dsh-vibe-math 2.7.2 → 2.8.0

package/tests/selfdrive-v5.mutants.mjs

@@ -0,0 +1,132 @@
+// Registered mutant harness for the v5 self-drive suite — the suite carries its own proof.
+// Mirrors tests/formal-verify-v4.mutants.mjs / formal-verify-v3.mutants.mjs: a SINGLE content-anchored
+// mutation of the v5 preset (no line numbers), run through the V5_PLUGIN seam, bounded child runs (a
+// timeout is a HANG, never a pass), skipped families printed and counted, and the family's TOTAL WALL
+// TIME reported (per-family numbers alone can hide a timeout).
+// Run: node tests/selfdrive-v5.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-v5'
+const MAIN = PRESET + '.js'
+const SUITE = 'tests/selfdrive-v5.mjs'
+const ENV = 'V5_PLUGIN'
+// Measured: a selfdrive-v5 child run is ~2 s; MUTANT_CHILD_TIMEOUT_MS overrides it.
+const CHILD_TIMEOUT_MS = Number(process.env.MUTANT_CHILD_TIMEOUT_MS || 300000)
+const TIMES = []
+const hangs = []
+const skipped = []
+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(), 'v5sd-mut-' + Math.random().toString(36).slice(2, 10))
+  mkdirSync(dest, { recursive: true })
+  copyGraph(MAIN, dest)
+  const target = join(dest, f.editFile || MAIN)
+  const before = readFileSync(target, 'utf8')
+  const n = before.split(f.from).length - 1
+  const okCount = f.lastOnly || f.all ? n >= 1 : n === 1
+  if (!okCount) {
+    skipped.push(f.name + ' (anchor x' + n + ')')
+    console.error('  SKIP - ' + f.name + ' (anchor x' + n + ')')
+    rmSync(dest, { recursive: true, force: true })
+    return false
+  }
+  writeFileSync(target, before.split(f.from).join(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, [SUITE], {
+      cwd: REPO, encoding: 'utf8', timeout: CHILD_TIMEOUT_MS, killSignal: 'SIGKILL',
+      env: Object.assign({}, process.env, { [ENV]: join(dest, 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 = [
+  {
+    // V5-A1: the LAST voter is replaced by a DUPLICATE of the first, so P stays 4 (the count assertion
+    // passes) while the voter id SET loses a member - the roster identity, not the head-count.
+    name: 'V5-A1: the last voter is a duplicate of the first (P still 4)',
+    from: "const voters = () => activeMembers().filter((m) => m.kind === 'academician' || m.kind === 'researcher')",
+    to: "const voters = () => { const v = activeMembers().filter((m) => m.kind === 'academician' || m.kind === 'researcher'); return v.length > 1 ? v.slice(0, -1).concat([v[0]]) : v }",
+    expect: /V5-A1 the voter id SET is exactly/,
+  },
+  {
+    // F6 tightening (roster): the published voter set drops one member while its COUNT stays put.
+    name: 'V5-A2: the published voter set disagrees with its own count',
+    from: 'voters: vs.map((m) => m.id),',
+    to: 'voters: vs.slice(1).map((m) => m.id),',
+    expect: /V5-A2 the published voter SET agrees with its own count/,
+  },
+  {
+    // F6 tightening (staff): the member persona ignores `params.staffPersona` again.
+    name: 'V5-A3: the staff persona never reaches the prompts',
+    from: "const extra = String(params.staffPersona || '').trim()",
+    to: "const extra = ''",
+    expect: /V5-A3 the staff persona reaches the member persona\/prompts/,
+  },
+  {
+    // ④ CAS: the stale-revision guard is disabled, so a stale update lands (the return-value
+    // assertions AND the new "unchanged after refusal" assertion redden).
+    // R17 scope refinement for V5-A4: it detects any refused-but-written change VISIBLE IN THE
+    // OBSERVABLE TRIPLE (revision|ownerId|status) - measured: an in-branch `task.status='done'`
+    // makes it RED (`before=1||pending after=1||done`), while a write OUTSIDE the triple (e.g.
