dsh-vibe-math 2.7.2 → 2.8.0

package/tests/v2-list-agents-and-next-step.mutants.mjs

@@ -0,0 +1,106 @@
+#!/usr/bin/env node
+/**
+ * Mutant harness for the two v2 harness-gap rows (ships: `tests/*.mutants.mjs`):
+ *   F3  v2 `vibe_math_list_agents` — the tool must EXIST and be MENTIONED
+ *   F4  v2 require-gate — a refusal must carry an EXECUTABLE NEXT STEP
+ *
+ * Each case copies the v2 preset (and, for the persona case, all four preset dirs) into a private root,
+ * applies ONE logical mutation, drives the SHIPPED owning suite through its own seam, and requires a NAMED
+ * red. Every target suite also runs GREEN on the unmutated copies first (faithful-copy baseline), so the
+ * mutation is the only red cause. Seams used: MC_V2_PLUGIN, V2_PLUGIN, PERSONA_ROOT.
+ *
+ * Run: node tests/v2-list-agents-and-next-step.mutants.mjs
+ */
+import { mkdtempSync, cpSync, copyFileSync, readFileSync, writeFileSync, rmSync } from 'node:fs'
+import { tmpdir } from 'node:os'
+import { join } from 'node:path'
+import { spawnSync } from 'node:child_process'
+import { fileURLToPath } from 'node:url'
+
+const HERE = fileURLToPath(new URL('./', import.meta.url))
+const REPO = join(HERE, '..')
+const PRESETS = ['vibe-math-v2', 'vibe-math-v3', 'vibe-math-v4', 'vibe-math-v5']
+let passed = 0, failed = 0
+const ok = (c, l, d) => { if (c) { passed++; console.log('  ok   ' + l) } else { failed++; console.log('  FAIL ' + l + (d ? ' — ' + d : '')) } }
+
+function runSuite(file, env) {
+  const r = spawnSync(process.execPath, [join(HERE, file)], { cwd: REPO, encoding: 'utf8', maxBuffer: 1 << 26, env: Object.assign({}, process.env, env || {}) })
+  const o = String(r.stdout || '') + String(r.stderr || '')
+  return {
+    status: r.status, out: o,
+    failLines: o.split('\n').filter((l) => /^\s*(FAIL|  FAIL)\b/.test(l)),
+    sum: (o.split('\n').filter((l) => /passed, \d+ failed/.test(l)).slice(-1)[0] || '').trim(),
+  }
+}
+const named = (r, needle) => r.failLines.some((l) => l.indexOf(needle) !== -1)
+const V2_SRC = join(REPO, 'vibe-math-v2', 'vibe-math-v2.js')
+const ORIGIN = readFileSync(V2_SRC, 'utf8')
+const REGISTER = "registerTool('vibe_math_list_agents'"
+const NEXT_STEP = '下一步：把该对象形式化到 Lean 通过'
+const COPY_AS = 'vibe_math_list_agentsX'
+
+// ---------- baselines: the shipped suites are green on the COPIED artifacts ----------
+const baseRoot = mkdtempSync(join(tmpdir(), 'v2gaps-base-'))
+cpSync(join(REPO, 'vibe-math-v2'), join(baseRoot, 'vibe-math-v2'), { recursive: true })
+{
+  const a = runSuite('math-computation-v2.test.mjs', { MC_V2_PLUGIN: join(baseRoot, 'vibe-math-v2', 'vibe-math-v2.js') })
+  ok(a.status === 0, 'baseline: math-computation-v2 is green on the COPIED v2 preset (MC_V2_PLUGIN)', a.sum.slice(0, 60))
+  const c = runSuite('v2-fix-probes.test.mjs', { V2_PLUGIN: join(baseRoot, 'vibe-math-v2', 'vibe-math-v2.js') })
+  ok(c.status === 0, 'baseline: v2-fix-probes is green on the COPIED v2 preset (V2_PLUGIN)', c.sum.slice(0, 60))
+}
+{
+  const pRoot = mkdtempSync(join(tmpdir(), 'v2gaps-persona-'))
+  for (const d of PRESETS) cpSync(join(REPO, d), join(pRoot, d), { recursive: true })
+  const p = runSuite('audit-persona-surface.test.mjs', { PERSONA_ROOT: pRoot })
+  ok(p.status === 0, 'baseline: audit-persona-surface is green on the COPIED preset tree (PERSONA_ROOT)', p.sum.slice(0, 60))
+  rmSync(pRoot, { recursive: true, force: true })
+}
+rmSync(baseRoot, { recursive: true, force: true })
+
