@holmes-lab/holmes-kit 0.40.0 → 0.40.1

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 (56) hide show
  1. package/CHANGELOG.md +234 -0
  2. package/README.md +70 -4
  3. package/dist/.build-id +1 -1
  4. package/dist/holmes/cli/index.js +25 -19
  5. package/dist/holmes/cli/interactive-prompt.d.ts +39 -2
  6. package/dist/holmes/cli/interactive-prompt.js +98 -16
  7. package/dist/holmes/config/config.d.ts +17 -0
  8. package/dist/holmes/config/config.js +13 -1
  9. package/dist/holmes/governance/autonomy.d.ts +7 -0
  10. package/dist/holmes/governance/autonomy.js +36 -5
  11. package/dist/holmes/governance/push-gate-runner.d.ts +2 -0
  12. package/dist/holmes/governance/push-gate-runner.js +40 -3
  13. package/dist/holmes/governance/push-gate.d.ts +13 -2
  14. package/dist/holmes/governance/push-gate.js +106 -14
  15. package/dist/holmes/guardrail/advisory-line.d.ts +102 -0
  16. package/dist/holmes/guardrail/advisory-line.js +718 -0
  17. package/dist/holmes/guardrail/impact-gate.d.ts +6 -0
  18. package/dist/holmes/guardrail/impact-gate.js +32 -5
  19. package/dist/holmes/hooks/pre-tool-use.d.ts +11 -0
  20. package/dist/holmes/hooks/pre-tool-use.js +44 -5
  21. package/dist/holmes/hooks/session-start.js +17 -1
  22. package/dist/holmes/mcp/handlers/maintenance-analysis.d.ts +13 -0
  23. package/dist/holmes/mcp/handlers/maintenance-analysis.js +59 -2
  24. package/dist/holmes/mcp/handlers/test-execution.d.ts +470 -3
  25. package/dist/holmes/mcp/handlers/test-execution.js +199 -21
  26. package/dist/holmes/mcp/handlers.d.ts +468 -3
  27. package/dist/holmes/mcp/handlers.js +5 -1
  28. package/dist/holmes/mcp/maintenance-analyze.d.ts +5 -0
  29. package/dist/holmes/mcp/maintenance-analyze.js +4 -2
  30. package/dist/holmes/mcp/maintenance-evidence.d.ts +9 -0
  31. package/dist/holmes/mcp/maintenance-evidence.js +4 -1
  32. package/dist/holmes/mcp/tool-schemas.js +13 -7
  33. package/dist/holmes/project/baseline.d.ts +15 -2
  34. package/dist/holmes/project/baseline.js +11 -6
  35. package/dist/holmes/project/bounded-io.d.ts +1 -0
  36. package/dist/holmes/project/bounded-io.js +4 -2
  37. package/dist/holmes/project/change-source.d.ts +4 -0
  38. package/dist/holmes/project/change-source.js +13 -5
  39. package/dist/holmes/project/root.js +3 -1
  40. package/dist/holmes/review/evidence-mode.d.ts +21 -0
  41. package/dist/holmes/review/evidence-mode.js +33 -0
  42. package/dist/holmes/review/head-worktree.d.ts +55 -0
  43. package/dist/holmes/review/head-worktree.js +329 -0
  44. package/dist/holmes/review/mrp.js +13 -3
  45. package/dist/holmes/review/run-provenance.d.ts +45 -0
  46. package/dist/holmes/review/run-provenance.js +314 -0
  47. package/dist/holmes/review/test-evidence.d.ts +174 -0
  48. package/dist/holmes/review/test-evidence.js +330 -2
  49. package/dist/holmes/review/test-reports.d.ts +94 -0
  50. package/dist/holmes/review/test-reports.js +382 -0
  51. package/dist/holmes/review/test-runner.d.ts +176 -65
  52. package/dist/holmes/review/test-runner.js +531 -316
  53. package/dist/holmes/review/test-units.d.ts +159 -0
  54. package/dist/holmes/review/test-units.js +814 -0
  55. package/package.json +1 -1
  56. package/playbooks/publish/PLAYBOOK.md +2 -1
@@ -33,18 +33,15 @@ var __importStar = (this && this.__importStar) || (function () {
33
33
  };
34
34
  })();
35
35
  Object.defineProperty(exports, "__esModule", { value: true });
36
- exports.TAIL_MAX_CHARS = exports.RUNNER_TIMEOUT_MS = void 0;
36
+ exports.TAIL_MAX_CHARS = exports.UNSUPPORTED = exports.readJestReport = exports.wellShapedJson = exports.parseJUnitXmlFiles = exports.parseJUnitXmlOutcomes = exports.parseJUnitXmlCounts = exports.classifyJestOutcomes = exports.parseExecutedCounts = exports.goPackagesWithTestEvents = exports.parseGoTestJson = exports.projectFiles = exports.pytestRules = exports.BUILD_ROOTS = exports.normaliseExcludeDirs = exports.excludedDir = exports.scanEcosystems = exports.RUNNER_TIMEOUT_MS = void 0;
37
37
  exports.runBounded = runBounded;
38
38
  exports.runnerStopNote = runnerStopNote;
39
39
  exports.planTestRun = planTestRun;
40
40
  exports.ecosystemOf = ecosystemOf;
41
- exports.parseGoTestJson = parseGoTestJson;
42
- exports.parseExecutedCounts = parseExecutedCounts;
43
- exports.classifyJestOutcomes = classifyJestOutcomes;
44
- exports.parseJUnitXmlCounts = parseJUnitXmlCounts;
45
41
  exports.jestEntry = jestEntry;
42
+ exports.testRunTimeoutMinutes = testRunTimeoutMinutes;
43
+ exports.testRunWorkers = testRunWorkers;
46
44
  exports.runJestOutcomes = runJestOutcomes;
47
- exports.parseJUnitXmlOutcomes = parseJUnitXmlOutcomes;
48
45
  exports.detectProjectEcosystems = detectProjectEcosystems;
49
46
  exports.runPytest = runPytest;
50
47
  exports.parseCargoTest = parseCargoTest;
@@ -53,12 +50,16 @@ exports.runGradle = runGradle;
53
50
  exports.runDotnet = runDotnet;
54
51
  exports.runGo = runGo;
55
52
  exports.runTestScope = runTestScope;
53
+ exports.runEcosystems = runEcosystems;
54
+ exports.coverageAfterRun = coverageAfterRun;
55
+ exports.environmentReason = environmentReason;
56
56
  exports.tailOf = tailOf;
57
57
  exports.failedTestSummary = failedTestSummary;
58
+ exports.redJestTail = redJestTail;
58
59
  exports.summarizeJestJson = summarizeJestJson;
59
- // @implements A-SPEC-102.1
60
+ // @implements A-SPEC-102.1, A-SPEC-128.4, A-SPEC-128.5, A-SPEC-128.6, A-SPEC-128.7, A-SPEC-128.8, A-SPEC-128.9, A-SPEC-128.10, A-SPEC-128.11, A-SPEC-128.12, A-SPEC-128.13
60
61
  // @implements A-SPEC-647
61
- // @implements A-SPEC-734
62
+ // @implements A-SPEC-734, A-SPEC-466.1, A-SPEC-466.2, A-SPEC-466.3, A-SPEC-466.4, A-SPEC-466.5, A-SPEC-466.6
62
63
  const node_child_process_1 = require("node:child_process");
63
64
  const bounded_group_1 = require("../project/bounded-group");
64
65
  const fs = __importStar(require("node:fs"));
@@ -68,6 +69,10 @@ const path = __importStar(require("node:path"));
68
69
  // `--version` probes above stay as they are: they do not judge anything.
69
70
  const root_1 = require("../project/root");
70
71
  const python_env_1 = require("./python-env");
72
+ const test_units_1 = require("./test-units");
73
+ const test_reports_1 = require("./test-reports");
74
+ const test_units_2 = require("./test-units");
75
+ const run_provenance_1 = require("./run-provenance");
71
76
  // Runner stdout ceiling. Measured 2026-09-01 on this repo: a full `jest --json` for 289 suites /
72
77
  // 4568 tests emits >2MB, and execFileSync's DEFAULT maxBuffer (1MB) kills the child with ENOBUFS —
73
78
  // so a GREEN suite reported `passed: false`, twice, with no failing test named (the baseline never
@@ -90,7 +95,9 @@ let STOPPED = null;
90
95
  * no groups; there the child alone is stopped, as before. Otherwise execFileSync's contract: stdout on exit 0, else a
91
96
  * throw carrying stdout, stderr, status and code.
92
97
  */
93
- function runBounded(cmd, args, opts, timeoutMs = exports.RUNNER_TIMEOUT_MS) {
98
+ /** @implements A-SPEC-466.6 — the bound of the run in progress; `runTestScope` sets it from timeoutMinutes. */
99
+ let RUN_BUDGET_MS = exports.RUNNER_TIMEOUT_MS;
100
+ function runBounded(cmd, args, opts, timeoutMs = RUN_BUDGET_MS) {
94
101
  // @implements A-SPEC-734.4 — the group's whole lifetime lives in bounded-group: a watch shim ends it when this process
95
102
  // goes, and the call ends it on the bound, an overflow or a signal, or when members outlive a normal exit
96
103
  const { result: r, stopped, leftovers } = (0, bounded_group_1.boundedGroupSpawnSync)(cmd, args, opts, timeoutMs);
@@ -123,6 +130,13 @@ function planTestRun(scope) {
123
130
  reason: `${scope.tier} scope — run ${scope.impactedTestFiles.length} impacted test file(s)`,
124
131
  };
125
132
  }
133
+ var test_units_3 = require("./test-units");
134
+ Object.defineProperty(exports, "scanEcosystems", { enumerable: true, get: function () { return test_units_3.scanEcosystems; } });
135
+ Object.defineProperty(exports, "excludedDir", { enumerable: true, get: function () { return test_units_3.excludedDir; } });
136
+ Object.defineProperty(exports, "normaliseExcludeDirs", { enumerable: true, get: function () { return test_units_3.normaliseExcludeDirs; } });
137
+ Object.defineProperty(exports, "BUILD_ROOTS", { enumerable: true, get: function () { return test_units_3.BUILD_ROOTS; } });
138
+ Object.defineProperty(exports, "pytestRules", { enumerable: true, get: function () { return test_units_3.pytestRules; } });
139
+ Object.defineProperty(exports, "projectFiles", { enumerable: true, get: function () { return test_units_3.projectFiles; } });
126
140
  /** Ecosystem of a test file, by extension/convention. `null` when no runner adapter exists for it. */
127
141
  function ecosystemOf(file) {
128
142
  if (/\.(ts|mts|cts|tsx|jsx|js|mjs|cjs)$/i.test(file))
@@ -150,150 +164,18 @@ function ecosystemOf(file) {
150
164
  return 'dotnet';
151
165
  return null; // cpp and non-test files — counted syntactically, not yet executable here
152
166
  }
153
- /**
154
- * @implements A-SPEC-137.1
155
- * Executed-test counts per Go package from `go test -json` output. A test is EXECUTED when a NAMED
156
- * test (`Test` present) reaches `Action: 'pass'` or `'fail'` — a skip does not count (matching the
157
- * jest/pytest rule), and a package-level result (no `Test`) is not a test case. Non-JSON lines (go
158
- * interleaves build output) are ignored. Pure; empty/garbage input yields `{}` rather than throwing.