+    // `task.notes`) keeps it green and remains a MEASURED BOUNDARY (invisible to the harness).
+    // assertions AND the new "unchanged after refusal" assertion redden).
+    name: 'V5-A4: the stale-revision guard is disabled (a stale CAS lands)',
+    from: 'if (expected !== task.revision) {',
+    to: 'if (false) {',
+    expect: /V5-A4 a REFUSED stale CAS leaves the task UNCHANGED|a stale revision is refused/,
+  },
+  {
+    // ⑤ path contract: ONLY the documented composition moves (instRootRel), while the framework
+    // keeps writing through instRoot() - exactly the v4 58fff7f defect class this guard exists for.
+    name: 'V5-A5: the documented member root diverges from the written root',
+    from: "const instRootRel = () => 'VibeMath/Projects/' + project + '/Institutes/' + instituteName",
+    to: "const instRootRel = () => 'VibeMath/Projects/' + project + '/Institutes/' + instituteName + 'X'",
+    expect: /V5-A5 the documented member root equals the root the framework writes/
+  },
+]
+let red = 0
+for (const f of FAMILIES) { const ok = runFamily(f); if (ok) red++ }
+const totalMs = TIMES.reduce((a, t) => a + t[1], 0)
+console.log('')
+console.log('mutant families reddening the v5 self-drive suite by name: ' + red + '/' + FAMILIES.length)
+console.log('timings (per family): ' + TIMES.map((t) => String(t[0]).split(':')[0] + '=' + t[1] + 'ms').join('  '))
+console.log('TOTAL WALL TIME (all families + setup): ' + totalMs + 'ms (' + Math.round(totalMs / 1000) + 's)')
+console.log('hangs=[' + hangs.join(' | ') + ']')
+console.log('skipped=[' + skipped.join(' | ') + ']')
+if (red !== FAMILIES.length || skipped.length || hangs.length) process.exit(1)
+console.log('ALL MUTANTS RED AS REQUIRED')

package/tests/temp-hygiene.mutants.mjs

@@ -0,0 +1,80 @@
+// Falsifiability driver for scripts/clean-temp.mjs (task-12): a SINGLE-SITE edit in the sweeper must
+// turn its own guard (`--self-test`) NAMED-red, while the unmutated control stays green. The mutated
+// copy is written as a DOT-SIBLING inside scripts/ so the module still resolves REPO/tests correctly
+// (a copy in %TEMP% would fail the prefix-extraction check for the wrong reason), and it is always
+// removed in `finally`.
+import { spawnSync } from 'node:child_process'
+import { readFileSync, rmSync, writeFileSync } from 'node:fs'
+import { dirname, join } from 'node:path'
+import { fileURLToPath } from 'node:url'
+
+const HERE = dirname(fileURLToPath(import.meta.url))
+const REPO = join(HERE, '..')
+const SRC = join(REPO, 'scripts', 'clean-temp.mjs')
+
+const FAMILIES = [
+  {
+    // The age threshold is the whole safety property: without it every suite dir looks deletable.
+    name: 'the age threshold is dropped (a FRESH suite dir starts looking stale)',
+    from: '.filter((e) => Number(e.mtimeMs) < cut)',
+    to: '.filter(() => true)',
+    expect: /a FRESH suite dir is never swept/,
+  },
+  {
+    // The prefix test keeps the sweeper from touching ANY top-level directory of the temp root.
+    name: 'the suite-prefix test is dropped (any top-level dir starts looking like ours)',
+    from: 'prefixes.some((p) => String(e.name).startsWith(p))',
+    to: 'true',
+    expect: /only exact suite prefixes are swept/,
+  },
+  {
+    // task-15: "parent is gone" is the safety property - dropping it makes OUR OWN running gate look like
+    // a leftover, which is exactly the false positive that would make this tool dangerous to auto-run.
+    name: 'the parent-gone test is dropped (a live-parented gate looks like a leftover)',
+    from: 'return e.parentGone === true',
+    to: 'return true',
+    expect: /a process with a LIVE parent .* is never called a leftover/,
+  },
+  {
+    // task-15: the command-line gate is what keeps unrelated programs (editors, other stacks) out.
