Files
msd-core/src/installer-migrations.cts
Tom Boucher 27aa40f65e fix(#3023): stage pi's shared hook bundle outside pi's reserved hooks/ directory (#3175)
* test(#3023): failing-first guard — pi must not stage hooks in its reserved dir

pi reserves <configDir>/hooks as its deprecated extension location and warns
on every startup when it exists. Assert a pi install stages the shared hook
bundle under gsd-hooks/ instead, manifests it there, and never creates hooks/.

Also adds pi to the local-scope dir table in install-shared.cjs: pi was in
RUNTIME_META but not LOCAL_DIR_NAME, so scope:'local' resolved
path.join(root, undefined) and no local pi install could be exercised.

Fails before the fix. Verified via the remote runner.

* fix(#3023): stage pi's shared hook bundle outside pi's reserved hooks/ dir

pi reserves <configDir>/hooks as its now-deprecated extension location and
warns on every startup when that directory merely exists — checkDeprecatedExtensionDirs()
guards the warning with a bare existsSync(), unlike its tools/ sibling. GSD staged
its shared hook bundle exactly there, and pi's advised remediation (move it to
extensions/) would break the adapter's paths and expose GSD's .js helpers to pi's
extension auto-discovery.

The bundle directory name is now runtime-descriptor-driven: hostBehaviors
.sharedHooksDirName, defaulting to 'hooks' so all 18 other runtimes are
byte-identical. pi sets 'gsd-hooks'. The name is validated as a single path
segment — separators, dot-only segments, trailing dots, absolute paths, NUL,
and Windows reserved device names all fall back to the default, because the
value is joined onto a user's config root and written to.

Renamed in place rather than relocated: hook scripts resolve siblings via
__dirname/.., so a depth change would silently break them.

- install / uninstall / manifest sites all read the resolved name
- pi/gsd.cjs probes gsd-hooks then hooks, so dev checkouts and half-upgraded
  trees still resolve; the never-throws contract is preserved
- new migration 009 retires the legacy pi hooks/ dir on upgrade, using a new
  non-recursive remove-empty-dir engine primitive (rmdirSync only,
  symlink-refusing, containment-guarded); ADR-0008 amended accordingly
- fixes two latent name-dependencies the rename exposed: the stale-hook scan
  and the injection scanner's self-exclusion both hardcoded 'hooks'

Verified on the remote runner.

Closes #3023

* fix(#3023): close review findings and align emitted provenance with the rename

Adversarial review found two defects, and the remote runner found four
failure clusters. All fixed here.

Review BLOCKER — detect-custom-files was blind to the renamed bundle.
GSD_PREFIX_MANAGED_DIRS in gsd-tools.cjs hardcoded 'hooks', so for pi the
whole gsd-hooks/ tree was invisible to the custom-file scan and user-added
files there were never backed up before the next update's clean-install wipe.
The dir set now resolves via the .gsd-runtime marker plus the shipped
capability registry (never bin/install.js, which is not shipped into installed
trees), and falls back to scanning every known candidate when the runtime
cannot be determined — over-scanning is safe, under-scanning is the data loss.

Review MAJOR — the pi adapter bound to an empty bundle. resolveSharedHooksDir
accepted any directory, so an interrupted install left gsd-hooks/ winning over
a fully-staged legacy hooks/ and every hook silently no-opped. A candidate now
qualifies only if it is non-empty.

Remote-runner clusters:
- emitted-provenance had no rule for the gsd-hooks/ family; added two pi-scoped
  rules pointing at the same sources the existing hooks/ rules use. The table is
  total, so an unattributed family is a hard failure by design.
- pi tests in install-minimal-hooks and the install integration suite asserted
  the old layout; updated to derive the dir name from the descriptor rather than
  hardcoding either name.
- 19 unrelated-looking failures on node22 only were a leaked fs mock: t.after()
  runs in registration order, cleanup was registered before mock.restoreAll(),
  and node22's JS rimraf calls the public fs.rmdirSync while node24's native
  path does not — so the EACCES stub leaked process-wide on one lane. Restore
  now runs first.

Verified on the remote runner.

* fix(#3023): honor PI_CODING_AGENT_DIR, ack the rename ripple, fix expandTilde

pi resolves its agent dir as PI_CODING_AGENT_DIR ?? ~/<CONFIG_DIR_NAME>/agent
(packages/coding-agent/src/config.ts). GSD's pi descriptor declared an empty
configHome.env, so a user with that variable set had GSD installed where pi
never looks. Added the env name; the dot-home-nested resolver already handled
the override, so no resolver logic changed.

