* test(#3412): failing-first suite for the pattern-construction seam Phase 1 of epic #3212 (ADR-3212 §1/§2/§7). Tests only — src/pattern.cts and eslint-rules/no-adhoc-regex-escape.cjs do not exist yet, so both suites fail with MODULE_NOT_FOUND, which is the intended RED. Locks the measured behavior rather than the assumed behavior: RegExp.escape hex-escapes the leading character of nearly every string ("abc" -> "\x61bc"), so the suite asserts match-equivalence against an inlined historical oracle (the implementation being deleted) rather than byte-equivalence of pattern text — 200 seeded fast-check runs plus a fixed corpus, 0 mismatches. Also locks the latent character-class range bug this phase fixes as a side effect: a hyphen-bearing value interpolated into [...] currently forms a real range and matches an unintended character; post-migration it must not. * chore(#3412): src/pattern.cts owns runtime-value regex construction Phase 1 of epic #3212 (ADR-3212 §1/§2/§6/§7). Adds the pattern seam delegating to the built-in RegExp.escape, deletes every hand-rolled copy, and raises the Node floor to the Active LTS line. The census was low, three times over. ADR-3212 counted 10 copies; a graph query found 12; the new lint rule — once live — found 27 more. The difference is that the census counted named helper FUNCTIONS while the rule counts the escape SHAPE, so inline .replace(<class>, '\$&') copies were never in scope. ADR §1's actual requirement is that no module outside the seam escapes a value for regex use, so all of them are, and CLAUDE.md's no-defer rule makes them this change's work. Fourth consecutive epic here whose copy count was low — the argument for ADR-3180 Amendment 3's "state N found by the guard" rule. Also corrected mid-implementation: the survey reported phase-id.cts's escapeRegex had 0 external importers. It had 8 production importers, making its removal a public-surface change to an ADR-2121-owned module and requiring an update to that ADR's locked-surface test. Blast radius revised Medium-High -> High. RegExp.escape is match-equivalent but NOT text-equivalent: it hex-escapes the leading char of nearly every string ("abc" -> "\x61bc"). Equivalence is proven by a seeded fast-check property test against the deleted implementation as oracle. It also fixes a latent bug: a hyphen-bearing value interpolated into a character class previously formed a real range and matched an unintended character. Node floor 22 -> 24 (RegExp.escape is Node 24+), across engines, .nvmrc, package-lock, 9 CI matrix entries, and 5 docs. The aggregate `required-tests` context is unchanged and no job was added or removed, so branch protection cannot be orphaned by the dropped lanes. Enforced by eslint-rules/no-adhoc-regex-escape.cjs (shape-matched, with structural provenance for reviewed pattern-fragment constants rather than a name heuristic) plus a whole-tree companion guard covering the directories ESLint's globs miss. * fix(#3412): close the _SOURCE guard evasion, correct two false claims Three findings from the orthogonal review pass, all fixed. 1. The ESLint rule's `_SOURCE` provenance fallback was pure identifier- name matching with no binding check, so `new RegExp(userInput_SOURCE)` — a function parameter — sailed past the guard. That is the same rename-evasion class issue #3410 documents, reopened by the very fallback meant to complement the structural check. Now bound to the identifier's actual binding kind: import, require-derived const, or module-scope const; parameters, `let`/`var`, and unresolvable bindings fail closed. Four RuleTester cases cover the evasion and prove the legitimate cross-module case still passes. 2. src/pattern.cts's own header carried the stale pre-correction counts (12 copies / 17 call sites) while CONTEXT.md and the design doc carried the corrected ones (~39 / ~44) — a self-contradiction inside the PR whose entire purpose is deleting divergent copies. Rewritten, preserving the durable lesson: a named-function census cannot see inline copies; only a shape-matching guard can. 3. The claim that all deleted copies threw TypeError on non-string was false. phase-id.cts's copy — the one with 8 external importers — did String(value).replace(...) and never threw. The seam's locked signature does not coerce, so this is a real, now-disclosed behavior change rather than the pure preservation the tests asserted. Audited all 32 invocations across the 8 importers and 6 in-file callers: every one is safe by construction (upstream truthy guard or a string-producing derivation), verified by runtime probe against the compiled modules rather than by TS compilation, which cannot see a runtime undefined. Corrected the false claim in both the test comment and the design doc, and added it to Known limits. * docs(#3412): add Changed changeset for the Node 24 floor The only user-visible break in this phase. The escape-behavior change is internal and match-equivalent, so it carries no user-facing note. * fix(#3412): resolve the seam's require graph in script fixtures and packaging Checkpoint 2 came back red with 90 failures on the node24 lane. Three distinct defects, all introduced by routing scripts/ through the new pattern seam, none reproducible by any local gate: 1. ~82 failures — tests/adr-index-gate.test.cjs and tests/removed-but-needed-lint.test.cjs copy a scripts/*.cjs into an mkdtemp fixture and spawn it there (necessary: those scripts resolve their scan root from __dirname/.., so running the real script would scan the real repo). Each harness hand-listed the dependencies to copy alongside. Adding require('../gsd-core/bin/lib/pattern.cjs') to gen-adr-index.cjs made both lists silently incomplete -> MODULE_NOT_FOUND, plus 17 downstream 'did not emit parseable JSON' failures from the same crash. Fixed as a class, not an instance: new tests/helpers/copy-script- fixture.cjs walks a script's transitive static relative-require graph and copies it, so dependencies are derived and never re-declared. It throws (naming the unbuilt artifact) instead of letting the child die with a bare MODULE_NOT_FOUND. Verified for all four seam-consuming scripts: gen-adr-index, lint-removed-but-needed, gen-loop-host- contract, sync-runtime-launcher. 2. 2 failures — scripts/ ships wholesale but eslint-rules/ does not, so the new scripts/lint-no-adhoc-regex-escape.cjs would be MODULE_NOT_FOUND in a published install (#2858 guard). Excluded from the tarball, matching the existing precedent for gen-emitted- baseline.cjs, which is excluded for the identical reason, and locked with a test modeled on that one. Confirmed against a real npm pack: 890 files, 0 from eslint-rules/, and gsd-core/bin/lib/pattern.cjs present (so the other four scripts' requires are legitimate). 3. 