159
- */
160
- function parseGoTestJson(out) {
161
- const counts = {};
162
- for (const line of out.split('\n')) {
163
- const s = line.trim();
164
- if (!s.startsWith('{'))
165
- continue;
166
- let evt;
167
- try {
168
- evt = JSON.parse(s);
169
- }
170
- catch {
171
- continue;
172
- }
173
- if (!evt.Test || !evt.Package)
174
- continue; // package-level result is not a test case
175
- if (evt.Action === 'pass' || evt.Action === 'fail') { // skip is excluded; pass/fail both executed
176
- counts[evt.Package] = (counts[evt.Package] ?? 0) + 1;
177
- }
178
- }
179
- return counts;
180
- }
181
- /**
182
- * Parse jest --json output into per-file EXECUTED case counts. This is the execution-proof upgrade
183
- * over the syntactic countTestCases lower bound: a case only appears here if jest actually ran it
184
- * (pending/skipped assertions are excluded), so `if(false){it(…)}`, aliasing, and other syntactic
185
- * fabrications cannot produce evidence. Returns {} for unparseable output (caller falls back).
186
- */
187
- function parseExecutedCounts(stdout, cwd) {
188
- const start = stdout.indexOf('{');
189
- if (start < 0)
190
- return {};
191
- let data;
192
- try {
193
- data = JSON.parse(stdout.slice(start));
194
- }
195
- catch {
196
- return {};
197
- }
198
- const out = {};
199
- for (const tr of data?.testResults ?? []) {
200
- const abs = tr.name ?? tr.testFilePath ?? '';
201
- if (!abs)
202
- continue;
203
- const ran = (tr.assertionResults ?? []).filter((a) => a.status === 'passed' || a.status === 'failed').length;
204
- out[relTestPath(abs, cwd)] = ran;
205
- }
206
- return out;
207
- }
208
- /** Repo-relative, forward-slashed key for a jest test-file path — shared by the parsers below. */
209
- function relTestPath(abs, cwd) {
210
- const rel = abs.startsWith(cwd) ? abs.slice(cwd.length).replace(/^[/\\]/, '') : abs;
211
- return rel.split('\\').join('/');
212
- }
213
- /**
214
- * @implements A-SPEC-534.1
215
- * Classify each jest-executed file into red-assertion | red-error | green from `jest --json` stdout.
216
- * Pure; unparseable/empty input → {} (same fail-soft convention as parseExecutedCounts).
217
- *
218
- * A failed assertion anywhere in a file wins (`red-assertion`) — that is a real red. A file that ran
219
- * NO assertions but errored (testExecError, or status 'failed' with an empty assertion list — a
220
- * load/collection/import failure) is `red-error`, which ART-8 must not accept as a valid RED. A file
221
- * that executed ≥1 assertion with none failing is `green`. A file that only pended/skipped carries no
222
- * evidence and is omitted entirely.
223
- */
224
- function classifyJestOutcomes(stdout, cwd) {
225
- const start = stdout.indexOf('{');
226
- if (start < 0)
227
- return {};
228
- let data;
229
- try {
230
- data = JSON.parse(stdout.slice(start));
231
- }
232
- catch {
233
- return {};
234
- }
235
- const out = {};
236
- for (const tr of data?.testResults ?? []) {
237
- const abs = tr.name ?? tr.testFilePath ?? '';
238
- if (!abs)
239
- continue;
240
- const rel = relTestPath(abs, cwd);
241
- const assertions = tr.assertionResults ?? [];
242
- if (assertions.some((a) => a.status === 'failed')) {
243
- out[rel] = 'red-assertion';
244
- continue;
245
- }
246
- const executed = assertions.filter((a) => a.status === 'passed' || a.status === 'failed').length;
247
- const errored = !!tr.testExecError || tr.status === 'failed';
248
- if (executed === 0) {
249
- if (errored)
250
- out[rel] = 'red-error'; // ran nothing AND failed → could not execute its cases
251
- continue; // otherwise all pending/skipped → no evidence, omit
252
- }
253
- out[rel] = errored ? 'red-error' : 'green';
254
- }
255
- return out;
256
- }
257
- /**
258
- * Parse pytest's built-in JUnit XML into per-file EXECUTED case counts — the pytest analogue of
259
- * parseExecutedCounts. `--junit-xml` ships with pytest core, so this needs no plugin, and it is the
260
- * only pytest output that is both machine-readable and per-file.
261
- *
262
- * A `<testcase>` counts as executed unless it carries `<skipped>`, matching the jest rule where
263
- * pending/skipped assertions are excluded. Errors and failures DO count: they executed.
264
- */
265
- function parseJUnitXmlCounts(xml, ext = 'py') {
266
- const out = {};
267
- for (const m of xml.matchAll(/<testcase\b([^>]*?)(\/>|>([\s\S]*?)<\/testcase>)/g)) {
268
- const attrs = m[1] ?? '';
269
- const body = m[3] ?? '';
270
- // `file` (xunit1 / dotnet loggers) is used verbatim when present. Otherwise the `classname`
271
- // fallback maps to a source path — but the class↔file relationship is language-specific:
272
- // @implements A-SPEC-137.3
273
- // py : the MODULE names the file, the class is a nested segment. `tests.test_calc.TestThing`
274
- // -> `tests/test_calc.py` (drop trailing Capitalized class segments).
275
- // java/cs : the CLASS names the file. `com.x.FooTest` -> `com/x/FooTest.java` (keep all).
276
- let file = /\bfile="([^"]*)"/.exec(attrs)?.[1];
277
- if (!file) {
278
- const cls = /\bclassname="([^"]*)"/.exec(attrs)?.[1];
279
- if (!cls)
280
- continue;
281
- const parts = cls.split('.').filter(Boolean);
282
- if (ext === 'py') {
283
- while (parts.length > 1 && /^[A-Z]/.test(parts[parts.length - 1]))
284
- parts.pop();
285
- }
286
- if (parts.length === 0)
287
- continue;
288
- file = `${parts.join('/')}.${ext}`;
289
- }
290
- if (/<skipped\b/.test(body))
291
- continue;
292
- const key = file.split('\\').join('/');
293
- out[key] = (out[key] ?? 0) + 1;
294
- }
295
- return out;
296
- }
167
+ // @implements A-SPEC-128.12 — the report readers live in review/test-reports (SRP); these names stay importable from here
168
+ // @implements A-SPEC-137.1, A-SPEC-137.3, A-SPEC-534.1 — the go, JUnit and jest-outcome readers they specify, re-exported
169
+ var test_reports_2 = require("./test-reports");
170
+ Object.defineProperty(exports, "parseGoTestJson", { enumerable: true, get: function () { return test_reports_2.parseGoTestJson; } });
171
+ Object.defineProperty(exports, "goPackagesWithTestEvents", { enumerable: true, get: function () { return test_reports_2.goPackagesWithTestEvents; } });
172
+ Object.defineProperty(exports, "parseExecutedCounts", { enumerable: true, get: function () { return test_reports_2.parseExecutedCounts; } });
173
+ Object.defineProperty(exports, "classifyJestOutcomes", { enumerable: true, get: function () { return test_reports_2.classifyJestOutcomes; } });
174
+ Object.defineProperty(exports, "parseJUnitXmlCounts", { enumerable: true, get: function () { return test_reports_2.parseJUnitXmlCounts; } });
175
+ Object.defineProperty(exports, "parseJUnitXmlOutcomes", { enumerable: true, get: function () { return test_reports_2.parseJUnitXmlOutcomes; } });
176
+ Object.defineProperty(exports, "parseJUnitXmlFiles", { enumerable: true, get: function () { return test_reports_2.parseJUnitXmlFiles; } });
177
+ Object.defineProperty(exports, "wellShapedJson", { enumerable: true, get: function () { return test_reports_2.wellShapedJson; } });
178
+ Object.defineProperty(exports, "readJestReport", { enumerable: true, get: function () { return test_reports_2.readJestReport; } });
297
179
  /**
298
180
  * @implements A-SPEC-466 — W-4. The TARGET repo's own jest entry, as its package declares it.
299
181
  *
@@ -322,8 +204,75 @@ function jestEntry(cwd) {
322
204
  return null;
323
205
  }
324
206
  }
207
+ /** @implements A-SPEC-128.6 — passed and failed cases only (skipped ones are not counted by the parsers). */
208
+ const casesRan = (executed) => Object.values(executed).reduce((a, b) => a + b, 0);
209
+ /**
210
+ * @implements A-SPEC-466.6 — test_run's `timeoutMinutes`, the run's time budget: absent or null is 30; an integer
211
+ * 1..180 is kept; anything else throws (never clamped). It is not a judgement input.
212
+ */
213
+ function testRunTimeoutMinutes(value) {
214
+ if (value === undefined || value === null)
215
+ return exports.RUNNER_TIMEOUT_MS / 60_000;
216
+ if (typeof value === 'number' && Number.isInteger(value) && value >= 1 && value <= 180)
217
+ return value;
218
+ throw new Error(`timeoutMinutes must be an integer from 1 to 180 (got ${typeof value === 'string' ? JSON.stringify(value) : String(value)})`);
219
+ }
220
+ /**
221
+ * @implements A-SPEC-466.1 — test_run's `workers`: absent stays absent; an integer 1..64 is kept; anything else
222
+ * throws, so a bad count is refused before any change set is derived or any runner starts.