Also fixes expandTilde in the shared runtime-homes resolver, found while adding
that: it hardcoded os.homedir() and ignored the opts.home every caller threads,
so EVERY runtime's tilde-valued env override (claude, antigravity, windsurf, pi)
silently resolved against the real home. That is a correctness bug and a
test-escape hazard — a sandboxed test asserting on a tilde override reached the
developer's actual home directory. Now threaded through every branch; behavior
with no injected home is unchanged.

Adds the emitted-drift ack fragment for the 58 pi paths whose emitted location
moved with the rename. The provenance rules satisfy the totality gate; the
differential gate needs the ack because the hook sources are byte-unchanged —
only the installer's target directory moved. The two hook files this branch
genuinely edits stay attributed and are not double-acked.

Note on piConfig.configDir: it is read from pi's OWN installed package.json
(getPackageDir walks up from pi's __dirname), alongside piConfig.name — a
white-label setting for a redistributed pi fork, not a per-project user setting.
Documented accordingly rather than treated as an unsupported override.

Verified on the remote runner.

* fix(#3023): reject blank env overrides, pin adapter/descriptor parity

Three review findings, all fixed.

A whitespace-only config-dir override was accepted verbatim: the guard was
`if (val)`, falsy only for the empty string, so PI_CODING_AGENT_DIR='   '
resolved to a literal three-space directory name instead of falling back to the
descriptor default. Fixed across every env-consuming branch — dot-home,
dot-home-nested, all three xdg steps, and generic-agents-root — not just pi's.
Non-blank values are still never trimmed, so '~/My Agent Dir' keeps working.

pi/gsd.cjs's probe list and the descriptor were two independent sources of truth
for the bundle directory name; a future rename would have desynced them silently
and left every pi hook quiet with no error. The probe list stays deliberate — it
must resolve in a dev checkout and a half-upgraded tree, where the registry's
answer would be wrong — so this adds the parity assertion the repo's
generative-fix-divergence rule calls for: the descriptor value must be the FIRST
candidate, and the default must remain present.

Changeset body rewritten to cover the two later user-facing fixes it had not
caught up with.

Verified on the remote runner.

* chore(#3023): backfill changeset PR number

* fix(#3023): anchor injection-scan patterns and fix a macOS detection hole

CI's security job flagged CONTEXT.md:124 — pre-existing prose reading 'not the
same fact as a genuinely empty or absent one'. The match was the 'act as a'
INSIDE 'f-act as a': the pattern had no left word boundary, so any word ending
in act tripped it (fact, impact, contract, artifact, interact, redact,
abstract). My four-line CONTEXT.md edit dragged the latent false positive into
this PR because the scan is diff-scoped by file but reads whole files. Anchored
with (^|[^[:alnum:]]) rather than rewording maintainer-owned prose, which would
have left the class alive for the next PR touching any file saying 'fact as a'.

Auditing the rest of the list for the same class surfaced a real detection hole:
the eval/exec/Function patterns matched a quote via \x27, a GNU-grep-only hex
escape. BSD/macOS grep reads it as four literal characters, so single-quoted
eval('...')/exec('...') payloads were NEVER detected there while passing on
GNU-grep CI. Replaced with a literal apostrophe class.

Boundaries were added only where a real word-suffix collision exists; exec,
jailbreak, developer mode and the role-manipulation family were audited and
deliberately left unanchored. 22 new cases cover both directions — the false
positives now scan clean, and every real payload still fires, including the
quote/punctuation/start-of-line boundary forms.

Also builds this branch's injection test fixture at runtime instead of carrying
the literal phrase, so the payload keeps its teeth without tripping the scan.

Verified on the remote runner.