+    name: 'the command-line gate is dropped (any orphaned program looks like ours)',
+    from: "if (!patterns.some((re) => re.test(String(e.cmd || '')))) return false",
+    to: 'if (false) return false',
+    expect: /an unrelated program is never reported/,
+  },
+]
+
+function runSelftest(scriptPath) {
+  const r = spawnSync(process.execPath, [scriptPath, '--self-test'], { encoding: 'utf8', timeout: 60000, cwd: REPO })
+  return { code: r.status, out: String(r.stdout || '') + String(r.stderr || '') }
+}
+
+const control = runSelftest(SRC)
+console.log('control (unmutated sweeper): exit=' + control.code + (control.code === 0 ? '  green, as required' : '  NOT GREEN'))
+if (control.code !== 0) console.log(control.out.split(/\r?\n/).filter((l) => /FAIL/.test(l)).slice(0, 3).join('\n'))
+
+let red = 0
+for (const f of FAMILIES) {
+  const dest = join(REPO, 'scripts', '.tmp-hygiene-' + Math.random().toString(36).slice(2, 10) + '.mjs')
+  let ok = false
+  try {
+    let text = readFileSync(SRC, 'utf8').replace(/\r\n?/g, '\n')
+    if (/\r/.test(text)) { console.error('SETUP-FAIL - the anchor source was not EOL-normalised (CR leaked into anchor matching)'); process.exit(1) }
+    const hits = text.split(f.from).length - 1
+    if (hits !== 1) { console.error('SETUP-FAIL - anchor ' + JSON.stringify(f.from) + ' matched ' + hits + ' time(s), expected exactly 1'); process.exit(1) }
+    writeFileSync(dest, text.replace(f.from, f.to), 'utf8')
+    const r = runSelftest(dest)
+    const named = r.out.split(/\r?\n/).filter((l) => /^\s*FAIL - /.test(l) && f.expect.test(l)).map((l) => l.trim())
+    ok = r.code !== 0 && named.length > 0
+    console.log((ok ? '  ok - ' : '  FAIL - ') + f.name + ' [exit=' + r.code + ']' + (named.length ? ' :: ' + named[0].slice(0, 140) : ' :: no named red (exit=' + r.code + ')'))
+    if (ok) red++
+  } finally {
+    try { rmSync(dest, { force: true }) } catch (e) { /* best effort */ }
+  }
+}
+
+console.log('')
+console.log('mutant families reddening the temp-hygiene guard by name: ' + red + '/' + FAMILIES.length)
+console.log('control green: ' + (control.code === 0))
+if (red !== FAMILIES.length || control.code !== 0) process.exit(1)
+console.log('ALL MUTANTS RED AS REQUIRED')

package/tests/v2-fix-probes.mutants.mjs

@@ -0,0 +1,104 @@
+// Registered mutant harness for tests/v2-fix-probes.test.mjs — the probe carries its own proof.
+// Defect classes: surface-shape contract (F1), declared asymmetry (F-A), path-base mixing (F6),
+// silent-failure surfacing (F6a/F6b/F6c). Single content-anchored mutations; V2_PLUGIN seam.
+// Exit criterion: every family reddens BY NAME; any skipped family is printed and counted (never a
+// silent NO-OP impersonating a pass); ALL MUTANTS RED AS REQUIRED + exit 0 only then.
+// Run: node tests/v2-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'
+
+// Bound every child run: run-tests has no per-suite timeout, so an unbounded child hangs the 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
+const HERE = dirname(fileURLToPath(import.meta.url))
+const REPO = join(HERE, '..')