6 failures — tests/phase-id.test.cjs asserted the literal escaped source text ('0*29', 'PROJ-42'). RegExp.escape is match-equivalent to the retired hand-rolled escaper but NOT text-equivalent: it hex- escapes the leading character and all hyphens ('0*\x329', '\x50ROJ\x2d42'). Verified NOT a behavior change — 576 match decisions across all three real interpolation prefixes, zero divergence. Those tests now compile each source into the same heading regex src/roadmap.cts's searchPhaseInContent builds and assert what matches and what does not, including the 'i'-flag canonicalization the hex escape has to preserve. Re-pinning the new literals would have rebuilt the same brittleness one layer down. Adds a test for the property the escape exists for: a dot in '1.2' must not act as a wildcard. Also shares one definition of 'a require' between the packaging guard and the fixture copier, so the two cannot disagree about what they scan. Co-Authored-By: Claude Opus 5 <noreply@anthropic.com> * fix(#3412): refuse to copy a fixture dependency outside the fixture root copyScriptWithDeps resolved each relative require and joined the repo-relative result onto fixtureRoot. A require resolving OUTSIDE the repo yields a '../'-prefixed relative path, so path.join climbed out of the fixture and wrote into the surrounding temp dir (verified: repoRoot=/repo + depAbs=/etc/passwd wrote /tmp/etc/passwd). No script in the tree does this today, so this closes an available escape rather than an active one. Refuses via the existing unresolved- require path so the failure names the offending specifier. Covered by a negative proof that the guard fires and that nothing lands outside the fixture. Co-Authored-By: Claude Opus 5 <noreply@anthropic.com> * fix(#3412): parse requires instead of pattern-matching them; restore the foreign-prefix contract Applies all findings from the second orthogonal review round, re-run because real code changed after round 1. HIGH (security) — extractRequires stripped BLOCK comments before LINE comments, so a '//' comment containing '/*' opened a phantom block comment, and a '//' inside a string literal truncated the line. Both hid real requires: 'const u="http://x"; require("./real.cjs")' returned [], and four real requires in gsd-core/bin/gsd-tools.cjs were invisible. Replaced with a real AST parse via espree. This is ADR-3212's own Decision 4 — tokenizer-first for stateful grammars — applied to the case it describes; comment/string/regex nesting is exactly such a grammar, which is why the regex version was wrong. The function was moved byte-identical out of the #2858 packaging guard, so the bug PRE-DATES this branch and has been a live blind spot there: a shipped script could have required an unshipped path undetected. Fixing it makes that guard strictly stronger than on next. espree is promoted from a transitive eslint dependency to an explicit devDependency rather than relying on hoisting. The script parse attempt sets ecmaFeatures.globalReturn because Node wraps CommonJS bodies in a function, making a top-level return legal — scripts/check-coverage-gate .cjs relies on it, and without the flag the guard throws on a file it is supposed to scan. Verified 0 unparseable across all 324 .cjs/.js under scripts/, bin/, and gsd-core/bin/, and 0 new violations against a real npm pack, so the exact extractor does not newly fail the guard. MEDIUM (security) — the repo-containment check guarded dependencies but not the entry path. One escapesContainment predicate now guards both. LOW (security) — containment was lexical while fs follows symlinks, and a directory symlink could mint a fresh dedupe key per level. realpath now resolves both repoRoot and each dependency before the decision, and the realpath-derived path is the dedupe key. Destination layout still uses the original repo-relative path, so copied trees are unchanged. MAJOR (standards) — the round-1 behavioral rewrite of phase-id tests lost the foreign-prefix contract: every assertion was satisfied by an impl returning [A-Z]+\x2d42, i.e. ANY project code — the exact #3599 bug class the exact-source prevents. The literal assertions it replaced were catching this. Now asserts the compiled regex REJECTS a different prefix with the same number. MAJOR (standards) — the test hand-duplicated production's heading regex with no parity guard (CLAUDE.md's 'Generative Fix Divergence'). Removed the parallel surface instead of policing it: src/roadmap.cts exports buildPhaseHeadingRegex, searchPhaseInContent calls it, the test imports it. Byte-identical .source and .flags verified for both escaped forms. MINOR — '..foo' no longer false-flagged as an escape; the inverted spurious-vs-missing doc claim corrected; the dead allow-test-rule header removed. Co-Authored-By: Claude Opus 5 <noreply@anthropic.com> * chore(#3412): backfill changeset pr number to 3416 * fix(#3412): make the escape guard's own regex linear, reword an injection-scan collision Two CI failures on PR #3416, both in code this branch added. CodeQL js/redos (high) — REPLACE_CALL_RE's outer alternation let a bracket run be consumed EITHER by the character-class branch OR one character at a time by the trailing catch-all, so a failing match explored both parses of every pair. Measured on the real regex: n=26 -> 204ms, n=28 -> 791ms, n=30 -> 3475ms, a clean 2^n. This script scans repo source, so a file with a long bracket run after '.replace(/' would hang CI outright — a guard against undisciplined pattern construction was itself the worst pattern in the diff. Fixed the way ADR-3212 already prescribes: the catch-all branch now excludes '[' and ']' so a bracket can only be consumed by the class branch (this is what makes it linear), and every quantifier is bounded (the locked bounded-quantifiers decision) as a second line of defense. Now 0ms at n=2000. Disclosed coverage tradeoff, recorded at the constant: a regex literal with a BARE unescaped ']' outside a class is no longer matched by this backstop. No census shape has that form, and the AST rule remains the primary detector. Verified the guard did not go blind doing it: a real census-shape violation is still reported, and an allow-adhoc-regex-escape suppression comment is still honored. Regression test drives the exported findViolations on a 2000-repetition adversarial input and asserts the RESULT. It makes no wall-clock assertion — elapsed-time tests are forbidden — so a regression surfaces as a harness timeout, which is the correct signal. Prompt injection scan — 'must not act as a regex wildcard' in a test comment matched the scanner's jailbreak pattern act\s+as\s+(a|an|if| my). Reworded to 'behave as'. Deliberately NOT allowlisted: silencing a whole test file over one phrase would blunt the scanner permanently, and the comment has nothing to do with injection. Neither failure was reachable from the remote runner — CodeQL and the injection scan are not in that matrix, so the sha it passed was green and still wrong. Co-Authored-By: Claude Opus 5 <noreply@anthropic.com> --------- Co-authored-by: sim <sim@local> Co-authored-by: Claude Opus 5 <noreply@anthropic.com>
934 lines
43 KiB
JavaScript
934 lines
43 KiB
JavaScript
#!/usr/bin/env node
|
||
'use strict';
|
||
|
||
/**
|
||
* Anti-divergence drift guard for the PHASE-COMPLETION predicate (epic
|
||
* #3180, issue #3186, ADR-3180 Decision 4, spec §7.4).