---------

Co-authored-by: sim <sim@local>
2026-08-07 13:41:21 -04:00

1102 lines
41 KiB
TypeScript

/**
* Installer migrations engine — plan, apply, and track filesystem-mutation
* migrations for GSD runtime config directories.
*
* ADR-457 build-at-publish: the hand-written bin/lib/installer-migrations.cjs
* collapsed to a TypeScript source of truth. Behaviour is preserved
* byte-for-behaviour from the prior hand-written .cjs; only types are added.
*/
import fs from 'node:fs';
import path from 'node:path';
import crypto from 'node:crypto';
import {
validateInstallerMigrationActions,
validateInstallerMigrationRecord,
type MigrationRecord,
type MigrationAction,
} from './installer-migration-authoring.cjs';
import { platformWriteSync, retryRenameSync, posixNormalize } from './shell-command-projection.cjs';
import { realClock, type Clock } from './clock.cjs';
const MANIFEST_NAME = 'gsd-file-manifest.json';
const INSTALL_STATE_NAME = 'gsd-install-state.json';
const INSTALL_MIGRATION_LOCK_NAME = 'gsd-install-migration.lock';
const DEFAULT_MIGRATIONS_DIR = path.join(__dirname, 'installer-migrations');
const DEFAULT_LOCK_TIMEOUT_MS = 30_000;
const STRICT_JSON = Symbol('strict-json');
function sha256File(filePath: string): string {
const hash = crypto.createHash('sha256');
const buffer = Buffer.allocUnsafe(1024 * 1024);
const fd = fs.openSync(filePath, 'r');
try {
while (true) {
const bytesRead = fs.readSync(fd, buffer, 0, buffer.length, null);
if (bytesRead === 0) break;
hash.update(buffer.subarray(0, bytesRead));
}
} finally {
fs.closeSync(fd);
}
return hash.digest('hex');
}
function sha256Text(value: string): string {
return crypto.createHash('sha256').update(value).digest('hex');
}
/**
* Copy a managed path for the rollback snapshot or the user-facing backup,
* WITHOUT dereferencing a symlink.
*
* `fs.copyFileSync` follows symlinks, so a managed path that has been replaced
* by a link (tampering, or an unexpected user layout) would have had the
* LINK TARGET's bytes copied into `gsd-migration-journal/…-backups/` — e.g. a
* `gsd.cjs` symlinked at `~/.ssh/id_rsa` would land that key's contents in the
* backup tree. Nothing GSD installs is ever a symlink, so the faithful snapshot
* of a symlinked managed path is the link itself: recreating it preserves
* rollback fidelity (restore re-creates the same link) while never reading the
* referent. Deletion was already safe — `fs.rmSync` unlinks the link, never the
* target.
*
* Windows note: `fs.symlinkSync` can throw EPERM for unprivileged users. That
* surfaces as an apply failure and triggers the normal rollback path, which is
* the correct outcome — refusing to proceed beats silently copying referent
* bytes.
*/
/**
* Evaluate and, if safe, perform a `remove-empty-dir` action against `fullPath`.
*
* This is deliberately WEAKER than a recursive directory-removal primitive
* (which 003's docblock records as an intentional absence in the ADR-0008
* design): it only ever calls `fs.rmdirSync` — never `fs.rmSync`, never
* `{ recursive: true }`, never `{ force: true }` — so a non-empty directory
* fails the underlying syscall rather than being swept. The emptiness check
* immediately above the call is what turns that failure mode into a
* deliberate, non-error "left in place" outcome instead of surfacing ENOTEMPTY.
*
* Guards, in order:
* - lstat (not stat): a symlinked directory is refused outright, never
* followed. A missing target is reported distinctly so callers can tell
* "nothing was ever there" from "something was there and is left alone".
* - must actually be a directory (not a file masquerading under the relPath).
* - containment: the REALPATH of the target must resolve strictly inside the
* REALPATH of configDir — never equal to it (removing the config root
* itself is never in scope) and never escaping it (e.g. via an ancestor
* symlink the lstat check alone would not catch).
* - emptiness, re-checked here rather than trusted from planning time: a
* directory that still holds any entry (managed-but-undeleted, unknown,
* or created between plan and apply) is left in place. This is reported
* as 'skipped-not-empty', a successful no-op, not a failure.
*
* Any unexpected error along the way (EACCES, EBUSY, a race that removes the