223
+ */
224
+ function testRunWorkers(value) {
225
+ if (value === undefined || value === null)
226
+ return undefined; // @implements A-SPEC-466.3 — null reads as omitted
227
+ if (typeof value === 'number' && Number.isInteger(value) && value >= 1 && value <= 64)
228
+ return value;
229
+ throw new Error(`workers must be an integer from 1 to 64 (got ${typeof value === 'string' ? JSON.stringify(value) : String(value)})`);
230
+ }
231
+ /**
232
+ * @implements A-SPEC-128.12 — the one message for a configuration outside the 0.40.x supported surface (jest with one
233
+ * project and one unit, pytest under default naming, go). It leads every such environment reason.
234
+ */
235
+ exports.UNSUPPORTED = '이 구성은 0.40.x가 증빙하지 않습니다 — 0.41 / 오너 결정';
236
+ /** @implements A-SPEC-128.12 — S7: a runner output that could not be read is a value (the unit's environment), never a throw. */
237
+ const unreadable = (what, out, workers) => ({ passed: false, tail: `runner output could not be read (${what}): ${firstLine(out) || '(empty)'}`, executed: {}, outcomes: {}, workers, testsRun: 0, environment: true });
238
+ /**
239
+ * @implements A-SPEC-128.12 — S2/S7: what `jest --showConfig` resolved for this unit, judged by test-units'
240
+ * `jestConfigSurface` (A-SPEC-128.13: one place) — the reason it is outside the surface or cannot be read, else null.
241
+ * No pattern is taken from it.
242
+ */
243
+ function jestConfigReason(cmd, pre, cwd) {
244
+ let out;
245
+ try {
246
+ out = runBounded(cmd, [...pre, '--showConfig'], { cwd, encoding: 'utf8', stdio: ['ignore', 'pipe', 'pipe'], maxBuffer: RUNNER_MAX_BUFFER, env: (0, root_1.cleanTestEnv)() });
247
+ }
248
+ catch (e) {
249
+ const err = e;
250
+ out = `${err.stderr ?? ''}\n${err.stdout ?? ''}\n${err.message ?? ''}`;
251
+ }
252
+ const cfg = (0, test_reports_1.wellShapedJson)(out, test_reports_1.isJestConfig);
253
+ if (cfg === null || cfg.configs.length === 0)
254
+ return `runner output could not be read (jest --showConfig): ${firstLine(out) || '(empty)'}`;
255
+ const outside = (0, test_units_2.jestConfigSurface)(cfg);
256
+ return outside === null ? null : `${exports.UNSUPPORTED}: ${outside}`;
257
+ }
325
258
  /** Run the jest half of a plan. */
326
- function runJest(files, mode, cwd) {
259
+ function runJest(files, mode, cwd, opts = {}) {
260
+ // @implements A-SPEC-128.11 — P3: a full run's report is read from a file jest writes, so the user's console output
261
+ // on stdout cannot break it; a scoped run keeps its 466.1 argv
262
+ const reportDir = mode === 'full' ? fs.mkdtempSync(path.join(os.tmpdir(), 'holmes-jest-')) : null;
263
+ try {
264
+ return runJestWith(files, mode, cwd, opts, reportDir === null ? null : path.join(reportDir, 'report.json'));
265
+ }
266
+ finally {
267
+ if (reportDir !== null) {
268
+ try {
269
+ fs.rmSync(reportDir, { recursive: true, force: true });
270
+ }
271
+ catch { /* best effort */ }
272
+ }
273
+ }
274
+ }
275
+ function runJestWith(files, mode, cwd, opts, reportFile) {
327
276
  // --runTestsByPath treats args as EXACT files (not testPathPattern REGEXes, which would over/mis-match
328
277
  // or throw on a metachar); `--` + a leading-`./` guard so a path can never be read as a flag. NO
329
278
  // --passWithNoTests on a scoped run (verification H5): if the resolved paths match ZERO tests
@@ -335,15 +284,36 @@ function runJest(files, mode, cwd) {
335
284
  // (structurally no .cmd shim to ENOENT on); the npx spelling only as the fallback.
336
285
  const entry = jestEntry(cwd);
337
286
  const [cmd, ...pre] = entry !== null ? [process.execPath, entry] : ['npx', 'jest'];
287
+ // @implements A-SPEC-466.1 — absent, jest keeps the project's own maxWorkers
288
+ // @implements A-SPEC-466.2 — `applied` is the one value both the argv and the report are built from
289
+ const applied = opts.workers ?? null;
290
+ const workers = applied === null ? [] : [`--maxWorkers=${applied}`];
291
+ // @implements A-SPEC-128.12 — S2: holmes passes jest no ignore argument, so the user's configuration is never replaced
292
+ // (a CLI pattern applies to every project and hid one project's red, R1); exclusions are judged after the run
338
293
  const args = mode === 'scoped'
339
- ? [...pre, '--silent', '--json', '--forceExit', '--runTestsByPath', '--', ...safeFiles]
340
- : [...pre, '--silent', '--json', '--forceExit']; // round 13: an open handle after the last test must not hold the run
294
+ ? [...pre, '--silent', '--json', '--forceExit', ...workers, '--runTestsByPath', '--', ...safeFiles]
295
+ : [...pre, '--silent', '--json', '--forceExit', ...workers, `--outputFile=${reportFile}`]; // round 13: an open handle after the last test must not hold the run
296
+ const ran = jestRun(cmd, args, cwd, mode, reportFile, applied);
297
+ if (mode !== 'full' || ran.environment === true)
298
+ return ran;
299
+ // @implements A-SPEC-128.12 — S2/S7: one project only; a configuration jest cannot print is environment
300
+ // @implements A-SPEC-128.13 — and no configured testNamePattern (a selection below the file)
301
+ const reason = jestConfigReason(cmd, pre, cwd);
302
+ return reason === null ? ran : { ...ran, passed: false, environment: true, tail: `${reason}\n${ran.tail}` };
303
+ }
304
+ function jestRun(cmd, args, cwd, mode, reportFile, applied) {
341
305
  try {
342
- const out = runBounded(cmd, args, { cwd, encoding: 'utf8', stdio: ['ignore', 'pipe', 'pipe'], maxBuffer: RUNNER_MAX_BUFFER, env: (0, root_1.cleanTestEnv)() });
306
+ const raw = runBounded(cmd, args, { cwd, encoding: 'utf8', stdio: ['ignore', 'pipe', 'pipe'], maxBuffer: RUNNER_MAX_BUFFER, env: (0, root_1.cleanTestEnv)() });
307
+ const out = mode === 'full' ? (0, test_reports_1.readJestReport)(reportFile, raw) : raw;
308
+ if (out === null)
309
+ return unreadable('jest --json', raw, applied);
343
310
  // @implements A-SPEC-515.1 — the green path says what happened instead of handing back the
344
311
  // whole `--json` document. `tailOf` wraps the result too, so an unexpectedly long summary is
345
312
  // still bounded: a cap that any one path can escape is not a cap.
346
- return { passed: true, tail: tailOf(summarizeJestJson(out) ?? out), executed: parseExecutedCounts(out, cwd), outcomes: classifyJestOutcomes(out, cwd) };
313
+ const executed = (0, test_reports_1.parseExecutedCounts)(out, cwd);
314
+ // @implements A-SPEC-128.13 — T1: exit 0 is not green when the report holds a failure (testFailureExitCode: 0)
315
+ const failed = (0, test_reports_1.jestReportFailed)(out);
316
+ return { passed: !failed, tail: tailOf(`${failed ? 'jest exited 0 but its report holds failures\n' : ''}${summarizeJestJson(out) ?? out}`), executed, outcomes: (0, test_reports_1.classifyJestOutcomes)(out, cwd), workers: applied, testsRun: casesRan(executed) };
347
317
  }
348
318
  catch (e) {
349
319
  const err = e;
@@ -351,9 +321,19 @@ function runJest(files, mode, cwd) {
351
321
  // classification (A-SPEC-534.1) — survives a red run.
352
322
  // @implements A-SPEC-577.1 — the names first, then jest's own summary from stderr. Without the
353
323
  // first half this tail was the tail of a JSON document: it said how many failed and never which.
354
- const named = failedTestSummary(err.stdout ?? '');
355
- const tail = tailOf(`${named === null ? (err.stdout ?? '') : named}\n${err.stderr ?? err.message ?? ''}`, { lines: named === null ? 6 : 40 });
356
- return { passed: false, tail, executed: parseExecutedCounts(err.stdout ?? '', cwd), outcomes: classifyJestOutcomes(err.stdout ?? '', cwd) };
324
+ // @implements A-SPEC-466.3 — the names and the stderr are bounded apart, so a long stderr cannot push the names out
325
+ // @implements A-SPEC-128.11 — P3: a red full run is read from its report file too; with no report at all the unit
326
+ // ran nothing readable and is environment (a jest that could not start: a configuration error, a missing preset)
327
+ const report = mode === 'full' ? (0, test_reports_1.readJestReport)(reportFile, err.stdout ?? '') : (err.stdout ?? '');
328
+ // jest's own answer for a unit with no test writes no report: it ran nothing, read as such (the tests, not environment)
329
+ if (report === null && /^No tests found, exiting with code 1/m.test(err.stderr ?? '')) {
330
+ return { passed: false, tail: tailOf(err.stderr ?? '', 3), executed: {}, outcomes: {}, workers: applied, testsRun: 0 };
331
+ }
332
+ if (report === null)
333
+ return unreadable('jest --json', `${err.stderr ?? ''}\n${err.stdout ?? ''}\n${err.message ?? ''}`, applied);
334
+ const tail = redJestTail(report, err.stderr ?? err.message ?? '');
335
+ const executed = (0, test_reports_1.parseExecutedCounts)(report, cwd);
336
+ return { passed: false, tail, executed, outcomes: (0, test_reports_1.classifyJestOutcomes)(report, cwd), workers: applied, testsRun: casesRan(executed) };
357
337
  }
358
338
  }
359
339
  /**
@@ -361,113 +341,28 @@ function runJest(files, mode, cwd) {
361
341
  * Run specific jest files and return only their RED-first outcomes — the runner `test_run --mutate`
362
342
  * injects into runKillsOnFile. Scoped, jest only (the classifier is jest-only), fail-soft to {}.
363
343
  */
364
- function runJestOutcomes(files, cwd) {
344
+ function runJestOutcomes(files, cwd, timeoutMinutes) {
365
345
  if (files.length === 0)
366
346
  return {};
367
- return runJest(files, 'scoped', cwd).outcomes;
347
+ // @implements A-SPEC-466.6 — a mutation's jest child is a runner child of the run too
348
+ return withRunBudget(timeoutMinutes, () => runJest(files, 'scoped', cwd).outcomes);
368
349
  }
369
350
  /** Run the pytest half of a plan, taking execution evidence from pytest's built-in JUnit XML. */
370
351
  // @implements A-SPEC-502.1 — exported for the wiring test (a fake venv python capturing argv).