|
||
*
|
||
* The derivation this guard protects is "is phase P complete?", under the
|
||
* DISK-STRICT rule §7.4 now locks (amended on this branch, commit
|
||
* af92fd4c9, per the #2957 maintainer decision): `readVerificationStatus`
|
||
* is called UNCONDITIONALLY (no plan-count precondition), and a ticked
|
||
* ROADMAP checkbox carries NO machine authority over disk state. The
|
||
* designated owner is `src/verification.cts` · `isPhaseComplete` (Decision
|
||
* 1). Per ADR-3180 Amendment 3's standing rule this guard was written BEFORE
|
||
* any src/ file was touched (guard-first discovery), with
|
||
* `FUNCTION_SCOPED_EXEMPTIONS` below INTENTIONALLY EMPTY at that point. The
|
||
* implementation step (issue #3186) has since landed `isPhaseComplete` and
|
||
* migrated its call sites onto it, and populated the map — see its own
|
||
* comment immediately above its definition for the per-entry reasoning,
|
||
* including two DECLARED DEVIATIONS (`buildPhaseCompletionProjection`'s
|
||
* retained `implementation_complete` and `buildWorkstreamInventory`'s
|
||
* pure-projection boundary) not named by the design doc's own
|
||
* DO-NOT-MIGRATE list.
|
||
*
|
||
* Per ADR-3180 Decision 4(a) this guard discovers call sites by SCANNING THE
|
||
* WHOLE `src/` TREE, never an allowlist of known files. Per Decision 4(d)
|
||
* that surface is widened further: `src/` alone is itself the forbidden
|
||
* allowlist, one directory wide, so this guard ALSO scans the prompt-layer
|
||
* markdown (`gsd-core/workflows`, `commands`, `agents`, `skills`) — see the
|
||
* "PROMPT-LAYER PROSE DETECTION" section below.
|
||
*
|
||
* FOUR INDEPENDENT SHAPES are detected, matching issue #3186's dispatch (shape
|
||
* (d) added by the Phase 4 follow-up that closed the `cmdStateSync` /
|
||
* `src/state.cts` gap the remote matrix exposed — see shape (d)'s own header
|
||
* below):
|
||
*
|
||
* (a) CHECKBOX-DERIVED COMPLETION — a ROADMAP `- [x] Phase N` tick treated
|
||
* as completion evidence. Categorically wrong under disk-strict
|
||
* (§7.4's DECIDED resolution: "a ticked ROADMAP checkbox is a human
|
||
* annotation with no machine authority"). Detected as the PAIRED
|
||
* shape actually observed twice in this tree: an `if` statement whose
|
||
* condition references a roadmap/checkbox-derived "complete" boolean
|
||
* AND tests some status variable `!== 'complete'`, followed (within a
|
||
* small bounded window — this pairs ONE conditional with its OWN
|
||
* consequent statement, the same tight-pairing role
|
||
* `lint-state-field-drift.cjs`'s `LADDER_WINDOW_LINES` plays, NOT the
|
||
* big function-scoped co-occurrence Decision 4(a)'s Goodhart lesson
|
||
* targets) by an assignment of that SAME variable to the literal
|
||
* `'complete'`.
|
||
* (b) PLAN-COUNT PRECONDITION GATING A VERIFICATION READ — §7.4 names this
|
||
* exactly as `buildPhaseCompletionProjection`'s divergence: a
|
||
* `planCount > 0`-shaped gate that decides WHETHER
|
||
* `readVerificationStatus(` runs at all, rather than calling it
|
||
* unconditionally. Detected as FUNCTION-SCOPED co-occurrence (no
|
||
* window — Decision 4(a)'s Goodhart lesson, and Phase 5's "bounded
|
||
* window missed 7 of 14 copies" trap) of a count-gate ANYWHERE in the
|
||
* function's own body, together with a ternary- OR block/`if`-gated
|
||
* call to `readVerificationStatus(` (a call that is NOT the function's
|
||
* unconditional top-level statement) — the ternary form is a same-line
|
||
* `?` before the call; the block form is a call sitting inside the
|
||
* BODY of an `if (...)` whose OWN condition matches the count-gate
|
||
* (`findIfCountGateBlockLines`), which also catches `if (planCount >
|
||
* 0) { x = readVerificationStatus(...) }` — a #3186 review finding
|
||
* (4a): the prior version matched only the same-line ternary and
|
||
* produced zero hits on the block form. Known, disclosed limit: a
|
||
* multi-line `if` condition whose opening `{` lands on a later line
|
||
* than the condition's own closing `)` is not detected (every
|
||
* count-gate condition actually present in this tree is one line).
|
||
* (c) LOCAL RE-IMPLEMENTATION OF "COMPLETE" FROM COUNTS — a
|
||
* `summaryCount >= planCount`-shaped comparison (either operand
|
||
* order, or the algebraic `summaryCount - planCount >= 0` restatement
|
||
* and its mirror), computed locally instead of calling the owner.
|
||
* Known, disclosed limit: further algebraic restatements (an
|
||
* intermediate difference variable, `Math.max`/`!()`-wrapped forms,
|
||
* etc.) are not reliably matchable by a bounded, non-backtracking
|
||
* regex and are not attempted — see the regexes' own comment. This is
|
||
* the single sharpest textual signature this epic's divergent copies
|
||
* share: `buildPhaseCompletionProjection`, `buildStateFrontmatter`
|
||
* (via `scanPhasePlans`'s own `completed` field),
|
||
* `cmdRoadmapAnalyze`, `cmdRoadmapUpdatePlanProgress`, and
|
||
* `buildWorkstreamInventory` all independently hand-roll this exact
|
||
* comparison.