* target between the lstat and the rmdir, etc.) degrades to 'left-in-place'.
* This action must never throw out of the executor, matching every sibling
* action type's failure posture.
*/
function evaluateRemoveEmptyDir(configDir: string, fullPath: string): string {
let stat: fs.Stats;
try {
stat = fs.lstatSync(fullPath);
} catch {
return 'missing';
}
if (stat.isSymbolicLink()) return 'left-in-place';
if (!stat.isDirectory()) return 'left-in-place';
let resolvedRoot: string;
let resolvedTarget: string;
try {
resolvedRoot = fs.realpathSync(configDir);
resolvedTarget = fs.realpathSync(fullPath);
} catch {
return 'left-in-place';
}
if (resolvedTarget === resolvedRoot || !resolvedTarget.startsWith(resolvedRoot + path.sep)) {
// Refuses both "target IS configDir" and "target escaped configDir".
return 'left-in-place';
}
let entries: string[];
try {
entries = fs.readdirSync(fullPath);
} catch {
return 'left-in-place';
}
if (entries.length > 0) return 'skipped-not-empty';
try {
fs.rmdirSync(fullPath);
return 'removed';
} catch {
return 'left-in-place';
}
}
function copyPreservingSymlink(srcPath: string, destPath: string): void {
if (fs.lstatSync(srcPath).isSymbolicLink()) {
// symlinkSync fails with EEXIST on an occupied path, so clear it first.
// Scoped to this branch on purpose: the regular-file path below keeps
// copyFileSync's overwrite-in-place, so a mid-restore failure cannot leave
// the destination destroyed.
fs.rmSync(destPath, { force: true });
fs.symlinkSync(fs.readlinkSync(srcPath), destPath);
return;
}
fs.copyFileSync(srcPath, destPath);
}
function readJsonIfPresent(filePath: string, fallback: unknown): unknown {
if (!fs.existsSync(filePath)) return fallback;
try {
return JSON.parse(fs.readFileSync(filePath, 'utf8'));
} catch (error) {
if (fallback === STRICT_JSON) {
throw new Error(`invalid installer migration state JSON: ${filePath}: ${(error as Error).message}`);
}
return fallback;
}
}
interface InstallManifest {
version: string | null;
timestamp: string | null;
mode: string | null;
files: Record<string, string>;
}
function readInstallManifest(configDir: string): InstallManifest {
const manifest = readJsonIfPresent(path.join(configDir, MANIFEST_NAME), null);
if (!manifest || typeof manifest !== 'object') {
return { version: null, timestamp: null, mode: null, files: {} };
}
const m = manifest as Record<string, unknown>;
return {
version: typeof m.version === 'string' ? m.version : null,
timestamp: typeof m.timestamp === 'string' ? m.timestamp : null,
mode: typeof m.mode === 'string' ? m.mode : null,
files: m.files && typeof m.files === 'object' ? m.files as Record<string, string> : {},
};
}
interface InstallState {
schemaVersion: number;
appliedMigrations: Array<Record<string, unknown>>;
}
function readInstallState(configDir: string): InstallState {
const state = readJsonIfPresent(path.join(configDir, INSTALL_STATE_NAME), STRICT_JSON);
if (!state || typeof state !== 'object') {
return { schemaVersion: 1, appliedMigrations: [] };
}
const s = state as Record<string, unknown>;
return {
schemaVersion: typeof s.schemaVersion === 'number' ? s.schemaVersion : 1,
appliedMigrations: Array.isArray(s.appliedMigrations) ? s.appliedMigrations as Array<Record<string, unknown>> : [],
};
}
// Strict atomic write for the install state: must never be left half-written.
// Bypasses the seam because platformWriteSync falls back to a direct write on
// rename failure, which would silently violate this invariant.
function atomicWriteInstallState(configDir: string, content: string): void {
fs.mkdirSync(configDir, { recursive: true });
const filePath = path.join(configDir, INSTALL_STATE_NAME);
const tmpPath = `${filePath}.tmp-${process.pid}-${Date.now()}`;
try {
fs.writeFileSync(tmpPath, content, 'utf8');
retryRenameSync(tmpPath, filePath);
} catch (error) {
try { fs.rmSync(tmpPath, { force: true }); } catch { /* best-effort */ }
throw error;
}
}
function writeInstallState(configDir: string, state: InstallState): InstallState {
atomicWriteInstallState(configDir, JSON.stringify(state, null, 2) + '\n');
return state;
}
interface ReadJsonResult {
exists: boolean;
value: unknown;
error: Error | null;
}
function readJson(configDir: string, relPath: string): ReadJsonResult {
const { fullPath } = ensureInsideConfig(configDir, relPath);