371
- /**
372
- * @implements A-SPEC-647
373
- * Per-file RED-first outcome from the same JUnit XML `parseJUnitXmlCounts` reads. Measured 2026-09-14
374
- * in an external pytest project: a ranged test_run executed the anchored case and recorded NOTHING,
375
- * because classification was jest-only — a Python project could never produce the red-assertion→green
376
- * sequence ART-8 reads. `<error>` (collection/import/setup) is a red-error, `<failure>` (an assertion)
377
- * a red-assertion, executed cases without either are green; a file with only skipped cases says nothing.
378
- */
379
- function parseJUnitXmlOutcomes(xml, ext = 'py') {
380
- const flags = new Map();
381
- for (const m of xml.matchAll(/<testcase\b([^>]*?)(\/>|>([\s\S]*?)<\/testcase>)/g)) {
382
- const attrs = m[1] ?? '';
383
- const body = m[3] ?? '';
384
- let file = /\bfile="([^"]*)"/.exec(attrs)?.[1];
385
- if (!file) {
386
- const cls = /\bclassname="([^"]*)"/.exec(attrs)?.[1];
387
- if (!cls)
388
- continue;
389
- const parts = cls.split('.').filter(Boolean);
390
- if (ext === 'py') {
391
- while (parts.length > 1 && /^[A-Z]/.test(parts[parts.length - 1]))
392
- parts.pop();
393
- }
394
- if (parts.length === 0)
395
- continue;
396
- file = `${parts.join('/')}.${ext}`;
397
- }
398
- const key = file.split('\\').join('/');
399
- const f = flags.get(key) ?? { error: false, failure: false, executed: false };
400
- if (/<error\b/.test(body))
401
- f.error = true;
402
- else if (/<failure\b/.test(body))
403
- f.failure = true;
404
- if (!/<skipped\b/.test(body))
405
- f.executed = true;
406
- flags.set(key, f);
407
- }
408
- const out = {};
409
- for (const [file, f] of flags) {
410
- if (f.error)
411
- out[file] = 'red-error';
412
- else if (f.failure)
413
- out[file] = 'red-assertion';
414
- else if (f.executed)
415
- out[file] = 'green';
416
- }
417
- return out;
418
- }
352
+ // @implements A-SPEC-647 — its per-file outcome reader (parseJUnitXmlOutcomes) lives in review/test-reports
419
353
  /**
420
354
  * @implements A-SPEC-647
421
355
  * The ecosystems a project's OWN test files belong to, by file-name convention, under a bounded walk
422
356
  * that skips vendored trees and virtualenvs. Measured: a full-tier run with no impacted list assumed
423
- * the jest suite on a pytest-only project, ran nothing and reported passed:false. This repository's
424
- * Python tests live under `reference/`, so it detects jest only and its runs are unchanged.
357
+ * the jest suite on a pytest-only project, ran nothing and reported passed:false.
358
+ * @implements A-SPEC-128.5 — configuration files name an ecosystem too (a pytest.ini, a go.mod), the root's are read
359
+ * even where the walk stops early, and the skipped trees a project vendors are its own declaration
360
+ * (`ax.config.json` testRun.excludeDirs), not a name built in here. This repository lists `reference`.
425
361
  */
426
- function detectProjectEcosystems(cwd) {
427
- const SKIP_DIRS = new Set(['node_modules', '.git', 'dist', 'dist-tools', 'build', 'vendor', 'third_party', 'reference']);
428
- const found = new Set();
429
- const walk = (dir, depth) => {
430
- if (depth > 6)
431
- return;
432
- let entries;
433
- try {
434
- entries = fs.readdirSync(dir, { withFileTypes: true });
435
- }
436
- catch {
437
- return;
438
- }
439
- if (entries.some((e) => e.isFile() && e.name === 'pyvenv.cfg'))
440
- return; // a virtualenv is not the project
441
- for (const e of entries) {
442
- const name = e.name;
443
- if (e.isDirectory()) {
444
- if (SKIP_DIRS.has(name) || (name.startsWith('.') && name !== '.'))
445
- continue;
446
- if (name === '__tests__')
447
- found.add('jest');
448
- walk(path.join(dir, name), depth + 1);
449
- continue;
450
- }
451
- if (!e.isFile())
452
- continue;
453
- if (/\.(test|spec)\.(ts|tsx|js|jsx|mjs|cjs|mts|cts)$/.test(name))
454
- found.add('jest');
455
- else if (/^(test_[^/]*|[^/]*_test)\.py$/.test(name) || name === 'conftest.py')
456
- found.add('pytest');
457
- else if (/_test\.go$/.test(name))
458
- found.add('go');
459
- else if (name === 'Cargo.toml')
460
- found.add('rust');
461
- else if (name === 'build.gradle' || name === 'build.gradle.kts')
462
- found.add('java');
463
- else if (/\.csproj$/.test(name))
464
- found.add('dotnet');
465
- }
466
- };
467
- walk(cwd, 0);
468
- return [...found].sort();
362
+ function detectProjectEcosystems(cwd, opts = {}) {
363
+ return (0, test_units_1.scanEcosystems)(cwd, opts).detected;
469
364
  }
470
- function runPytest(files, mode, cwd) {
365
+ function runPytest(files, mode, cwd, ignore = [], pythonFrom, run) {
471
366
  const dir = fs.mkdtempSync(path.join(os.tmpdir(), 'holmes-pytest-'));
472
367
  const report = path.join(dir, 'report.xml');
473
368
  // `python3 -m pytest` rather than a bare `pytest`, so the interpreter on PATH decides which
@@ -482,24 +377,48 @@ function runPytest(files, mode, cwd) {
482
377
  // @implements A-SPEC-502.1 — BUG-2, both halves: the interpreter resolves from pyvenv.cfg
483
378
  // evidence (HOLMES_PYTHON above all, python3 fallback byte-identical to the old path), and a
484
379
  // conftest.py argument widens to its directory instead of collection-ERRORing the whole run.
485
- const args = ['-m', 'pytest', '-q', '-o', 'junit_family=xunit1', `--junit-xml=${report}`,
380
+ // @implements A-SPEC-128.9 — `-p no:cacheprovider`: a test run leaves no .pytest_cache in the project; E3: an excluded
381
+ // or ignored tree is not collected
382
+ // @implements A-SPEC-128.11 — P3: `--rootdir` is the unit, so a pytest.ini above the repository cannot move the
383
+ // report's file keys out of it (pt-rootdir-above)
384
+ const rootdir = fs.existsSync(cwd) ? (0, root_1.canonicalPath)(cwd) : cwd;
385
+ // @implements A-SPEC-128.13 — W2: the work tree's interpreter (a .venv is not in HEAD)
386
+ const python = (0, python_env_1.pythonFor)(pythonFrom && fs.existsSync(pythonFrom) ? pythonFrom : cwd, { env: process.env, exists: fs.existsSync, platform: process.platform });
387
+ // @implements A-SPEC-128.14, A-SPEC-128.15, A-SPEC-128.16 — in HEAD's worktree: the pinned configuration and its confcutdir,
388
+ // the environment asked where HEAD's names resolve (refused before any test), and the loaded-module record
389
+ const prov = run ? (0, run_provenance_1.pytestRunProvenance)(dir, run, rootdir, python, (0, root_1.cleanTestEnv)(), (c, a, o) => runBounded(c, a, { ...o, encoding: 'utf8', stdio: ['ignore', 'pipe', 'pipe'], maxBuffer: RUNNER_MAX_BUFFER })) : null;
390
+ if (prov?.refused) {
391
+ fs.rmSync(dir, { recursive: true, force: true });
392
+ return { passed: false, tail: prov.refused, executed: {}, outcomes: {}, testsRun: 0, environment: true };
393
+ }
394
+ const args = ['-m', 'pytest', '-q', '-p', 'no:cacheprovider', '-o', 'junit_family=xunit1', `--junit-xml=${report}`, `--rootdir=${rootdir}`, ...(prov?.args ?? []),
395
+ ...(mode === 'full' ? [`--confcutdir=${prov?.confcutdir ?? rootdir}`, ...ignore.map((d) => `--ignore=${d.replace(/\/+$/, '')}`)] : []),
486
396
  ...(mode === 'scoped' ? ['--', ...(0, python_env_1.pytestTargets)(files)] : [])];
487
- const python = (0, python_env_1.pythonFor)(cwd, { env: process.env, exists: fs.existsSync, platform: process.platform });
397
+ const judged = (r) => { const why = prov?.judge(r.passed) ?? null; return why === null ? r : { passed: false, tail: why, executed: {}, outcomes: {}, testsRun: 0, environment: true }; };
488
398
  const read = () => {
489
399
  try {
490
- return parseJUnitXmlCounts(fs.readFileSync(report, 'utf8'));
400
+ return (0, test_reports_1.parseJUnitXmlCounts)(fs.readFileSync(report, 'utf8'));
491
401
  }
492
402
  catch {
493
403
  return {};
494
404
  }
495
405
  };
406
+ // @implements A-SPEC-128.10 — I3: every file pytest collected, skipped ones included
407
+ const reported = () => {
408
+ try {
409
+ return (0, test_reports_1.parseJUnitXmlFiles)(fs.readFileSync(report, 'utf8'));
410
+ }
411
+ catch {
412
+ return [];
413
+ }
414
+ };
496
415
  // @implements A-SPEC-647 — per-file outcomes from the same report; in a scoped run a planned file the
497
416
  // failed run never reached (no case at all — a collection error before its cases) is a red-error, so a
498
417
  // RED that never ran is not mistaken for evidence of nothing.
499
418
  const readOutcomes = (passed) => {
500
419
  let outcomes = {};
501
420
  try {
502
- outcomes = parseJUnitXmlOutcomes(fs.readFileSync(report, 'utf8'));
421
+ outcomes = (0, test_reports_1.parseJUnitXmlOutcomes)(fs.readFileSync(report, 'utf8'));
503
422
  }
504
423
  catch {
505
424
  outcomes = {};
@@ -518,13 +437,18 @@ function runPytest(files, mode, cwd) {
518
437
  // @implements A-SPEC-651.1 — launch from the canonical spelling: pytest keys its JUnit `file`
519
438
  // attribute relative to its rootdir, and an 8.3 invocation dir made that key climb out of the
520
439
  // project (`../../…/tests/test_calc.py`, measured 2026-09-14 on Windows).