|
||
* (d) A BARE FIELD READ OF `scanPhasePlans(...).completed` OUTSIDE THE
|
||
* OWNER (`src/plan-scan.cts`) — the shape the remote matrix exposed:
|
||
* `cmdStateSync` (`src/state.cts`, now fixed) destructured
|
||
* `scanPhasePlans(dirPath).completed` directly and used it AS a
|
||
* completion verdict, with no comparison for shapes (a)/(b)/(c) to
|
||
* catch — a bare property read is not a re-derivation shape any of the
|
||
* other three detectors match. `scanPhasePlans` legitimately EXPOSES
|
||
* `.completed` (it is Phase 1's own owner for "are all plans
|
||
* summarized?", a real and distinct question from "is the phase
|
||
* complete?"), so reading the field is not inherently wrong — USING it
|
||
* as a completion verdict is. That distinction is DATA-FLOW (what the
|
||
* caller does with the value), which no textual guard can decide.
|
||
* KNOWN, DISCLOSED, HONEST LIMIT: this detector cannot tell a
|
||
* legitimate "are summaries caught up with plans" read from an illegal
|
||
* "is the phase complete" read — it flags EVERY read of `.completed`
|
||
* off a `scanPhasePlans(` result outside `src/plan-scan.cts` and
|
||
* relies on `FUNCTION_SCOPED_EXEMPTIONS` carrying a WRITTEN REASON per
|
||
* exempted function (see that map's own comment for each current
|
||
* entry's reasoning) rather than silently deciding the question for
|
||
* itself. Detected in three independent textual forms, ALL
|
||
* function-scoped (Decision 4(a), no line window — see below):
|
||
* - the direct chained form, `scanPhasePlans(...).completed`, on one
|
||
* statement — the call's own argument list is walked with a plain
|
||
* paren-depth counter (not a `[^)]*` regex), so a nested-paren
|
||
* argument (`scanPhasePlans(path.join(phasesDir, dir))`, the
|
||
* majority real shape in this tree) is still matched correctly;
|
||
* - the destructured form, `const { completed } = scanPhasePlans(…)`
|
||
* or the renamed-alias form `const { completed: isDone } =
|
||
* scanPhasePlans(…)`, on one statement;
|
||
* - the INDIRECT form — `const scan = scanPhasePlans(…)` binding the
|
||
* whole result to a variable, with `.completed` read off that same
|
||
* variable ANYWHERE else in the SAME named function, including on a
|
||
* different line — no bounded window (the same Phase-5 "bounded
|
||
* window missed 7 of 14 copies" lesson shape (b) already learned;
|
||
* `cmdStateSync`'s own real shape bound the result to a
|
||
* destructured `{ planCount: plans, summaryCount: summaries }`
|
||
* rather than a whole-object variable, so this indirect form is
|
||
* precautionary breadth for a shape not yet observed in this tree,
|
||
* not a shape already caught in the wild).
|
||
*
|
||
* DETECTION IS FUNCTION-SCOPED, NOT A BOUNDED LINE-WINDOW, for the
|
||
* *discovery* co-occurrence in each shape (shape (a)'s if/assignment pairing
|
||
* and shape (b)'s ternary-clause pairing are each ONE conditional's own two
|
||
* halves — the same narrow, unavoidable pairing `LADDER_WINDOW_LINES` bounds
|
||
* in the sibling state-field guard, not a re-derivation-hiding Goodhart
|
||
* target). Phase 5's first guard used a bounded line window between a
|
||
* ladder and its consuming call and MISSED 7 of 14 live copies inside the
|
||
* very function it scanned — that lesson is why shape (b)'s outer
|
||
* count-gate/gated-call pairing has NO line-distance bound at all: both
|
||
* signals need only appear somewhere in the SAME named function.
|
||
*
|
||
* COMMENT-AWARE. Phase 3's first `lint-phase-enumeration-drift.cjs` flagged
|
||
* JSDoc/inline comments that merely DOCUMENTED the derivation, not code that
|
||
* re-derives it (ADR-3180 Amendment 3). This guard reuses the SAME
|
||
* comment/string-stripping tokenizer `lint-state-field-drift.cjs` proved
|
||
* (`scanCode` below): comments and string/template literal CONTENTS are
|
||
* blanked before any detection regex runs, cross-line-aware for block
|
||
* comments and multi-line template literals.
|
||
*
|
||
* FUNCTION-SCOPED EXEMPTIONS ONLY, NEVER A WHOLE-FILE ALLOWLIST (Decision
|
||
* 4(a)/(d)): a whole-file exemption on the owner is precisely how
|
||
* `getMilestoneInfo` stayed invisible to an earlier guard (Decision 4(d)).
|
||
* `FUNCTION_SCOPED_EXEMPTIONS` is a `Map<relPath, Set<functionName>>`, kept
|
||
* EMPTY here because the owner (`src/verification.cts` · `isPhaseComplete`)
|
||
* does not exist yet — this comment, not a populated map, is what the
|
||
* implementation step (Phase 4's migration PR) replaces.
|
||
*
|
||
* Every regex below is small, bounded, and has no nested/overlapping
|
||
* quantifiers — the same ReDoS discipline `npm run lint:ci`'s CodeQL
|
||
* js/redos query verifies over every sibling drift guard.
|
||
*
|
||
* The tree-walk / root-confinement / sanitizer machinery is SHARED with the
|
||
* sibling drift guards via `scripts/lib/drift-scan.cjs` (ADR-3180 Decision
|
||
* 4).
|
||
*/
|
||
|
||
const path = require('node:path');
|
||
const driftScan = require('./lib/drift-scan.cjs');
|
||
const { MAX_REGEX_LITERAL_LEN, sanitizeForReport, scanTree } = driftScan;
|
||
const { escapeRegex } = require('../gsd-core/bin/lib/pattern.cjs');
|
||
|
||
// ─── SHARED TOKENIZER + FUNCTION ATTRIBUTION (mirrors lint-state-field-drift.cjs) ──
|
||
//
|
||
// Two parallel per-line views from ONE single-pass, escape-aware character
|
||
// scan (not a regex — nothing for a backtracking engine to explore):
|
||
// - `detect[i]`: comments stripped, string/template CONTENTS kept verbatim
|
||
// (this is what every detection regex below runs against).
|
||
// - `braces[i]`: comments AND string/template CONTENTS stripped, used only
|
||
// for brace-depth counting, so a brace inside a string/comment never
|
||
// perturbs the depth count.