+// ---------- F3a: the tool must EXIST (rename the registration -> the documented call fails) ----------
+{
+  const dir = mkdtempSync(join(tmpdir(), 'v2gaps-exist-'))
+  cpSync(join(REPO, 'vibe-math-v2'), join(dir, 'vibe-math-v2'), { recursive: true })
+  const p = join(dir, 'vibe-math-v2', 'vibe-math-v2.js')
+  const src = readFileSync(p, 'utf8')
+  writeFileSync(p, src.replace(REGISTER, REGISTER.replace('list_agents', 'list_agentsX')))
+  const r = runSuite('math-computation-v2.test.mjs', { MC_V2_PLUGIN: p })
+  ok(r.status !== 0 && named(r, '对照：vibe_math_list_agents 可调用'),
+    '★ F3 (exists): renaming the registration reddens the named "list_agents is callable" assertion (unit: tool name)',
+    r.sum.slice(0, 70))
+  if (!named(r, '对照：vibe_math_list_agents 可调用')) for (const l of r.failLines.slice(0, 3)) console.log('        ' + l.trim().slice(0, 130))
+  rmSync(dir, { recursive: true, force: true })
+}
+// ---------- F3b: the tool must be MENTIONED (persona still names the old name) ----------
+{
+  const root = mkdtempSync(join(tmpdir(), 'v2gaps-mention-'))
+  for (const d of PRESETS) cpSync(join(REPO, d), join(root, d), { recursive: true })
+  const p = join(root, 'vibe-math-v2', 'vibe-math-v2.js')
+  writeFileSync(p, readFileSync(p, 'utf8').split(REGISTER).join(REGISTER.replace('list_agents', 'list_agentsX')))
+  const r = runSuite('audit-persona-surface.test.mjs', { PERSONA_ROOT: root })
+  ok(r.status !== 0 && named(r, 'every registered tool is mentioned in the persona, or explicitly allow-listed'),
+    '★ F3 (mentioned): a registered tool the persona never names reddens the named allow-list assertion (unit: registered tool names)',
+    r.sum.slice(0, 70))
+  if (!named(r, 'every registered tool is mentioned in the persona, or explicitly allow-listed')) for (const l of r.failLines.slice(0, 3)) console.log('        ' + l.trim().slice(0, 130))
+  rmSync(root, { recursive: true, force: true })
+}
+// ---------- F4: a require-gate refusal must carry an executable NEXT STEP ----------
+{
+  const dir = mkdtempSync(join(tmpdir(), 'v2gaps-next-'))
+  cpSync(join(REPO, 'vibe-math-v2'), join(dir, 'vibe-math-v2'), { recursive: true })
+  const p = join(dir, 'vibe-math-v2', 'vibe-math-v2.js')
+  const src = readFileSync(p, 'utf8')
+  if (src.indexOf(NEXT_STEP) === -1) ok(false, 'F4: the single-site next-step anchor applies', 'ANCHOR MISS')
+  else {
+    writeFileSync(p, src.replace(NEXT_STEP, '下一步：见文档'))
+    const r = runSuite('v2-fix-probes.test.mjs', { V2_PLUGIN: p })
+    ok(r.status !== 0 && named(r, 'require-gate 反馈行携带下一步'),
+      '★ F4: dropping the executable next step from the require-gate refusal reddens its named assertion (unit: refusal sites)',
+      r.sum.slice(0, 70))
+    if (!named(r, 'require-gate 反馈行携带下一步')) for (const l of r.failLines.slice(0, 3)) console.log('        ' + l.trim().slice(0, 130))
+  }
+  rmSync(dir, { recursive: true, force: true })
+}
+console.log('')
+console.log('=== V2 LIST-AGENTS / NEXT-STEP MUTANTS: ' + passed + ' passed, ' + failed + ' failed ===')
+process.exit(failed === 0 ? 0 : 1)

package/tests/v2v3-interrupt.mutants.mjs

@@ -0,0 +1,111 @@
+// Registered mutant harness for the v2/v3 math-computation suites (task-14 / SLV probe P3): the suites'
+// §19 D1 section already covers "unknown child" and "empty childId" — this family makes the missing one
+// executable: the "host REFUSES to interrupt" branch must keep its NAMED code and its actionable `next`.