521
- const out = runBounded(python, args, { cwd: fs.existsSync(cwd) ? (0, root_1.canonicalPath)(cwd) : cwd, encoding: 'utf8', stdio: ['ignore', 'pipe', 'pipe'], maxBuffer: RUNNER_MAX_BUFFER, env: (0, root_1.cleanTestEnv)() });
522
- return { passed: true, tail: tailOf(out), executed: read(), outcomes: readOutcomes(true) };
440
+ const out = runBounded(python, args, { cwd: fs.existsSync(cwd) ? (0, root_1.canonicalPath)(cwd) : cwd, encoding: 'utf8', stdio: ['ignore', 'pipe', 'pipe'], maxBuffer: RUNNER_MAX_BUFFER, env: prov ? prov.env((0, root_1.cleanTestEnv)()) : (0, root_1.cleanTestEnv)() });
441
+ const executed = read();
442
+ // @implements A-SPEC-128.13 — T1: exit 0 is not green when the JUnit report holds a failure or an error
443
+ const outcomes = readOutcomes(true);
444
+ const red = Object.values(outcomes).some((o) => o !== 'green');
445
+ return judged({ passed: !red, tail: tailOf(`${red ? 'pytest exited 0 but its report holds failures\n' : ''}${out}`), executed, outcomes, testsRun: casesRan(executed), reported: reported() });
523
446
  }
524
447
  catch (e) {
525
448
  const err = e;
526
449
  // A red pytest run still writes the XML, so execution evidence survives — same property as jest.
527
- return { passed: false, tail: tailOf(`${err.stdout ?? ''}\n${err.stderr ?? err.message ?? ''}`), executed: read(), outcomes: readOutcomes(false) };
450
+ const executed = read();
451
+ return judged({ passed: false, tail: tailOf(`${err.stdout ?? ''}\n${err.stderr ?? err.message ?? ''}`), executed, outcomes: readOutcomes(false), testsRun: casesRan(executed), reported: reported() });
528
452
  }
529
453
  finally {
530
454
  try {
@@ -586,7 +510,8 @@ function runCargo(files, _mode, cwd, opts) {
586
510
  executed[f] = count; // crate-granular
587
511
  // Executing nothing is not a pass, mirroring no-passWithNoTests.
588
512
  const passed = ok && count > 0;
589
- return { passed, tail: `[rust] cargo test: ${count} executed`, executed };
513
+ // @implements A-SPEC-128.5 — the crate's count once, whatever the per-file credit (a full run credits no file)
514
+ return { passed, tail: `[rust] cargo test: ${count} executed`, executed, testsRun: count };
590
515
  }
591
516
  /** Read every JUnit XML report under a directory (recursively) and sum their counts for `ext`. */
592
517
  function readJUnitReports(dir, ext) {
@@ -605,7 +530,7 @@ function readJUnitReports(dir, ext) {
605
530
  walk(p);
606
531
  else if (e.name.endsWith('.xml')) {
607
532
  try {
608
- const counts = parseJUnitXmlCounts(fs.readFileSync(p, 'utf8'), ext);
533
+ const counts = (0, test_reports_1.parseJUnitXmlCounts)(fs.readFileSync(p, 'utf8'), ext);
609
534
  for (const [k, v] of Object.entries(counts))
610
535
  acc[k] = (acc[k] ?? 0) + v;
611
536
  }
@@ -668,12 +593,15 @@ function runGradle(files, _mode, cwd, opts) {
668
593
  try {
669
594
  runBounded(cmd, args, { cwd, encoding: 'utf8', stdio: ['ignore', 'pipe', 'pipe'], maxBuffer: RUNNER_MAX_BUFFER, env: (0, root_1.cleanTestEnv)() });
670
595
  }
671
- catch {
672
- ok = false;
673
- } // a failing test run still wrote the JUnit XML
596
+ catch (e) {
597
+ // @implements A-SPEC-128.10 — R4: the build tool itself missing is the environment, not the tests
598
+ if (e.code === 'ENOENT')
599
+ return { passed: false, tail: `${cmd} not found on PATH — cannot execute Java tests`, executed: {}, unsupported: files };
600
+ ok = false; // a failing test run still wrote the JUnit XML
601
+ }
674
602
  const executed = readJUnitReports(path.join(cwd, reportDir), 'java');
675
603
  const total = Object.values(executed).reduce((a, b) => a + b, 0);
676
- return { passed: ok && total > 0, tail: `[java] ${cmd} ${args.join(' ')}: ${total} executed`, executed };
604
+ return { passed: ok && total > 0, tail: `[java] ${cmd} ${args.join(' ')}: ${total} executed`, executed, testsRun: total };
677
605
  }
678
606
  /**
679
607
  * @implements A-SPEC-137.3
@@ -705,7 +633,7 @@ function runDotnet(files, _mode, cwd, opts) {
705
633
  }
706
634
  let executed = {};
707
635
  try {
708
- executed = parseJUnitXmlCounts(fs.readFileSync(report, 'utf8'), 'cs');
636
+ executed = (0, test_reports_1.parseJUnitXmlCounts)(fs.readFileSync(report, 'utf8'), 'cs');
709
637
  }
710
638
  catch { /* no report */ }
711
639
  try {
@@ -713,7 +641,7 @@ function runDotnet(files, _mode, cwd, opts) {
713
641
  }
714
642
  catch { /* best effort */ }
715
643
  const total = Object.values(executed).reduce((a, b) => a + b, 0);
716
- return { passed: ok && total > 0, tail: `[dotnet] dotnet test: ${total} executed`, executed };
644
+ return { passed: ok && total > 0, tail: `[dotnet] dotnet test: ${total} executed`, executed, testsRun: total };
717
645
  }
718
646
  /**
719
647
  * @implements A-SPEC-137.1
@@ -723,6 +651,7 @@ function runDotnet(files, _mode, cwd, opts) {
723
651
  * `unsupported`, never a pass. `opts.goPresent` overrides the probe for testing.
724
652
  */
725
653
  function runGo(files, mode, cwd, opts) {
654
+ const goEnv = { ...(0, root_1.cleanTestEnv)(), ...(opts?.env ?? {}) };
726
655
  const present = opts?.goPresent ?? (() => {
727
656
  try {
728
657
  (0, node_child_process_1.execFileSync)('go', ['version'], { cwd, stdio: 'ignore', timeout: PROBE_TIMEOUT_MS });
@@ -742,10 +671,10 @@ function runGo(files, mode, cwd, opts) {
742
671
  const tails = [];
743
672
  let passed = true;
744
673
  let anyExecuted = 0;
745
- const runDir = (dirArg, creditFiles) => {
674
+ const runDir = (dirArg, creditFiles, onOutput) => {
746
675
  let out = '';
747
676
  try {
748
- out = runBounded('go', ['test', '-json', dirArg], { cwd, encoding: 'utf8', stdio: ['ignore', 'pipe', 'pipe'], maxBuffer: RUNNER_MAX_BUFFER, env: (0, root_1.cleanTestEnv)() });
677
+ out = runBounded('go', ['test', '-json', dirArg], { cwd, encoding: 'utf8', stdio: ['ignore', 'pipe', 'pipe'], maxBuffer: RUNNER_MAX_BUFFER, env: goEnv });
749
678
  }
750
679
  catch (e) {
751
680
  const err = e;
@@ -753,15 +682,63 @@ function runGo(files, mode, cwd, opts) {
753
682
  tails.push(err.stderr ? tailOf(err.stderr, 2) : (err.message ?? 'go test failed'));
754
683
  passed = false;
755
684
  }
756
- const perPkg = parseGoTestJson(out);
685
+ // @implements A-SPEC-128.13 — T1: a named test's fail event is red whatever the exit code (TestMain's os.Exit(0))
686
+ const failedTests = (0, test_reports_1.goFailedTests)(out);
687
+ if (failedTests.length > 0) {
688
+ passed = false;
689
+ tails.push(`[go] failed: ${failedTests.slice(0, 5).join(', ')}`);
690
+ }
691
+ const perPkg = (0, test_reports_1.parseGoTestJson)(out);
757
692
  const dirCount = Object.values(perPkg).reduce((a, b) => a + b, 0);
758
693
  anyExecuted += dirCount;
759
694
  for (const f of creditFiles)
760
695
  executed[f] = (executed[f] ?? 0) + dirCount; // package-granular
696
+ if (onOutput)
697
+ onOutput(out);
761
698
  tails.push(`[go] ${dirArg}: ${dirCount} executed`);
762
699
  };
763
700
  if (mode === 'full') {
764
- runDir('./...', []);
701
+ // @implements A-SPEC-128.9 — E3: the module's packages minus the excluded or ignored trees
702
+ // @implements A-SPEC-128.10 — I3 at package granularity: each package's count is credited to its _test.go files, so a
703
+ // package whose tests did not run leaves its files unreported
704
+ let pkgs = [];
705
+ const base = fs.existsSync(cwd) ? (0, root_1.canonicalPath)(cwd) : cwd;
706
+ try {
707
+ pkgs = runBounded('go', ['list', '-f', '{{.ImportPath}}\t{{.Dir}}', './...'], { cwd, encoding: 'utf8', stdio: ['ignore', 'pipe', 'pipe'], maxBuffer: RUNNER_MAX_BUFFER, env: goEnv })
708
+ .split('\n').map((l) => l.trim()).filter(Boolean)
709
+ .map((l) => { const [importPath, dir] = l.split('\t'); return { importPath, rel: path.relative(base, (0, root_1.canonicalPath)(dir ?? base)).split(path.sep).join('/') }; });
710
+ }
711
+ catch {
712
+ pkgs = [];
713
+ }
714
+ const skip = (opts?.ignore ?? []).map((x) => x.replace(/\/+$/, ''));
715
+ const kept = pkgs.filter((p) => !skip.some((x) => p.rel === x || p.rel.startsWith(`${x}/`)));
716
+ const before = new Map();
717
+ const credit = (out) => {
718
+ const perPkg = (0, test_reports_1.parseGoTestJson)(out);
719
+ const seen = (0, test_reports_1.goPackagesWithTestEvents)(out);
720
+ for (const p of kept) {
721
+ const n = perPkg[p.importPath] ?? 0;
722
+ // @implements A-SPEC-128.11 — P5: a package whose tests all skipped ran them (go-skip); the count stays pass/fail
723
+ if (!seen.has(p.importPath) || before.has(p.importPath))
724
+ continue;
725
+ before.set(p.importPath, n);
726
+ let names = [];
727
+ try {
728
+ names = fs.readdirSync(path.join(base, p.rel)).filter((f) => f.endsWith('_test.go'));
729
+ }
730
+ catch {
731
+ names = [];
732
+ }
733
+ for (const f of names)
734
+ executed[p.rel === '' ? f : `${p.rel}/${f}`] = n;
735
+ }
736
+ };
737
+ if (skip.length === 0)
738
+ runDir('./...', [], credit);
739
+ else
740
+ for (const k of kept)
741
+ runDir(k.rel === '' ? '.' : `./${k.rel}`, [], credit);
765
742
  }
766
743
  else {
767
744
  const byDir = new Map();
@@ -778,7 +755,8 @@ function runGo(files, mode, cwd, opts) {
778
755
  // A run that executed NOTHING is not a pass — the same no-passWithNoTests rule jest/pytest hold.