|
||
// `inBlockComment` / `inTemplate` are threaded ACROSS lines. Regex literals
|
||
// are not specially recognised — same documented, narrow, known limitation
|
||
// as the sibling guards (harmless for every regex literal actually present
|
||
// in this repo's completion-predicate call sites today, each balanced on
|
||
// its own line).
|
||
function scanCode(lines) {
|
||
const detect = new Array(lines.length);
|
||
const braces = new Array(lines.length);
|
||
let inBlockComment = false;
|
||
let inTemplate = false;
|
||
for (let li = 0; li < lines.length; li++) {
|
||
const line = lines[li];
|
||
let outDetect = '';
|
||
let outBraces = '';
|
||
let i = 0;
|
||
if (inTemplate) {
|
||
const start = i;
|
||
while (i < line.length) {
|
||
if (line[i] === '\\') {
|
||
i += 2;
|
||
continue;
|
||
}
|
||
if (line[i] === '`') {
|
||
i++;
|
||
inTemplate = false;
|
||
break;
|
||
}
|
||
i++;
|
||
}
|
||
outDetect += line.slice(start, i);
|
||
if (inTemplate) {
|
||
detect[li] = outDetect;
|
||
braces[li] = '';
|
||
continue;
|
||
}
|
||
}
|
||
while (i < line.length) {
|
||
if (inBlockComment) {
|
||
const close = line.indexOf('*/', i);
|
||
if (close === -1) {
|
||
i = line.length;
|
||
break;
|
||
}
|
||
i = close + 2;
|
||
inBlockComment = false;
|
||
continue;
|
||
}
|
||
const ch = line[i];
|
||
if (ch === '/' && line[i + 1] === '/') {
|
||
i = line.length;
|
||
break;
|
||
}
|
||
if (ch === '/' && line[i + 1] === '*') {
|
||
inBlockComment = true;
|
||
i += 2;
|
||
continue;
|
||
}
|
||
if (ch === "'" || ch === '"') {
|
||
const quote = ch;
|
||
const start = i;
|
||
let j = i + 1;
|
||
while (j < line.length) {
|
||
if (line[j] === '\\') {
|
||
j += 2;
|
||
continue;
|
||
}
|
||
if (line[j] === quote) {
|
||
j++;
|
||
break;
|
||
}
|
||
j++;
|
||
}
|
||
outDetect += line.slice(start, j);
|
||
i = j;
|
||
continue;
|
||
}
|
||
if (ch === '`') {
|
||
const start = i;
|
||
let j = i + 1;
|
||
let closed = false;
|
||
while (j < line.length) {
|
||
if (line[j] === '\\') {
|
||
j += 2;
|
||
continue;
|
||
}
|
||
if (line[j] === '`') {
|
||
j++;
|
||
closed = true;
|
||
break;
|
||
}
|
||
j++;
|
||
}
|
||
if (!closed) {
|
||
outDetect += line.slice(start);
|
||
inTemplate = true;
|
||
i = line.length;
|
||
break;
|
||
}
|
||
outDetect += line.slice(start, j);
|
||
i = j;
|
||
continue;
|
||
}
|
||
outDetect += ch;
|
||
outBraces += ch;
|
||
i++;
|
||
}
|
||
detect[li] = outDetect;
|
||
braces[li] = outBraces;
|
||
}
|
||
return { detect, braces };
|
||
}
|
||
|
||
// Named function scope openers (mirrors lint-state-field-drift.cjs exactly):
|
||
// - `function NAME(...) {` — top-level OR nested, any indentation, an
|
||
// optional leading `export ` tolerated by `\b` alone.
|
||
const FUNCTION_DECL_RE = /\bfunction\s+([A-Za-z_$][\w$]*)\s*\(/;
|
||
// - `const NAME = (...): ReturnType => {` — block-bodied arrow assigned to
|
||
// a const (an expression-bodied arrow `=> ({...})` never opens a new
|
||
// function frame; its `{` is an object literal, still counted toward
|
||
// brace depth, but attributes no name).
|
||
const ARROW_CONST_RE = /\bconst\s+([A-Za-z_$][\w$]*)\s*=\s*\([^)]*\)\s*(?::\s*[^=]+)?=>\s*\{/;
|
||
|
||
/**
|
||
* Pure: walk `lines` once, maintaining a brace-depth stack of open named
|
||
* function frames (deferring a multi-line `function NAME(` signature until
|
||
* the line whose brace count actually increases — see
|
||
* `lint-state-field-drift.cjs`'s `buildFunctionInfo` header for the full
|
||
* rationale this mirrors verbatim). Returns `{ innermostAt, detect }`:
|
||
* `innermostAt[i]` is the name of the innermost named function open at line
|
||
* `i` (or `null` at module scope), `detect[i]` is the comment/string-
|
||
* preserving-but-stripped-of-comments view every detector regex runs
|
||
* against.
|
||
*/
|
||
function buildFunctionInfo(lines) {
|
||
const { detect, braces } = scanCode(lines);
|
||
const innermostAt = new Array(lines.length).fill(null);
|
||
const stack = []; // { name, openDepth }
|
||
let depth = 0;
|
||
let pendingDeclName = null;
|
||
for (let i = 0; i < lines.length; i++) {
|
||
const detectCode = detect[i];
|
||
const braceCode = braces[i];
|
||
|
||
let immediateName = null;
|
||
if (detectCode.trim()) {
|
||
const arrowMatch = ARROW_CONST_RE.exec(detectCode);
|
||
if (arrowMatch) {
|
||
immediateName = arrowMatch[1];
|
||
} else {
|
||
const declMatch = FUNCTION_DECL_RE.exec(detectCode);
|
||
if (declMatch) pendingDeclName = declMatch[1];
|
||
}
|
||
}
|
||
|
||
const opens = (braceCode.match(/\{/g) || []).length;
|
||
const closes = (braceCode.match(/\}/g) || []).length;
|
||
depth += opens - closes;
|
||
|
||
if (immediateName) stack.push({ name: immediateName, openDepth: depth });
|
||
|
||
if (pendingDeclName) {
|
||
if (opens > 0) {
|
||
stack.push({ name: pendingDeclName, openDepth: depth });
|
||
pendingDeclName = null;
|
||
} else if (detectCode.includes(';')) {
|
||
pendingDeclName = null;
|
||
}
|
||
}
|
||
|
||
while (stack.length > 0 && depth < stack[stack.length - 1].openDepth) stack.pop();
|
||
|
||
innermostAt[i] = stack.length > 0 ? stack[stack.length - 1].name : null;
|
||
}
|
||
return { innermostAt, detect };
|
||
}
|
||
|
||
/**
|
||
* §7.4/#3186 review finding 4(a): the BLOCK form of shape (b)'s gate — `if
|
||
* (planCount > 0) { … readVerificationStatus(…) … }` — is textually
|
||
* indistinguishable from an unrelated `if` block by a same-line regex; the
|
||
* prior ternary-only `GATED_VERIFICATION_READ_RE` produced zero hits on it.