+//
+// Measured traps reused from tests/math-computation-v4.mutants.mjs:
+//   · the seam must point at the PLUGIN file (`MC_V2_PLUGIN` / `MC_V3_PLUGIN`), and the copied graph must
+//     include the plugin's local dependencies, or the suite dies before reaching any assertion;
+//   · an anchor that matches a different number of times than expected is a SETUP failure, not a red.
+//
+// Usage: node tests/v2v3-interrupt.mutants.mjs
+import { readFileSync, writeFileSync, mkdirSync, copyFileSync, rmSync, readdirSync } from 'node:fs'
+import { execFileSync } from 'node:child_process'
+import { dirname, join } from 'node:path'
+import { tmpdir } from 'node:os'
+import { fileURLToPath } from 'node:url'
+
+const HERE = dirname(fileURLToPath(import.meta.url))
+const REPO = join(HERE, '..')
+const CHILD_TIMEOUT_MS = Number(process.env.MUTANT_CHILD_TIMEOUT_MS || 300000)
+const TIMES = []
+const hangs = []
+const skipped = []
+
+function copyGraph(preset, file, dest) {
+  const src = join(REPO, preset, file)
+  copyFileSync(src, join(dest, file))
+  for (const m of readFileSync(src, 'utf8').matchAll(/from\s+'(\.\/[A-Za-z0-9_.-]+\.js)'/g)) {
+    const dep = m[1].slice(2)
+    if (!readdirSync(dest).includes(dep)) copyGraph(preset, dep, dest)
+  }
+}
+
+function runFamily(f) {
+  const dest = join(tmpdir(), f.preset.replace(/[^a-z0-9]/gi, '') + '-mut-' + Math.random().toString(36).slice(2, 10))
+  mkdirSync(dest, { recursive: true })
+  copyGraph(f.preset, f.main, dest)
+  const target = join(dest, f.main)
+  // task-7 discipline: anchor matching is done on an EOL-normalised copy, and a leaked CR is a SETUP
+  // failure (a lone CR would silently change the anchor's match count).
+  let before = readFileSync(target, 'utf8')
+  if (/\r/.test(before.replace(/\r\n/g, '\n'))) {
+    console.error('SETUP-FAIL - the anchor source was not EOL-normalised (a lone CR leaked in)')
+    rmSync(dest, { recursive: true, force: true })
+    process.exit(1)
+  }
+  const n = before.split(f.from).length - 1
+  if (n !== 1) {
+    skipped.push(f.name + ' (anchor x' + n + ')')
+    console.error('  SKIP - ' + f.name + ' (anchor x' + n + ', expected exactly 1)')
+    rmSync(dest, { recursive: true, force: true })
+    return false
+  }
+  writeFileSync(target, before.replace(f.from, f.to), 'utf8')
+  try { execFileSync(process.execPath, ['--check', target], { stdio: 'pipe' }) }
+  catch (e) {
+    writeFileSync(target, before, 'utf8')
+    skipped.push(f.name + ' (mutation did not compile)')
+    console.error('  SKIP - ' + f.name + ' (mutation did not compile; restored)')
+    rmSync(dest, { recursive: true, force: true })
+    return false
+  }
+  const t0 = Date.now()
+  let out = ''
+  let code = 0
+  let hang = false
+  try {
+    out = execFileSync(process.execPath, [f.suite], {
+      cwd: REPO, encoding: 'utf8', timeout: CHILD_TIMEOUT_MS, killSignal: 'SIGKILL',
+      env: Object.assign({}, process.env, { [f.env]: join(dest, f.main) }),
+    })
+  } catch (e) {