779
756
  if (anyExecuted === 0)
780
757
  passed = false;
781
- return { passed, tail: tails.join('\n'), executed };
758
+ // @implements A-SPEC-128.5 — a full module run credits no file but its tests ran: the count rides apart from the credit
759
+ return { passed, tail: tails.join('\n'), executed, testsRun: anyExecuted };
782
760
  }
783
761
  /**
784
762
  * Execute a plan (I/O). Deterministic inputs → deterministic command; the pass/fail is the suite's.
@@ -789,59 +767,234 @@ function runGo(files, mode, cwd, opts) {
789
767
  * ecosystem has no adapter are returned in `unsupported` — never dropped quietly, because a scoped
790
768
  * gate that ran part of its scope has verified less than it reports.
791
769
  */
792
- function runTestScope(scope, cwd) {
793
- const plan = planTestRun(scope);
794
- if (plan.mode === 'noop') {
795
- return { tier: scope.tier, mode: 'noop', ranFiles: [], passed: true, skipped: true, tail: plan.reason };
770
+ function runTestScope(scope, cwd, opts = {}) {
771
+ return withRunBudget(opts.timeoutMinutes, () => runScopeWithinBudget(scope, cwd, opts));
772
+ }
773
+ /** @implements A-SPEC-466.6 — each runner child started inside `fn` takes the budget; the default comes back after. */
774
+ function withRunBudget(timeoutMinutes, fn) {
775
+ RUN_BUDGET_MS = (timeoutMinutes ?? exports.RUNNER_TIMEOUT_MS / 60_000) * 60_000;
776
+ try {
777
+ return fn();
778
+ }
779
+ finally {
780
+ RUN_BUDGET_MS = exports.RUNNER_TIMEOUT_MS;
796
781
  }
797
- // planTestRun deliberately empties testFiles for a `full` run (the point is to run everything), so
798
- // ecosystem detection has to read the SCOPE's impacted files instead. Grouping on plan.testFiles
799
- // alone made every full-tier run fall back to jest — i.e. a Python project's full regression
800
- // executed the JavaScript suite and reported on it. Found by review, not by a failing test.
801
- const detectFrom = plan.mode === 'full' ? scope.impactedTestFiles : plan.testFiles;
782
+ }
783
+ /**
784
+ * @implements A-SPEC-128.4 — which ecosystem groups a plan runs. A scoped plan groups its own files. A full plan runs
785
+ * every ecosystem the PROJECT has, together with those of the impacted test files: detecting from the impacted files
786
+ * alone let a full run of a jest+pytest project skip pytest and still record `mode: full` (review 2026-11-05, A H1).
787
+ * Jest is assumed only when the project shows no ecosystem at all, and the run says so.
788
+ */
789
+ function runEcosystems(scope, plan, cwd, opts = {}) {
802
790
  const groups = new Map();
803
791
  const unsupported = [];
804
- for (const f of detectFrom) {
792
+ const notRun = [];
793
+ if (plan.mode !== 'full') {
794
+ for (const f of plan.testFiles) {
795
+ const eco = ecosystemOf(f);
796
+ if (!eco) {
797
+ unsupported.push(f);
798
+ continue;
799
+ }
800
+ groups.set(eco, [...(groups.get(eco) ?? []), f]);
801
+ }
802
+ const units = [...groups.entries()].map(([ecosystem, files]) => ({ ecosystem, dir: '', ignore: [], files }));
803
+ return { groups, units, unsupported, assumedJest: false, detected: [], detectedBy: {}, notRun, notDetected: [],
804
+ implementationCount: 0, implementationSample: [], notTests: [], expected: {}, codeRoots: [], buildRoots: [], sparseTests: [], outsideSurface: [], testFiles: [], notBuilt: [] };
805
+ }
806
+ // @implements A-SPEC-647 — run what the project HAS; @implements A-SPEC-128.4 — always, not only with no impacted file
807
+ // @implements A-SPEC-128.7 — I1: an impacted test file of an ecosystem with no unit is named in notRun, never run
808
+ // @implements A-SPEC-128.9 — E1: what runs is the project's execution units, each where its build root is
809
+ const scan = (0, test_units_1.scanEcosystems)(cwd, opts);
810
+ // @implements A-SPEC-128.12 — S1: only jest, pytest and go units run; the others are named outside the surface
811
+ const units = scan.units.filter((u) => SUPPORTED.has(u.ecosystem)).map((u) => ({ ...u, files: [] }));
812
+ let assumedJest = false;
813
+ // @implements A-SPEC-128.8 — jest is assumed only beside a root package.json (A M1);
814
+ // @implements A-SPEC-128.9 — and only where there is code to test (E2): nothing testable runs nothing
815
+ if (units.length === 0 && scan.implementationCount > 0 && fs.existsSync(path.join(cwd, 'package.json'))) {
816
+ units.push({ ecosystem: 'jest', dir: '', ignore: scan.excluded, files: [] });
817
+ assumedJest = true;
818
+ }
819
+ for (const u of units)
820
+ groups.set(u.ecosystem, []);
821
+ for (const f of scope.impactedTestFiles) {
805
822
  const eco = ecosystemOf(f);
806
823
  if (!eco) {
807
824
  unsupported.push(f);
808
825
  continue;
809
826
  }
810
- // In full mode the runners ignore the file list; the group only records which ecosystems exist.
811
- groups.set(eco, plan.mode === 'full' ? [] : [...(groups.get(eco) ?? []), f]);
827
+ if (groups.has(eco) || notRun.some((n) => n.ecosystem === eco))
828
+ continue;
829
+ notRun.push({ ecosystem: eco, reason: `impacted ${eco} test files, but no ${eco} unit holds them — a full run runs the project's units only` });
812
830
  }
813
- // A full run with no impacted files at all gives no signal about which ecosystems are present.
814
- // Defaulting to jest keeps the historical behavior; the assumption is stated in the evidence tail
815
- // rather than left implicit, because "we ran the wrong suite" and "we ran nothing" look identical
816
- // to a caller that only sees passed=true.
817
- let assumedJest = false;
818
- if (plan.mode === 'full' && groups.size === 0) {
819
- // @implements A-SPEC-647 — run what the project HAS; assume jest only when nothing is detected.
820
- for (const eco of detectProjectEcosystems(cwd))
821
- groups.set(eco, []);
822
- if (groups.size === 0) {
823
- groups.set('jest', []);
824
- assumedJest = true;
831
+ // @implements A-SPEC-128.13 — T4: a go unit expects only the _test.go files go builds here (go list, no tags); the rest
832
+ // are named in notBuilt and never refused (a red behind a build tag is a named limit)
833
+ const expected = { ...scan.expected };
834
+ const notBuilt = [];
835
+ for (const u of units.filter((x) => x.ecosystem === 'go')) {
836
+ const key = `go@${u.dir}`;
837
+ const listed = (0, test_units_2.goListTests)(cwd, u.dir);
838
+ if (listed === null || !expected[key])
839
+ continue;
840
+ for (const f of expected[key])
841
+ if (!listed.built.has(f))
842
+ notBuilt.push(f);
843
+ expected[key] = expected[key].filter((f) => listed.built.has(f));
844
+ }
845
+ return { groups, units, unsupported, assumedJest, detected: scan.detected, detectedBy: scan.detectedBy, notRun, notDetected: scan.notDetected,
846
+ implementationCount: scan.implementationCount, implementationSample: scan.implementationSample, notTests: scan.notTests,
847
+ expected, codeRoots: scan.codeRoots, buildRoots: scan.buildRoots, sparseTests: scan.sparseTests,
848
+ outsideSurface: scan.outsideSurface, testFiles: scan.testFiles, notBuilt: notBuilt.sort() };
849
+ }
850
+ const SUPPORTED = new Set(['jest', 'pytest', 'go']);
851
+ /**
852
+ * @implements A-SPEC-128.10 — I3 and R2, judged AFTER the run from what the runners reported: the expected test files
853
+ * a unit did not report (a runner configuration, a root narrower than the tree, a package that ran nothing), and the
854
+ * build roots whose code saw no executed test. Units that could not run (environment, stop) are judged elsewhere.
855
+ */
856
+ function coverageAfterRun(expected, codeRoots, reported, executedByFile, skipUnits, buildRoots = [], testFiles) {
857
+ const notExecuted = [];
858
+ for (const [unit, files] of Object.entries(expected)) {
859
+ if (skipUnits.has(unit))
860
+ continue;
861
+ for (const f of files)
862
+ if (!reported.has(f))
863
+ notExecuted.push(f);
864
+ }
865
+ // @implements A-SPEC-128.12 — S5 (repairs 128.11 P2): a root's code is tested only by an executed test file whose OWN
866
+ // nearest build root is that root; that a unit ran is not evidence (root-unit-credit, ws-root-credit)
867
+ const roots = new Set(buildRoots);
868
+ const nearestRoot = (rel) => {
869
+ const parts = rel.split('/').slice(0, -1);
870
+ for (let i = parts.length; i > 0; i--) {
871
+ const d = parts.slice(0, i).join('/');
872
+ if (roots.has(d))
873
+ return d;
825
874
  }
875
+ return '';
876
+ };
877
+ // @implements A-SPEC-128.13 — T3: only an executed test file of the scanned universe credits code (a test outside the
878
+ // repository or outside the scan credits nothing); when no universe is given, every executed file counts (scoped)
879
+ const universe = testFiles ? new Set(testFiles) : null;
880
+ const testedRoots = new Set();
881
+ for (const [f, n] of Object.entries(executedByFile))
882
+ if (n > 0 && (universe === null || universe.has(f)))
883
+ testedRoots.add(nearestRoot(f));
884
+ const tested = (root) => testedRoots.has(root);
885
+ const testsAt = (root) => (testFiles ?? []).filter((f) => nearestRoot(f) === root).length;
886
+ const untestedCode = codeRoots.filter((g) => !tested(g.root))
887
+ .map((g) => ({ root: g.root, sample: g.sample, ...(g.missingBuildFile ? { missingBuildFile: g.missingBuildFile } : {}), ...(testsAt(g.root) > 0 ? { testFiles: testsAt(g.root) } : {}) }));
888
+ return { notExecuted: notExecuted.sort(), untestedCode };
889
+ }
890
+ /**
891
+ * @implements A-SPEC-128.9 — E4: why a unit that ran nothing failed, when the cause is its environment, not its tests: a
892
+ * missing runner. Null when the unit executed a test or failed for any other reason (then it is the tests).