|
||
* Deliberately narrower than a generic "is this line nested at all" check
|
||
* (which would false-positive on every unrelated wrapper — a `try` block, a
|
||
* `withPlanningLock(cwd, () => { … })` callback, a `for` loop — none of
|
||
* which are a plan-count GATE): only lines inside the BODY of an `if (...)`
|
||
* whose OWN condition matches `COUNT_GATE_RE` are marked. Line-granularity
|
||
* brace bookkeeping (mirrors `buildFunctionInfo`'s own style), not a
|
||
* per-character brace matcher — a multi-line `if` condition whose `{` lands
|
||
* on a later line than `extractIfCondition`'s reported `endLine` is a known,
|
||
* disclosed limitation (real count-gates in this tree are one-line
|
||
* conditions; see the module header).
|
||
*/
|
||
function findIfCountGateBlockLines(lines) {
|
||
const { detect, braces } = scanCode(lines);
|
||
const gated = new Array(lines.length).fill(false);
|
||
const stack = []; // { openDepth } for open if-blocks whose condition is a count-gate
|
||
let depth = 0;
|
||
for (let i = 0; i < lines.length; i++) {
|
||
const detectCode = detect[i];
|
||
const braceCode = braces[i];
|
||
|
||
let isCountGateIfHeader = false;
|
||
if (detectCode.trim()) {
|
||
const ifMatch = IF_OPEN_RE.exec(detectCode);
|
||
if (ifMatch) {
|
||
const startCol = ifMatch.index + ifMatch[0].length;
|
||
const condition = extractIfCondition(detect, i, startCol);
|
||
if (condition && COUNT_GATE_RE.test(condition.text)) isCountGateIfHeader = true;
|
||
}
|
||
}
|
||
|
||
const opens = (braceCode.match(/\{/g) || []).length;
|
||
const closes = (braceCode.match(/\}/g) || []).length;
|
||
depth += opens - closes;
|
||
|
||
// A line is "inside" a count-gate if-block when either a PRIOR line
|
||
// already opened one and it has not yet closed, or THIS line's own `if`
|
||
// header both matches the gate and opens its body on the same line
|
||
// (`if (planCount > 0) { … }` — the exact shape in the finding).
|
||
gated[i] = stack.length > 0 || (isCountGateIfHeader && opens > closes);
|
||
|
||
if (isCountGateIfHeader && opens > closes) stack.push({ openDepth: depth });
|
||
|
||
while (stack.length > 0 && depth < stack[stack.length - 1].openDepth) stack.pop();
|
||
}
|
||
return gated;
|
||
}
|
||
|
||
// ─── SHAPE (a): CHECKBOX-DERIVED COMPLETION ────────────────────────────────
|
||
//
|
||
// The `if` half of the pairing: a condition referencing a roadmap/checkbox-
|
||
// derived "complete" boolean (`roadmapComplete`, `roadmap_complete`, any
|
||
// `\w*roadmap\w*complete\w*` spelling — case-insensitive, both real sites in
|
||
// this tree use exactly `roadmapComplete`) AND testing some OTHER status
|
||
// variable against the literal `!== 'complete'` (single or double quotes).
|
||
const IF_OPEN_RE = /\bif\s*\(/;
|
||
const ROADMAP_COMPLETE_IDENT_RE = /\broadmap\w*complete\w*\b/i;
|
||
const STATUS_NEQ_COMPLETE_RE = /\b([A-Za-z_$][\w$]*)\s*!==\s*['"]complete['"]/;
|
||
|
||
// The consequent half: an assignment of the SAME status variable to the
|
||
// literal `'complete'`. The negative lookbehind excludes `!==`/`<=`/`>=`/`==`
|
||
// (each of which also contains a bare `=` immediately before a quote) so
|
||
// this never re-matches the `if` line's own `!== 'complete'` clause — a
|
||
// single bounded character class, not a nested quantifier.
|
||
const ASSIGN_COMPLETE_RE = /(?<![!<>=])=\s*['"]complete['"]/;
|
||
|
||
// How many lines the consequent assignment may trail its own `if` line by.
|
||
// This bounds ONE conditional's own two halves (condition, then its direct
|
||
// consequent statement) — the same narrow role `LADDER_WINDOW_LINES` plays
|
||
// in the sibling state-field guard, not the function-scoped, unbounded
|
||
// co-occurrence Decision 4(a)'s Goodhart lesson targets for shape (b) below.
|
||
const CHECKBOX_OVERRIDE_WINDOW_LINES = 4;
|
||
|
||
// The `if (...)` condition in both real sites nests a SECOND, unrelated
|
||
// parenthesised group (`(completion.phase_complete || planCount === 0)`), so
|
||
// a single-line, no-nested-parens regex over the whole condition
|
||
// systematically MISSES the second site. Extracted by plain paren-depth
|
||
// counting instead — a linear character walk, not a regex, so there is
|
||
// nothing for a backtracking engine to explore regardless of nesting depth.
|
||
// Bounded to IF_CONDITION_MAX_LINES so a pathologically unterminated `if (`
|
||
// cannot walk the whole file.