+    if (e && (e.killed || e.signal === 'SIGKILL')) hang = true
+    code = (e && e.status) || 1
+    out = String((e && e.stdout) || '') + String((e && e.stderr) || '')
+  }
+  const ms = Date.now() - t0
+  TIMES.push([f.name, ms])
+  if (hang) hangs.push(f.name + '(' + Math.round(ms / 1000) + 's)')
+  const named = out.split(/\r?\n/).filter((l) => /^\s*(FAIL - | {2}- )/.test(l) && f.expect.test(l)).map((l) => l.trim())
+  const ok = !hang && code !== 0 && named.length > 0
+  console.log((hang ? '  HANG - ' : (ok ? '  ok - ' : '  FAIL - ')) + f.name + ' [' + ms + 'ms]' + (named.length ? ' :: ' + named[0].slice(0, 150) : ' :: no named red (exit=' + code + ')'))
+  rmSync(dest, { recursive: true, force: true })
+  return ok
+}
+
+const FAMILIES = [
+  {
+    name: 'P3/v2: the host-refused-interrupt code is renamed (the named failure disappears)',
+    preset: 'vibe-math-v2', main: 'vibe-math-v2.js',
+    suite: 'tests/math-computation-v2.test.mjs', env: 'MC_V2_PLUGIN',
+    from: 'VIBE_MATH_INTERRUPT_FAILED', to: 'VIBE_MATH_INTERRUPT_RENAMED',
+    expect: /\[P3\/v2\] 宿主拒绝中断/,
+  },
+  {
+    name: 'P3/v3: the host-refused-interrupt code is renamed (the named failure disappears)',
+    preset: 'vibe-math-v3', main: 'vibe-math-v3.js',
+    suite: 'tests/math-computation-v3.test.mjs', env: 'MC_V3_PLUGIN',
+    from: 'VIBE_MATH_INTERRUPT_FAILED', to: 'VIBE_MATH_INTERRUPT_RENAMED',
+    expect: /\[P3\/v3\] 宿主拒绝中断/,
+  },
+]
+
+let red = 0
+for (const f of FAMILIES) { if (runFamily(f)) red++ }
+console.log('')
+console.log('mutant families reddening the v2/v3 host-refused-interrupt guard by name: ' + red + '/' + FAMILIES.length)
+console.log('skipped=[' + skipped.join(' | ') + ']')
+console.log('timings: ' + TIMES.map((t) => String(t[0]).split(':')[0] + '=' + t[1] + 'ms').join('  '))
+console.log('hangs=[' + hangs.join(' | ') + ']')
+if (red !== FAMILIES.length || skipped.length || hangs.length) process.exit(1)
+console.log('ALL MUTANTS RED AS REQUIRED')

package/tests/v3-fix-probes.mutants.mjs

@@ -0,0 +1,99 @@
+// Registered mutant harness for tests/v3-fix-probes.test.mjs — the probe carries its own proof.
+// Defect class covered: silent-failure surfacing (F6a/F6b/F6c), cap sharing (F2cap), declared asymmetry (F-A),
+// fence-root drift (F5), shell-fallback silence (F4c). Every family is a SINGLE content-anchored mutation
+// of the v3 preset (no line numbers) and the suite runs through its V3_PLUGIN seam.
+// Exit criterion (audit-path-discipline.mutants.mjs paradigm): every family must redden BY NAME;
+// ALL MUTANTS RED AS REQUIRED + exit 0 only when all do.
+// Run: node tests/v3-fix-probes.mutants.mjs
+import { readFileSync, writeFileSync, mkdirSync, copyFileSync, rmSync, readdirSync } from 'node:fs'
+import { execFileSync } from 'node:child_process'
+import { dirname, join } from 'node:path'
+import { tmpdir } from 'node:os'
+import { fileURLToPath } from 'node:url'
+const HERE = dirname(fileURLToPath(import.meta.url))
+const REPO = join(HERE, '..')