893
+ */
894
+ function environmentReason(r, counted) {
895
+ // @implements A-SPEC-128.12 — an adapter that says environment is believed even when tests ran (jest projects, a
896
+ // configuration jest cannot print: the run executed, but it is not 0.40.x evidence)
897
+ if (r.environment === true)
898
+ return firstLine(r.tail);
899
+ if (r.passed || counted > 0)
900
+ return null;
901
+ if ((r.unsupported?.length ?? 0) > 0)
902
+ return firstLine(r.tail);
903
+ // @implements A-SPEC-128.10 — R4: only the RUNNER being absent; an ENOENT inside a test, or a module a test imports,
904
+ // is the tests (review 10:51, A H4 / A M4)
905
+ const m = /No module named '?pytest'?(?![\w.])|toolchain not found on PATH|java runtime not found|dotnet not found on PATH|not found on PATH — cannot execute|no gradle\/maven build file found/.exec(r.tail);
906
+ return m ? m[0] : null;
907
+ }
908
+ // @implements A-SPEC-128.12 — S2/S6: withoutExcluded (a unit's report without its excluded trees) lives in test-reports
909
+ function firstLine(s) {
910
+ return (s.split('\n').find((l) => l.trim() !== '') ?? s).trim().slice(0, 200);
911
+ }
912
+ function runScopeWithinBudget(scope, cwd, opts) {
913
+ const plan = planTestRun(scope);
914
+ if (plan.mode === 'noop') {
915
+ return { tier: scope.tier, mode: 'noop', ranFiles: [], passed: true, skipped: true, tail: plan.reason, workers: null };
826
916
  }
827
- if (groups.size > 0) {
828
- const runners = { jest: runJest, pytest: runPytest, go: runGo, rust: runCargo, java: runGradle, dotnet: runDotnet };
917
+ const { units, unsupported, assumedJest, detected, detectedBy, notRun, notDetected, implementationCount, implementationSample, notTests, expected, codeRoots, buildRoots, sparseTests, outsideSurface, testFiles, notBuilt } = runEcosystems(scope, plan, cwd, { excludeDirs: opts.excludeDirs, isIgnored: opts.isIgnored });
918
+ // @implements A-SPEC-128.7 — what a full plan detected, left out and did not run travels with the result
919
+ // @implements A-SPEC-128.9 — with the implementation count E2 reads and the test-named files that hold no test
920
+ // @implements A-SPEC-128.13 — and the test-file count E2 also needs, and the go files this platform does not build
921
+ const named = plan.mode === 'full'
922
+ ? { detectedBy, implementationCount, implementationSample, testFileCount: testFiles.length, ...(notRun.length ? { notRun } : {}),
923
+ ...(notDetected.length ? { notDetected } : {}), ...(notTests.length ? { notTests } : {}), ...(notBuilt.length ? { notBuilt } : {}) } : {};
924
+ // @implements A-SPEC-128.12 — S1, S2, S3: outside the surface, one message, environment
925
+ // @implements A-SPEC-128.13 — W3: the work-tree rule is gone (a full run is HEAD's worktree); a configuration known to
926
+ // be outside the surface before the run starts no unit (B M2)
927
+ const surface = plan.mode !== 'full' ? [] : outsideSurface.map((o) => ({ ecosystem: o.ecosystem, dir: o.dir, reason: `${exports.UNSUPPORTED}: ${o.reason}` }));
928
+ if (units.length > 0 && surface.length === 0) {
929
+ const reported = new Set(); // @implements A-SPEC-128.10 — I3: every file a runner said it ran or collected
930
+ const skipUnits = new Set(); // units that could not run: their coverage is not the comparison's to judge
931
+ const runners = {
932
+ jest: (f, m, c) => (m === 'full' && jestEntry(c) === null
933
+ ? { passed: false, tail: `jest is not installed for this unit (no local node_modules/jest from ${c})`, executed: {}, workers: null, environment: true }
934
+ : runJest(f, m, c, opts)),
935
+ pytest: (f, m, c, ignore) => runPytest(f, m, c, ignore, opts.sourceRoot ? path.join(opts.sourceRoot, path.relative(cwd, c)) : undefined, opts.sourceRoot && m === 'full' ? { root: cwd, sourceRoot: opts.sourceRoot } : undefined),
936
+ go: (f, m, c, ignore) => runGo(f, m, c, { ignore, env: (0, test_units_2.goWorkEnv)(cwd) }),
937
+ rust: (f, m, c) => runCargo(f, m, c), java: (f, m, c) => runGradle(f, m, c), dotnet: (f, m, c) => runDotnet(f, m, c),
938
+ };
829
939
  const ran = [];
830
940
  let passed = true;
831
941
  const tails = [];
832
942
  const executedByFile = {};
833
943
  const outcomeByFile = {}; // @implements A-SPEC-534.5 — jest only
834
- for (const eco of [...groups.keys()].sort()) {
944
+ let workers = null; // @implements A-SPEC-466.2 — what the jest adapter placed on its argv
945
+ let testsRun = 0; // @implements A-SPEC-128.5
946
+ const testsRunBy = {}; // @implements A-SPEC-128.6 — per ecosystem, so none hides behind another
947
+ let stopped = false;
948
+ const failedBy = []; // @implements A-SPEC-128.7 — which runner was red, so a refusal can name its tree
949
+ const unitResults = []; // @implements A-SPEC-128.9 — per unit: every unit must run a test
950
+ const environment = [];
951
+ const keyed = (dir, rec) => (dir === '' ? rec : Object.fromEntries(Object.entries(rec).map(([k, v]) => [`${dir}/${k}`, v])));
952
+ for (const u of [...units].sort((a, b) => (a.ecosystem === b.ecosystem ? a.dir.localeCompare(b.dir) : a.ecosystem.localeCompare(b.ecosystem)))) {
953
+ const eco = u.ecosystem;
835
954
  STOPPED = null;
836
955
  LEFTOVER = null;
837
- const r = runners[eco](groups.get(eco), plan.mode, cwd);
838
- ran.push(eco);
839
- passed = passed && r.passed && STOPPED === null;
840
- tails.push(`[${eco}] ${r.tail}${STOPPED !== null ? `\n${STOPPED}` : ''}${LEFTOVER !== null ? `\n${LEFTOVER}` : ''}`);
841
- Object.assign(executedByFile, r.executed);
956
+ const unitCwd = u.dir === '' ? cwd : path.join(cwd, u.dir);
957
+ const raw = runners[eco](u.files, plan.mode, unitCwd, u.ignore);
958
+ // @implements A-SPEC-128.12 — S2/S6: exclusions judged after the run. A reported test under an excluded tree leaves
959
+ // the judgement and the counts; one that ran red is environment, named — never hidden
960
+ const { r, excludedRed } = plan.mode === 'full' ? (0, test_reports_1.withoutExcluded)(raw, u.ignore, u.dir) : { r: raw, excludedRed: [] };
961
+ if (eco === 'jest' && workers === null)
962
+ workers = r.workers ?? null;
963
+ if (!ran.includes(eco))
964
+ ran.push(eco);
965
+ passed = passed && r.passed && STOPPED === null && excludedRed.length === 0;
966
+ stopped = stopped || STOPPED !== null;
967
+ if ((!r.passed || STOPPED !== null) && !failedBy.includes(eco))
968
+ failedBy.push(eco);
969
+ // @implements A-SPEC-128.5 — an adapter that counts apart from its credit says so; jest and pytest credit every test
970
+ const own = r.testsRun;
971
+ // @implements A-SPEC-128.6 — every adapter counts for itself; one that does not reports 0, never a guess
972
+ const counted = typeof own === 'number' ? own : 0;
973
+ testsRunBy[eco] = (testsRunBy[eco] ?? 0) + counted;
974
+ testsRun += counted;
975
+ // @implements A-SPEC-128.13 — T3: a test the runner reported from outside the repository (roots or testpaths that
976
+ // leave it) is named, never credited; A-SPEC-128.14 — judged from the repository root (`../shared` from backend is inside)
977
+ const outside = Object.keys(r.executed).filter((k) => path.isAbsolute(k) || path.posix.normalize(u.dir === '' ? k : `${u.dir}/${k}`).startsWith('../'));
978
+ // every reason is kept: a jest projects configuration whose excluded tree also ran red names both
979
+ const reasons = STOPPED !== null ? [] : [environmentReason(r, counted),
980
+ excludedRed.length > 0 ? `제외한 트리의 시험이 실행됨 — 러너 설정에서도 빼야 함: ${excludedRed.slice(0, 5).join(', ')}` : null,
981
+ outside.length > 0 ? `tests outside the repository ran (${outside.slice(0, 3).join(', ')}) — they are not this commit's evidence` : null].filter((x) => x !== null);
982
+ if (outside.length > 0)
983
+ passed = false;
984
+ for (const reason of reasons)
985
+ environment.push({ ecosystem: eco, dir: u.dir, reason });
986
+ const env = reasons.length > 0 ? reasons.join(' | ') : null;
987
+ unitResults.push({ ecosystem: eco, dir: u.dir, testsRun: counted, passed: r.passed && STOPPED === null, ...(env !== null ? { environment: env } : {}) });
988
+ const label = u.dir === '' ? eco : `${eco} @ ${u.dir}`;
989
+ tails.push(`[${label}] ${r.tail}${STOPPED !== null ? `\n${STOPPED}` : ''}${LEFTOVER !== null ? `\n${LEFTOVER}` : ''}`);
990
+ Object.assign(executedByFile, keyed(u.dir, r.executed));
991
+ for (const f of [...Object.keys(r.executed), ...(r.reported ?? [])])
992
+ reported.add(u.dir === '' ? f : `${u.dir}/${f}`);
993
+ if (env !== null || STOPPED !== null)
994
+ skipUnits.add(`${eco}@${u.dir}`);
842
995
  const adapterOutcomes = r.outcomes;
843
996
  if (adapterOutcomes)
844
- Object.assign(outcomeByFile, adapterOutcomes);
997
+ Object.assign(outcomeByFile, keyed(u.dir, adapterOutcomes));
845
998
  // @implements A-SPEC-137.1 — an adapter whose toolchain is absent reports its files as
846
999
  // unsupported; merge them so a scope that could not run them never reads as fully covered.