|
||
const IF_CONDITION_MAX_LINES = 10;
|
||
|
||
function extractIfCondition(detectLines, startLine, startCol) {
|
||
let depth = 1; // the `(` at startCol already opened the condition
|
||
let text = '';
|
||
const endLineLimit = Math.min(detectLines.length, startLine + IF_CONDITION_MAX_LINES);
|
||
for (let li = startLine; li < endLineLimit; li++) {
|
||
const line = detectLines[li];
|
||
const from = li === startLine ? startCol : 0;
|
||
for (let ci = from; ci < line.length; ci++) {
|
||
const ch = line[ci];
|
||
if (ch === '(') depth++;
|
||
else if (ch === ')') {
|
||
depth--;
|
||
if (depth === 0) return { text, endLine: li };
|
||
}
|
||
text += ch;
|
||
}
|
||
text += '\n';
|
||
}
|
||
return null; // unterminated within the bound — treated as no match
|
||
}
|
||
|
||
function findChecklistOverrideDrift(text, relPath, exemptFunctions) {
|
||
const out = [];
|
||
const lines = text.split('\n');
|
||
const { innermostAt, detect } = buildFunctionInfo(lines);
|
||
for (let i = 0; i < lines.length; i++) {
|
||
const detectCode = detect[i];
|
||
if (!detectCode.trim()) continue;
|
||
const ifMatch = IF_OPEN_RE.exec(detectCode);
|
||
if (!ifMatch) continue;
|
||
const startCol = ifMatch.index + ifMatch[0].length;
|
||
const condition = extractIfCondition(detect, i, startCol);
|
||
if (!condition) continue;
|
||
if (!ROADMAP_COMPLETE_IDENT_RE.test(condition.text)) continue;
|
||
const neqMatch = STATUS_NEQ_COMPLETE_RE.exec(condition.text);
|
||
if (!neqMatch) continue;
|
||
const varName = neqMatch[1];
|
||
const limit = Math.min(lines.length, condition.endLine + 1 + CHECKBOX_OVERRIDE_WINDOW_LINES);
|
||
for (let j = condition.endLine + 1; j < limit; j++) {
|
||
const assignCode = detect[j];
|
||
if (!assignCode.trim()) continue;
|
||
if (!assignCode.includes(varName)) continue;
|
||
if (!ASSIGN_COMPLETE_RE.test(assignCode)) continue;
|
||
const fn = innermostAt[j] || innermostAt[i];
|
||
if (fn && exemptFunctions && exemptFunctions.has(fn)) break;
|
||
out.push({ line: j + 1, found: lines[j].trim().slice(0, MAX_REGEX_LITERAL_LEN), shape: 'a', fn: fn || null });
|
||
break;
|
||
}
|
||
}
|
||
return out;
|
||
}
|
||
|
||
// ─── SHAPE (b): PLAN-COUNT PRECONDITION GATING A VERIFICATION READ ─────────
|
||
//
|
||
// A "count > 0"-shaped gate — any identifier containing `count`
|
||
// (case-insensitive) compared `> 0`. Function-scoped presence only (no
|
||
// window): §7.4 names this exact shape as `planCount > 0`, but the
|
||
// identifier is matched generically so a differently-named count (or a
|
||
// future `isPhaseComplete` re-implementation reusing the same gate under a
|
||
// new name) is still caught.
|
||
const COUNT_GATE_RE = /\b[A-Za-z_$][\w$]*count\b\s*>\s*0\b/i;
|
||
|
||
// The call this derivation's owner (`readVerificationStatus`, wrapped by the
|
||
// not-yet-existing `isPhaseComplete`) must run UNCONDITIONALLY per §7.4. A
|
||
// line where `readVerificationStatus(` is reached via a ternary — a `?`
|
||
// appearing anywhere earlier on the SAME line — is a GATED call, not an
|
||
// unconditional one. `[^\n]*` is bounded by the line itself (no backtracking
|
||
// blow-up: a single non-newline character class followed by one literal).
|
||
const VERIFICATION_READ_CALL_RE = /\breadVerificationStatus\(/;
|
||
const GATED_VERIFICATION_READ_RE = /\?[^\n]*\breadVerificationStatus\(/;
|
||
|
||
function findGatedVerificationReadDrift(text, relPath, exemptFunctions) {
|
||
const out = [];
|
||
const lines = text.split('\n');
|
||
const { innermostAt, detect } = buildFunctionInfo(lines);
|
||
const ifCountGateBlockLines = findIfCountGateBlockLines(lines);
|
||
|
||
const countGateFns = new Set();
|
||
for (let i = 0; i < lines.length; i++) {
|
||
const detectCode = detect[i];
|
||
if (!detectCode.trim()) continue;
|
||
if (COUNT_GATE_RE.test(detectCode)) {
|
||
const fn = innermostAt[i];
|
||
if (fn) countGateFns.add(fn);
|
||
}
|
||
}
|
||
|
||
for (let i = 0; i < lines.length; i++) {
|
||
const detectCode = detect[i];
|
||
if (!detectCode.trim()) continue;
|
||
if (!VERIFICATION_READ_CALL_RE.test(detectCode)) continue;
|
||
// #3186 review finding 4(a): "gated" is EITHER a same-line ternary `?`
|
||
// before the call, OR the call sitting inside the BODY of an `if (...)`
|
||
// block whose own condition is a count-gate (`findIfCountGateBlockLines`)
|
||
// — the latter catches the block form (`if (planCount > 0) { …
|
||
// readVerificationStatus(…) … }`), which the ternary-only regex produced
|
||
// zero hits on.
|
||
const ternaryGated = GATED_VERIFICATION_READ_RE.test(detectCode);
|
||
const blockGated = ifCountGateBlockLines[i];
|
||
if (!ternaryGated && !blockGated) continue;
|
||
const fn = innermostAt[i];
|
||
if (!fn || !countGateFns.has(fn)) continue;
|
||
if (exemptFunctions && exemptFunctions.has(fn)) continue;
|
||
out.push({ line: i + 1, found: lines[i].trim().slice(0, MAX_REGEX_LITERAL_LEN), shape: 'b', fn });
|
||
}
|
||
return out;
|
||
}
|
||
|
||
// ─── SHAPE (c): LOCAL RE-IMPLEMENTATION OF "COMPLETE" FROM COUNTS ──────────
|
||
//
|
||
// `summaryCount >= planCount` (either identifier order, either comparison
|
||
// direction) — the single textual signature `buildPhaseCompletionProjection`,
|
||
// `scanPhasePlans`, `cmdRoadmapAnalyze`, `cmdRoadmapUpdatePlanProgress`, and
|
||
// `buildWorkstreamInventory` each independently hand-roll. Case-insensitive
|
||
// so `SummaryCount`/`summary_count` spellings are still caught; both operand
|
||
// orders are covered by two small, non-overlapping alternatives.