+const PRESET = 'vibe-math-v3'
+const SUITE = 'tests/v3-fix-probes.test.mjs'
+const ENV = 'V3_PLUGIN'
+// Bound every child run: run-tests has no per-suite timeout, so an unbounded child hangs the whole gate.
+// A timeout is reported as HANG, counts as a failure for that family, and is listed in hangs=[].
+const CHILD_TIMEOUT_MS = Number(process.env.MUTANT_CHILD_TIMEOUT_MS || 120000)
+const TIMES = []
+// R15: whole-process wall time, printed at the end (the timeout row quotes this line).
+const START_MS = Date.now()
+const hangs = []
+let lastHang = false
+function copyGraph(file, dest) {
+  const src = join(REPO, PRESET, file)
+  copyFileSync(src, join(dest, file))
+  for (const m of readFileSync(src, 'utf8').matchAll(/from\s+'(\.\/[A-Za-z0-9_.-]+\.js)'/g)) {
+    const dep = m[1].slice(2)
+    if (!readdirSync(dest).includes(dep)) copyGraph(dep, dest)
+  }
+}
+function runFamily(f) {
+  const dest = join(tmpdir(), 'v3probes-mut-' + Math.random().toString(36).slice(2, 10))
+  mkdirSync(dest, { recursive: true })
+  copyGraph(PRESET + '.js', dest)
+  const target = join(dest, f.editFile || (PRESET + '.js'))
+  const before = readFileSync(target, 'utf8')
+  const n = before.split(f.from).length - 1
+  if (f.lastOnly ? n < 1 : n !== 1) { console.error('  NO-OP - ' + f.name + ' (anchor x' + n + (f.lastOnly ? ', lastOnly' : '') + ')'); rmSync(dest, { recursive: true, force: true }); return false }
+  const mutated = f.lastOnly
+    ? (() => { const i = before.lastIndexOf(f.from); return before.slice(0, i) + f.to + before.slice(i + f.from.length) })()
+    : before.split(f.from).join(f.to)
+  writeFileSync(target, mutated, 'utf8')
+  try { execFileSync(process.execPath, ['--check', target], { stdio: 'pipe' }) }
+  catch (e) { writeFileSync(target, before, 'utf8'); console.error('  FAIL - ' + f.name + ' (mutation did not compile; restored)'); rmSync(dest, { recursive: true, force: true }); return false }
+  let out = ''
+  let code = 0
+  try { out = execFileSync(process.execPath, [SUITE], { cwd: REPO, encoding: 'utf8', timeout: CHILD_TIMEOUT_MS, killSignal: 'SIGKILL', env: Object.assign({}, process.env, { [ENV]: join(dest, PRESET + '.js') }) }) }
+  catch (e) { if (e && (e.killed || e.signal === 'SIGKILL')) lastHang = true; code = (e && e.status) || 1; out = String((e && e.stdout) || '') + String((e && e.stderr) || '') }
+  const named = out.split(/\r?\n/).filter((l) => /^\s*(FAIL - | {2}- )/.test(l) && f.expect.test(l)).map((l) => l.trim())
+  const ok = !lastHang && code !== 0 && named.length > 0
+  console.log((lastHang ? '  HANG - ' : (ok ? '  ok - ' : '  FAIL - ')) + f.name + (named.length ? ' :: ' + named[0].slice(0, 140) : ' :: no named red (exit=' + code + ')'))
+  rmSync(dest, { recursive: true, force: true })
+  return ok
+}
+const FAMILIES = [
+  { name: 'F2cap: status site reverted to a literal 10', expect: /\[F2cap\] status 与 report 两处都走同一常量/,
+    from: "recentActivity: activityLog.slice(-Math.min(ACTIVITY_REPORT_MAX, Number(params.activityLogCap) || 100)), params: params,",
+    to: "recentActivity: activityLog.slice(-Math.min(10, Number(params.activityLogCap) || 100)), params: params," },
+  { name: 'F2cap: param description hard-codes the number', expect: /\[F2cap\] 参数描述\*\*引用\*\*该常量/,