+const PRESET = 'vibe-math-v2'
+const MAIN = PRESET + '.js'
+const SUITE = 'tests/v2-fix-probes.test.mjs'
+const ENV = 'V2_PLUGIN'
+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)
+  }
+}
+const skipped = []
+function runFamily(f) {
+  const dest = join(tmpdir(), 'v2probes-mut-' + Math.random().toString(36).slice(2, 10))
+  mkdirSync(dest, { recursive: true })
+  copyGraph(MAIN, dest)
+  const target = join(dest, f.editFile || MAIN)
+  const before = readFileSync(target, 'utf8')
+  const n = before.split(f.from).length - 1
+  const okCount = f.lastOnly || f.firstOnly || f.all ? n >= 1 : n === 1
+  if (!okCount) { skipped.push(f.name + ' (anchor x' + n + ')'); console.error('  SKIP - ' + f.name + ' (anchor x' + n + ')'); rmSync(dest, { recursive: true, force: true }); return false }
+  let mutated
+  if (f.lastOnly) { const i = before.lastIndexOf(f.from); mutated = before.slice(0, i) + f.to + before.slice(i + f.from.length) }
+  else if (f.firstOnly) { const i = before.indexOf(f.from); mutated = before.slice(0, i) + f.to + before.slice(i + f.from.length) }
+  else if (f.all) mutated = before.split(f.from).join(f.to)
+  else mutated = 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'); 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 }
+  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, MAIN) }) }) }
+  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 = code !== 0 && named.length > 0
+  console.log((ok ? '  ok - ' : '  FAIL - ') + f.name + (named.length ? ' :: ' + named[0].slice(0, 130) : ' :: no named red (exit=' + code + ')'))
+  rmSync(dest, { recursive: true, force: true })
+  return ok
+}
+const ANCHOR_LEN = "pendingDecisions: decisionQueue.filter(function (d) { return d.status === 'pending' }).length,"
+const FAMILIES = [
+  { name: 'F1: report pendingDecisions back to an array', expect: /\[F1\] pendingDecisions 两面同形/, from: ANCHOR_LEN, firstOnly: true,
+    to: "pendingDecisions: decisionQueue.filter(function (d) { return d.status === 'pending' }).map(function (d) { return { id: d.id, node: d.node, context: d.context } })," },
+  { name: 'F-A: status gains pendingDecisionItems', expect: /\[F-A\] status \*\*不得\*\*携带明细/, from: ANCHOR_LEN,
+    to: ANCHOR_LEN + "\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: 'F6: paths base dropped from both surfaces', expect: /\[F6\] status\.formal\.paths 带 base \+ note/,
+    from: "paths: { base: frameworkRoot() + '/', project: 'Formal/', proofs: 'Verified/Lean/', lib:",
+    to: "paths: { project: 'Formal/', proofs: 'Verified/Lean/', lib:", all: true },
+  { name: 'F6a: provider list-failure warning removed', expect: /\[F6a\]/,
+    from: "warnProviderFallback('subagents.list() failed: ' + ((e && e.message) || e))", to: 'void e' },
+  { name: 'F6b: interrupt emit statement deleted (anchored by content)', expect: /\[F6b\/v2\]/,
+    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: /\[F6c\]/,
+    from: "noteStateWriteFailure('current.' + safeId(sessionId) + '.json', (e && e.message) || e)", to: 'void e' },
+]
+let red = 0
+FAMILIES.push(
+  { name: 'F5/v2: fence-root comparison removed', expect: /\[F5\/v2\] 围栏根漂移/,
+    from: 'if (actual && expected && actual !== expected) warnFenceDriftOnce(actual, expected)', to: 'void actual' },
+)
+FAMILIES.push(
+  { name: 'F-B: derivation drops the source descriptions', expect: /\[F-B\/v2\] 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: /\[F-B\/v2\] 例外集合 == docs\/parameter-schema\.md/,
+    from: "solverToolAllow: { type: 'array', items: { type: 'string' } },", to: '' },
+  { name: 'F-B: one registration site reverted to a literal', expect: /\[F-B\/v2\] 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/v2: the not-detected log drops the actionable sentence', expect: /\[task-9\/v2\]/,
+    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 v2 probe 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('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 || skipped.length || hangs.length) process.exit(1)
+console.log('ALL MUTANTS RED AS REQUIRED')