if (!fs.existsSync(fullPath)) {
return { exists: false, value: null, error: null };
}
try {
return { exists: true, value: JSON.parse(fs.readFileSync(fullPath, 'utf8')), error: null };
} catch (error) {
return { exists: true, value: null, error: error as Error };
}
}
function normalizeRelPath(relPath: string): string {
if (typeof relPath !== 'string' || relPath.trim() === '') {
throw new Error('migration action relPath must be a non-empty string');
}
const normalized = posixNormalize(relPath);
if (path.isAbsolute(normalized) || path.win32.isAbsolute(normalized)) {
throw new Error(`migration action relPath must stay inside configDir: ${relPath}`);
}
const segments = normalized.split('/');
if (segments.some((segment) => segment === '' || segment === '.' || segment === '..')) {
throw new Error(`migration action relPath must stay inside configDir: ${relPath}`);
}
return segments.join('/');
}
interface ArtifactClassification {
classification: string;
originalHash: string | null;
currentHash: string | null;
}
function classifyArtifact(configDir: string, relPath: string, manifest: InstallManifest): ArtifactClassification {
const normalized = normalizeRelPath(relPath);
const originalHash = manifest.files[normalized] || null;
const fullPath = path.join(configDir, normalized);
if (!fs.existsSync(fullPath)) {
return { classification: originalHash ? 'managed-missing' : 'missing', originalHash, currentHash: null };
}
const currentHash = sha256File(fullPath);
if (!originalHash) {
return { classification: 'unknown', originalHash: null, currentHash };
}
if (currentHash === originalHash) {
return { classification: 'managed-pristine', originalHash, currentHash };
}
return { classification: 'managed-modified', originalHash, currentHash };
}
function appliedMigrationIds(state: InstallState): Set<string> {
return new Set(
state.appliedMigrations
.filter((entry) => entry && typeof entry.id === 'string')
.map((entry) => entry.id as string)
);
}
function appliedMigrationEntries(state: InstallState): Map<string, Record<string, unknown>> {
const entries = new Map<string, Record<string, unknown>>();
for (const entry of state.appliedMigrations) {
if (entry && typeof entry.id === 'string' && !entries.has(entry.id)) {
entries.set(entry.id, entry);
}
}
return entries;
}
function migrationChecksum(migration: MigrationRecord): string {
const checksum = migration.checksum;
if (typeof checksum === 'string' && checksum) return checksum;
const serializable = {
id: migration.id,
title: migration.title || null,
description: migration.description || null,
introducedIn: migration.introducedIn || null,
runtimes: migration.runtimes || null,
scopes: migration.scopes || null,
destructive: migration.destructive === true,
runtimeContract: migration.runtimeContract || null,
plan: typeof migration.plan === 'function' ? (migration.plan as (...args: unknown[]) => unknown).toString() : null,
};
return `sha256:${sha256Text(JSON.stringify(serializable))}`;
}
// Rewrite the stored checksum of any already-applied entry whose id drifted, so the
// drift is reconciled durably and not re-detected on every subsequent run (issue #670).
// Returns the number of entries actually changed (so callers know whether a write is needed).
function reconcileDriftedChecksums(
appliedEntries: Array<Record<string, unknown>>,
checksumDrift: Array<{ id: string; currentChecksum: string }> | undefined
): number {
if (!Array.isArray(checksumDrift) || checksumDrift.length === 0) return 0;
const reconcile = new Map(checksumDrift.map((d) => [d.id, d.currentChecksum]));
let changed = 0;
for (let i = 0; i < appliedEntries.length; i++) {
const existing = appliedEntries[i];
if (existing && typeof existing.id === 'string' && reconcile.has(existing.id)) {
const next = reconcile.get(existing.id) as string;
if (existing.checksum !== next) {
appliedEntries[i] = { ...existing, checksum: next };
changed += 1;
}
}
}
return changed;
}
function collectAppliedChecksumDrift(
applied: Map<string, Record<string, unknown>>,
migrations: MigrationRecord[]
): Array<{ id: string; storedChecksum: string; currentChecksum: string }> {
const drift: Array<{ id: string; storedChecksum: string; currentChecksum: string }> = [];
for (const migration of migrations) {
const entry = applied.get(migration.id as string);