+    from: "' + ACTIVITY_REPORT_MAX + ' 条）', suggestion: 100 },", to: "' + 30 + ' 条）', suggestion: 100 }," },
+  { name: 'F-A: status gains pendingDecisionItems', expect: /\[F-A\] status \*\*不得\*\*携带明细/,
+    from: "pendingDecisions: decisionQueue.filter(function (d) { return d.status === 'pending' }).length,",
+    to: "pendingDecisions: decisionQueue.filter(function (d) { return d.status === 'pending' }).length,\n  pendingDecisionItems: decisionQueue.filter(function (d) { return d.status === 'pending' }).map(function (d) { return { id: d.id, node: d.node, context: d.context } }),",
+    lastOnly: true },
+  { name: 'F5: fence-root comparison removed', expect: /\[F5\/v3\] 围栏根漂移/,
+    from: 'if (actual && expected && actual !== expected) warnFenceDriftOnce(actual, expected)', to: 'void actual' },
+  { name: 'F4c: ensureDirs failure warning removed', expect: /\[F4c\] ensureDirs 失败不是沉默/,
+    from: "if (!r || !r.ok) warnShellOnce('ensureDirs (mkdir ' + base + ')', r)", to: 'void r' },
+  { name: 'F6a: provider list-failure warning removed', expect: /\[F6a\/v3\]/,
+    from: "warnProviderFallback('subagents.list() failed: ' + ((e && e.message) || e))", to: 'void e' },
+  { name: 'F6b: interrupt emit statement deleted (anchored by content)', expect: /\[F6b\/v3\]/,
+    from: "if (r && r.ok === false) logActivity('interrupt', '中断失败（' + why + '）：' + String(cid) + ' — ' + String(r.message || r.code || ''))", to: 'void r' },
+  { name: 'F6c: project-pointer write failure re-silenced', expect: /\[F6c\/v3\]/,
+    from: "noteStateWriteFailure('current.' + safeId(sessionId) + '.json', (e && e.message) || e)", to: 'void e' },
+]
+let red = 0
+FAMILIES.push(
+  { name: 'F-B: derivation drops the source descriptions', expect: /\[F-B\/v3\] 每个机读描述都来自真源且非空/,
+    from: 'if (e.description) p.description = e.description', to: 'void e /* MUTANT F-B: no source description */' },
+  { name: 'F-B: one exception entry deleted', expect: /\[F-B\/v3\] 例外集合 == docs\/parameter-schema\.md/,
+    from: "mode: { type: 'string', enum: ['manual', 'auto'] },", to: '' },
+  { name: 'F-B: one registration site reverted to a literal', expect: /\[F-B\/v3\] 两处注册点都由 paramProps\(\)/,
+    from: "objParams(paramProps()), 'vibe_math_set_params')", to: "objParams({ mode: { type: 'string', enum: ['manual', 'auto'] } }), 'vibe_math_set_params')" },
+)
+FAMILIES.push(
+  { name: 'task-9/v3: the not-detected log drops the actionable sentence', expect: /\[task-9\/v3\]/,
+    from: '已探测 PATH 与文档化的常见 TeX 根；可用 paperLatexCommand 指定绝对路径。', to: '' },
+)
+for (const f of FAMILIES) { const t0 = Date.now(); lastHang = false; const ok = runFamily(f); const ms = Date.now() - t0; TIMES.push([f.name, ms]); if (lastHang) hangs.push(f.name + '(' + Math.round(ms / 1000) + 's)'); if (ok) red++ }
+console.log('')
+console.log('mutant families reddening the v3 probe by name: ' + red + '/' + FAMILIES.length)
+console.log('timings: ' + TIMES.map((t) => String(t[0]).split(':')[0] + '=' + t[1] + 'ms').join('  '))
+console.log('TOTAL WALL TIME (all families + setup): ' + (Date.now() - START_MS) + 'ms (' + Math.round((Date.now() - START_MS) / 1000) + 's)')
+console.log('hangs=[' + hangs.join(' | ') + ']')
+if (red !== FAMILIES.length || hangs.length) process.exit(1)
+console.log('ALL MUTANTS RED AS REQUIRED')