847
1000
  const adapterUnsupported = r.unsupported;
@@ -849,14 +1002,30 @@ function runTestScope(scope, cwd) {
849
1002
  unsupported.push(...adapterUnsupported);
850
1003
  }
851
1004
  if (assumedJest) {
852
- tails.push('[assumption] full run with no impacted test files — defaulted to the jest suite');
1005
+ tails.push('[assumption] full run found no test ecosystem in the project — defaulted to the jest suite');
853
1006
  }
854
1007
  if (unsupported.length) {
855
1008
  tails.push(`[unsupported] ${unsupported.length} test file(s) have no runner adapter: ${unsupported.slice(0, 5).join(', ')}`);
856
1009
  }
1010
+ // @implements A-SPEC-128.11 — P6: tests a sparse checkout left out of the work tree cannot run here
1011
+ if (plan.mode === 'full' && sparseTests.length > 0) {
1012
+ passed = false;
1013
+ environment.push({ ecosystem: 'git', dir: '', reason: `sparse checkout — 증빙할 수 없음: ${sparseTests.length} test file(s) outside the work tree, e.g. ${sparseTests.slice(0, 3).join(', ')}` });
1014
+ tails.push(`[sparse] ${sparseTests.length} test file(s) are not in the work tree (skip-worktree): ${sparseTests.slice(0, 5).join(', ')}`);
1015
+ }
1016
+ // @implements A-SPEC-128.10 — I3 and R2, after the run
1017
+ const coverage = plan.mode === 'full' ? coverageAfterRun(expected, codeRoots, reported, executedByFile, skipUnits, buildRoots, testFiles) : null;
1018
+ if (coverage?.notExecuted.length)
1019
+ tails.push(`[not run] ${coverage.notExecuted.length} detected test file(s) did not run: ${coverage.notExecuted.slice(0, 5).join(', ')}`);
1020
+ if (coverage?.untestedCode.length)
1021
+ tails.push(`[untested] code with no executed test under: ${coverage.untestedCode.map((g) => g.root || '(root)').join(', ')}`);
857
1022
  return {
858
1023
  tier: scope.tier, mode: plan.mode, ranFiles: plan.testFiles, passed, skipped: false,
859
- tail: tails.join('\n'), executedByFile, ranWith: ran,
1024
+ tail: tails.join('\n'), executedByFile, ranWith: ran, workers, testsRun, testsRunBy, stopped, detected, failedBy, ...named,
1025
+ ...(plan.mode === 'full' ? { units: unitResults } : {}), ...(environment.length ? { environment } : {}),
1026
+ ...(coverage?.notExecuted.length ? { notExecuted: coverage.notExecuted.slice(0, 200) } : {}),
1027
+ ...(coverage?.untestedCode.length ? { untestedCode: coverage.untestedCode } : {}),
1028
+ workersIgnoredBy: opts.workers === undefined ? [] : ran.filter((eco) => eco !== 'jest'),
860
1029
  ...(Object.keys(outcomeByFile).length ? { outcomeByFile } : {}),
861
1030
  ...(unsupported.length ? { unsupported } : {}),
862
1031
  };
@@ -866,8 +1035,15 @@ function runTestScope(scope, cwd) {
866
1035
  // the no-passWithNoTests rule exists to prevent.
867
1036
  return {
868
1037
  tier: scope.tier, mode: plan.mode, ranFiles: plan.testFiles, passed: false, skipped: false,
869
- tail: `no runner adapter for ${unsupported.length} test file(s): ${unsupported.slice(0, 5).join(', ')}`,
870
- executedByFile: {}, unsupported, ranWith: [],
1038
+ // @implements A-SPEC-128.8 — a full plan with nothing detected at the root says so, not "no adapter for 0 files"
1039
+ tail: surface.length > 0 ? `no unit started — outside the 0.40.x surface:\n${surface.map((s) => s.reason).join('\n')}`
1040
+ : plan.mode === 'full' && unsupported.length === 0
1041
+ ? `no test unit in the project — nothing ran${implementationCount > 0 ? ` (implementation code without tests, e.g. ${implementationSample.join(', ')})` : ''}`
1042
+ : `no runner adapter for ${unsupported.length} test file(s): ${unsupported.slice(0, 5).join(', ')}`,
1043
+ executedByFile: {}, unsupported, ranWith: [], workers: null, testsRun: 0, testsRunBy: {}, stopped: false, detected, ...named,
1044
+ ...(plan.mode === 'full' ? { units: [] } : {}), ...(surface.length ? { environment: surface } : {}),
1045
+ // @implements A-SPEC-128.10 — R2: with no unit, every build root's code is untested (and names a missing build file)
1046
+ ...(plan.mode === 'full' && surface.length === 0 && codeRoots.length ? { untestedCode: codeRoots.map((g) => ({ root: g.root, sample: g.sample, ...(g.missingBuildFile ? { missingBuildFile: g.missingBuildFile } : {}) })) } : {}),
871
1047
  };
872
1048
  }
873
1049
  // @implements A-SPEC-515.1
@@ -933,18 +1109,32 @@ function failedTestSummary(stdout, limit = 5) {
933
1109
  if (j === null || typeof j !== 'object' || !Array.isArray(j.testResults))
934
1110
  return null;
935
1111
  const failures = [];
1112
+ const broken = [];
936
1113
  for (const file of j.testResults) {
937
1114
  const base = typeof file?.name === 'string' ? file.name.split(/[\\/]/).pop() ?? file.name : '(unknown file)';
1115
+ const before = failures.length;
938
1116
  for (const a of Array.isArray(file?.assertionResults) ? file.assertionResults : []) {
939
1117
  if (a?.status !== 'failed')
940
1118
  continue;
941
1119
  const msg = Array.isArray(a.failureMessages) ? a.failureMessages.find((m) => typeof m === 'string' && m.trim() !== '') : undefined;
1120
+ // @implements A-SPEC-466.4 \u2014 one line per name, each part bounded on its own, so one long message pushes no name out
942
1121
  failures.push({
943
- where: `${base} \u203a ${String(a.title ?? '(untitled)')}`,
944
- first: typeof msg === 'string' ? (msg.trim().split('\n')[0] ?? '') : '',
1122
+ where: clip(`${base} \u203a ${String(a.title ?? '(untitled)').replace(/\s+/g, ' ').trim()}`, WHERE_MAX),
1123
+ first: clip(typeof msg === 'string' ? (msg.trim().split('\n')[0] ?? '') : '', FIRST_MAX),
945
1124
  });
946
1125
  }
1126
+ // @implements A-SPEC-466.3 \u2014 a suite that failed with no failing assertion (it did not compile, its teardown threw)
1127
+ // is named too, with the first line of its message
1128
+ // @implements A-SPEC-466.4 \u2014 the cause line: jest opens the message with a `\u25cf Test suite failed to run` header
1129
+ if (file?.status === 'failed' && failures.length === before) {
1130
+ const message = typeof file.message === 'string' ? file.message.replace(/\u001b\[[0-9;]*m/g, '') : '';
1131
+ const cause = message.split('\n').map((l) => l.trim()).find((l) => l !== '' && !l.startsWith('\u25cf')) ?? '';
1132
+ broken.push({ where: clip(`${base} \u203a (suite failed to run)`, WHERE_MAX), first: clip(cause, FIRST_MAX) });
1133
+ }
947
1134
  }
1135
+ // @implements A-SPEC-466.5 \u2014 suites that failed to run come first: in a real mixed run (24 failing assertions, 2
1136
+ // broken suites) the five-name bound was spent on assertions and no broken suite was named
1137
+ failures.unshift(...broken);
948
1138
  if (failures.length === 0)
949
1139
  return null;
950
1140
  const shown = failures.slice(0, Math.max(0, limit));
@@ -954,6 +1144,31 @@ function failedTestSummary(stdout, limit = 5) {
954
1144
  lines.push(` \u2026 \uc678 ${rest}\uac74`);
955
1145
  return lines.join('\n');
956
1146
  }
1147
+ /** @implements A-SPEC-466.4 \u2014 each failure line's own bound: `file \u203a title` and its first message line. */
1148
+ const WHERE_MAX = 200;
1149
+ const FIRST_MAX = 300;
1150
+ /** @implements A-SPEC-466.4 \u2014 `text` cut to at most `max` characters, marked `\u2026`, never inside a surrogate pair. */
1151
+ function clip(text, max) {
1152
+ if (text.length <= max)
1153
+ return text;
1154
+ let end = max - 1;
1155
+ if (/[\uD800-\uDBFF]/.test(text[end - 1] ?? ''))
1156
+ end -= 1;
1157
+ return `${text.slice(0, end)}\u2026`;
1158
+ }
1159
+ /**
1160
+ * @implements A-SPEC-466.3, A-SPEC-466.4 \u2014 the tail of a red jest run. The failing names come first; jest's stderr
1161
+ * follows with a bound of its own. One `tailOf` over both kept the last lines and cut the front, so a long stderr pushed
1162
+ * every name out (2026-11-04, `27bc0c40`: five failures, none named). The name block is bounded by its parts: five names
1163
+ * of at most WHERE_MAX and FIRST_MAX characters and one count line, so the whole tail stays under 7,000 characters.
1164
+ * Without readable jest JSON the tail is what it always was.
1165
+ */
1166
+ function redJestTail(stdout, stderr) {
1167
+ const named = failedTestSummary(stdout);
1168
+ if (named === null)
1169
+ return tailOf(`${stdout}\n${stderr}`, { lines: 6 });
1170
+ return `${named}\n${tailOf(stderr, { lines: 30 })}`;
1171
+ }
957
1172
  function summarizeJestJson(stdout) {
958
1173
  // The payload is not the whole of stdout. A suite that logs anything prints it BEFORE the JSON,
959
1174
  // and this repository's own suite does: measured, the first live run after this function landed