|
||
const SUMMARY_GE_PLAN_RE = /\bsummar\w*count\w*\s*>=\s*\w*plan\w*count\w*/i;
|
||
const PLAN_LE_SUMMARY_RE = /\bplan\w*count\w*\s*<=\s*\w*summar\w*count\w*/i;
|
||
|
||
// §7.4/#3186 review finding 4(b): the literal `>=`/`<=` regexes above missed
|
||
// the algebraic restatement `summaryCount - planCount >= 0` (and its mirror,
|
||
// `planCount - summaryCount <= 0`) — same comparison, no literal `>=`/`<=`
|
||
// between the two count identifiers. Widened to cover exactly these two
|
||
// zero-compared-difference shapes; each is a single bounded alternative, no
|
||
// nested/overlapping quantifiers.
|
||
//
|
||
// KNOWN, DISCLOSED LIMIT (not claimed covered): arbitrary further algebraic
|
||
// restatements — `!(planCount > summaryCount)`, a difference stored in an
|
||
// intermediate variable before the comparison, `Math.max(0, planCount -
|
||
// summaryCount) === 0`, etc. — are NOT reliably detectable by a bounded,
|
||
// non-backtracking regex and are not attempted here. This is a genuine gap,
|
||
// not swept under "et cetera": the header above disclosed it as such rather
|
||
// than claiming a wider net than the regexes actually cast.
|
||
const SUMMARY_MINUS_PLAN_GE_ZERO_RE = /\bsummar\w*count\w*\s*-\s*\w*plan\w*count\w*\s*>=\s*0\b/i;
|
||
const PLAN_MINUS_SUMMARY_LE_ZERO_RE = /\bplan\w*count\w*\s*-\s*\w*summar\w*count\w*\s*<=\s*0\b/i;
|
||
|
||
function findLocalCompletionCountDerivationDrift(text, relPath, exemptFunctions) {
|
||
const out = [];
|
||
const lines = text.split('\n');
|
||
const { innermostAt, detect } = buildFunctionInfo(lines);
|
||
for (let i = 0; i < lines.length; i++) {
|
||
const detectCode = detect[i];
|
||
if (!detectCode.trim()) continue;
|
||
if (
|
||
!SUMMARY_GE_PLAN_RE.test(detectCode)
|
||
&& !PLAN_LE_SUMMARY_RE.test(detectCode)
|
||
&& !SUMMARY_MINUS_PLAN_GE_ZERO_RE.test(detectCode)
|
||
&& !PLAN_MINUS_SUMMARY_LE_ZERO_RE.test(detectCode)
|
||
) continue;
|
||
const fn = innermostAt[i];
|
||
if (fn && exemptFunctions && exemptFunctions.has(fn)) continue;
|
||
out.push({ line: i + 1, found: lines[i].trim().slice(0, MAX_REGEX_LITERAL_LEN), shape: 'c', fn: fn || null });
|
||
}
|
||
return out;
|
||
}
|
||
|
||
// ─── SHAPE (d): scanPhasePlans(...).completed READ AS A COMPLETION VERDICT ─
|
||
//
|
||
// See the module header's shape (d) entry for the full rationale and the
|
||
// three textual forms detected below. `FUNCTION_SCOPED_EXEMPTIONS` is
|
||
// SHARED with shapes (a)/(b)/(c) — the same per-file, per-function map, so a
|
||
// function already exempted for one shape (e.g. `scanPhasePlans` itself,
|
||
// which legitimately builds the `completed` field it returns) is exempted
|
||
// for shape (d) too, and a function newly exempted for shape (d) must carry
|
||
// its own written reason in that map's comment exactly like the others.
|
||
const SCAN_CALL_TOKEN = 'scanPhasePlans(';
|
||
const DOT_COMPLETED_RE = /^\.completed\b/;
|
||
const SCAN_ASSIGN_VAR_RE = /\b(?:const|let|var)\s+([A-Za-z_$][\w$]*)\s*=\s*scanPhasePlans\(/;
|
||
// Bounded: `[^{}]*` is a single, non-nested, non-overlapping quantifier — no
|
||
// backtracking blow-up regardless of destructure-pattern length. Real
|
||
// destructuring assignments in this tree are single-line (the same
|
||
// assumption `SCAN_ASSIGN_VAR_RE` and every sibling shape's regexes make).
|
||
const DESTRUCTURE_ASSIGN_RE = /\{([^{}]*)\}\s*=\s*scanPhasePlans\(/;
|
||
const COMPLETED_KEY_RE = /\bcompleted\b/;
|
||
|
||
function isWordChar(ch) {
|
||
return !!ch && /[A-Za-z0-9_$]/.test(ch);
|
||
}
|
||
|
||
// Plain paren-depth counter over a SINGLE line (not a regex — nothing for a
|
||
// backtracking engine to explore), so a nested-paren call argument (e.g.
|
||
// `scanPhasePlans(path.join(phasesDir, dir))`, the majority real shape in
|
||
// this tree) still resolves to the call's own true closing `)`. Confined to
|
||
// one line: every real `scanPhasePlans(` call site in this tree closes on
|
||
// the line it opens on (the same assumption the rest of this file's
|
||
// detectors make about this specific call).
|
||
function findCallEndOnLine(line, afterOpenParenIdx) {
|
||
let depth = 1;
|
||
for (let i = afterOpenParenIdx; i < line.length; i++) {
|
||
const ch = line[i];
|
||
if (ch === '(') depth++;
|
||
else if (ch === ')') {
|
||
depth--;
|
||
if (depth === 0) return i + 1;
|
||
}
|
||
}
|
||
return -1;
|
||
}
|
||
|
||
function findScanPhasePlansCompletedReadDrift(text, relPath, exemptFunctions) {
|
||
const out = [];
|
||
const lines = text.split('\n');
|
||
const { innermostAt, detect } = buildFunctionInfo(lines);
|
||
const isExemptAt = (i) => {
|
||
const fn = innermostAt[i];
|
||
return !!(fn && exemptFunctions && exemptFunctions.has(fn));
|
||
};
|
||
|
||
// fnKey: the innermost function name at the ASSIGNMENT site, or this
|
||
// sentinel for module-scope assignments (never collides with a real
|
||
// identifier — function names cannot start with U+0000).
|
||
const FN_KEY_MODULE = ' |