wanted.set(sc.id, specs);
322
343
  }
323
344
 
324
- for (const file of readdirSync(dir).filter(f => f.endsWith(".test.ts"))) {
345
+ for (const file of walkTestFiles(dir)) {
325
346
  const source = readFileSync(join(dir, file), "utf-8");
326
347
  const credits: ScCredit[] = [];
327
348
  for (const { id, at, kind, form } of coverageClaimSites(source)) {
@@ -334,6 +355,21 @@ export function findTestFilesForSCs(
334
355
  // crediting, and a todo is the clearest case there is: nothing can go
335
356
  // red, so nothing can be measured.
336
357
  if (kind === "unrun-title") continue;
358
+ // #810. That refusal was per-POSITION, so it dropped the todo's own
359
+ // claim and then accepted a comment in ANOTHER file for the same id —
360
+ // two statements about one criterion, and the weaker one decided. A
361
+ // todo is a statement about execution; a comment is not. `intendedOnly`
362
+ // is the one binding that answers it, so there is one thing to mutate.
363
+ //
364
+ // Deliberately NOT `kind === "comment" && intended.has(id)`. The first
365
+ // draft was, it reads as the narrower and therefore safer form, and it
366
+ // is unreachable: `intendedOnly` excludes any id with a running title,
367
+ // so the only kinds that can reach here for an id it names are the ones
368
+ // the conjunct would have admitted anyway. Measured — the battery row
369
+ // for widening it to this was 0 of 10, which is the fourth shape in
370
+ // `.claude/rules/checks-must-be-able-to-fail.md` and the reason AC-2
371
+ // had no witness until the conjunct came out.
372
+ if (intended.has(id)) continue;
337
373
  for (const specFile of wanted.get(id) ?? []) {
338
374
  if (requiresQualification(id, reg) && !(kind === "title" && isQualifiedBy(source, at, specFile))) continue;
339
375
  if (!credits.some(c => c.id === id && c.specFile === specFile)) {
@@ -350,24 +386,48 @@ export function findTestFilesForSCs(
350
386
  * Ids whose only declaring position in the test tree is a title that never
351
387
  * runs (#574), with the files that name them.
352
388
  *
353
- * An id that ALSO has a running title or a declaring comment anywhere is
354
- * not here: the question is whether anything can measure it, not whether a
355
- * todo exists. That distinction is the whole report — otherwise every
356
- * criterion with a todo beside a real test would be listed as unmeasured
357
- * and the list would be ignored within a week.
389
+ * An id that ALSO has a running title anywhere is not here: the question is
390
+ * whether anything can measure it, not whether a todo exists. That
391
+ * distinction is the whole report — otherwise every criterion with a todo
392
+ * beside a real test would be listed as unmeasured and the list would be
393
+ * ignored within a week.
394
+ *
395
+ * ## A declaring comment used to count as measurement, and it does not (#810)
396
+ *
397
+ * It was the natural reading and it is wrong in one direction only, which is
398
+ * why nobody noticed: a comment is where this project DECLARES coverage, so a
399
+ * declaring comment plus a todo read as "somebody wrote the test and left a
400
+ * stale todo beside it". The population it actually describes is the
401
+ * opposite — a file about something else quoting the criterion as fixture
402
+ * input, in a docstring, on its own line, which is a declaring form.
403
+ *
404
+ * Measured on this tree at the time of #810: **five** ids were in that state
405
+ * and **zero** of them had a running title. Four were `SC-1930`, `SC-1940`,
406
+ * `SC-1943` and `SC-1944`, cited by `audit-specs-fallback-scope.test.ts` and
407
+ * `audit-specs-rewrite-fidelity.test.ts` — both tests of the spec-rewriting
408
+ * tool, neither a test of a rule-lifecycle criterion. `sync-sc-status.ts`
409
+ * ticked all four, and the tick then broke the citing file's own premise,
410
+ * which asserts the spec still reads `- [ ] SC-1940: …`.
411
+ *
412
+ * So the comparison this makes is between a statement about EXECUTION and a
413
+ * statement about intent. A comment is neither, and crediting it here made
414
+ * `SC-1957`'s invariant — *"an id is reported unmeasured AND credited by a
415
+ * test claim"* — true by construction: an id with a declaring comment could
416
+ * not enter this report, so the two readings could not disagree. Narrowing
417
+ * `measured` is what gives that case something to see.
358
418
  */
359
419
  export function intendedOnly(testDir: string): Map<string, string[]> {
360
420
  const out = new Map<string, string[]>();
361
421
  if (!existsSync(testDir)) return out;
362
422
  const measured = new Set<string>();
363
423
  const intended = new Map<string, string[]>();
364
- for (const file of readdirSync(testDir).filter(f => f.endsWith(".test.ts"))) {
424
+ for (const file of walkTestFiles(testDir)) {
365
425
  for (const site of coverageClaimSites(readFileSync(join(testDir, file), "utf-8"))) {
366
426
  if (site.kind === "unrun-title") {
367
427
  const files = intended.get(site.id) ?? [];
368
428
  if (!files.includes(file)) files.push(file);
369
429
  intended.set(site.id, files);
370
- } else if (site.kind === "title" || (site.kind === "comment" && site.form)) {
430
+ } else if (site.kind === "title") {
371
431
  measured.add(site.id);
372
432
  }
373
433
  }
@@ -508,9 +568,10 @@ export function checkConformitySCs(uncheckedSCs: UncheckedSC[], root: string): C
508
568
  const failing = new Set<string>();
509
569
  const unmatchable = new Set<string>();
510
570
 
511
- // Read behavioral cache for behavioral SCs
512
- const cachePath = join(root, ".rungate", "behavioral-results.json");
513
- const behavioralCache = readFreshCache(cachePath);
571
+ // Read behavioral cache for behavioral SCs. Same binding the conformity
572
+ // suite asserts through, so the tick-decider and the gate cannot disagree
573
+ // about where the evidence lives (#328).
574
+ const behavioralCache = readFreshCache(behavioralCachePath(root));
514
575
 
515
576
  for (const sc of uncheckedSCs) {
516
577
  const statement = sc.statement || "";
@@ -661,7 +722,7 @@ if (isDirectExecution) {
661
722
  // Mentions that look like claims and are not credited. Visible on purpose:
662
723
  // under-crediting is the safe direction and still a measurement error.
663
724
  const unqualified = new Map<string, string>();
664
- for (const file of readdirSync(TEST_DIR).filter(f => f.endsWith(".test.ts"))) {
725
+ for (const file of walkTestFiles(TEST_DIR)) {
665
726
  for (const u of unqualifiedClaims(readFileSync(join(TEST_DIR, file), "utf-8"), registry)) {
666
727
  unqualified.set(`${file} ${u.id}`, ` ${file}: ${u.id} — ${u.reason}`);
667
728
  }