if (!entry || !entry.checksum) continue;
const currentChecksum = migrationChecksum(migration);
if (entry.checksum !== currentChecksum) {
// An already-applied migration is never re-run (it is filtered out of `pending`),
// so a checksum drift here is functionally inert. A prior release may have edited a
// shipped migration body (see issue #670). Surface it for reconciliation instead of
// hard-aborting the user's upgrade.
drift.push({
id: migration.id as string,
storedChecksum: entry.checksum as string,
currentChecksum,
});
}
}
return drift;
}
function migrationMatchesContext(migration: MigrationRecord, { runtime, scope }: { runtime: string | null; scope: string | null }): boolean {
if (Array.isArray(migration.runtimes) && (migration.runtimes as string[]).length > 0) {
if (!runtime || !(migration.runtimes as string[]).includes(runtime)) return false;
}
if (Array.isArray(migration.scopes) && (migration.scopes as string[]).length > 0) {
if (!scope || !(migration.scopes as string[]).includes(scope)) return false;
}
return true;
}
function discoverInstallerMigrations({ migrationsDir }: { migrationsDir: string }): MigrationRecord[] {
if (!migrationsDir || !fs.existsSync(migrationsDir)) return [];
return fs.readdirSync(migrationsDir, { withFileTypes: true })
.filter((entry) => entry.isFile() && entry.name.endsWith('.cjs'))
.map((entry) => entry.name)
.sort()
.flatMap((fileName) => {
const source = path.join(migrationsDir, fileName);
delete require.cache[require.resolve(source)];
// eslint-disable-next-line @typescript-eslint/no-require-imports
const exported: unknown = require(source);
const records = Array.isArray(exported) ? exported : [exported];
return records.map((record) => validateInstallerMigrationRecord(record as MigrationRecord, source));
});
}
function journalTimestamp(now: () => string): string {
return now().replace(/[:.]/g, '-');
}
function migrationRunId(appliedAt: string): string {
return `${journalTimestamp(() => appliedAt)}-${crypto.randomBytes(8).toString('hex')}`;
}
function sleepSync(ms: number): void {
const buffer = new SharedArrayBuffer(4);
Atomics.wait(new Int32Array(buffer), 0, 0, ms);
}
/**
* Check whether a given PID is alive on the current host.
* Uses process.kill(pid, 0) which works on POSIX and Windows (Node's
* implementation maps it to OpenProcess + GetExitCodeProcess on win32).
* Returns true if alive or permission-denied (live but not ours),
* false if ESRCH (no such process).
*/
function isPidAlive(pid: number): boolean {
if (typeof pid !== 'number' || !Number.isFinite(pid) || pid <= 0) return false;
try {
process.kill(pid, 0);
return true; // alive (or permission denied — treat as live)
} catch (err) {
return (err as NodeJS.ErrnoException).code !== 'ESRCH';
}
}
interface LockFileData {
pid: number;
acquiredAt: string;
}
/**
* Try to read and parse the lock file JSON. Returns null on any error
* (missing, invalid JSON, I/O failure).
*/
function readLockFile(lockPath: string): LockFileData | null {
try {
const raw = fs.readFileSync(lockPath, 'utf8');
const parsed: unknown = JSON.parse(raw);
if (parsed && typeof parsed === 'object' && typeof (parsed as Record<string, unknown>).pid === 'number') {
return parsed as LockFileData;
}
return null;
} catch {
return null;
}
}
function acquireInstallMigrationLock(
configDir: string,
{ timeoutMs = DEFAULT_LOCK_TIMEOUT_MS }: { timeoutMs?: number } = {},
clock: Clock = realClock,
): () => void {
fs.mkdirSync(configDir, { recursive: true });
const lockPath = path.join(configDir, INSTALL_MIGRATION_LOCK_NAME);
const started = clock.now();
while (true) {
let fd: number | null = null;
let lockCreatedByUs = false;
try {
fd = fs.openSync(lockPath, 'wx');
// Close the open descriptor before writing so the file handle is
// released on Windows before the release closure unlinks it.
// Write payload via writeFileSync with the path (not the fd) so we
// don't hold an open fd across the lifetime of the lock.
fs.closeSync(fd);
fd = null;
lockCreatedByUs = true; // we own the file; clean it up on any subsequent error
fs.writeFileSync(lockPath, JSON.stringify({
pid: process.pid,
acquiredAt: new Date().toISOString(),
}) + '\n');
lockCreatedByUs = false; // release closure owns cleanup from here
return () => {
const failures: Error[] = [];
// Use unlinkSync (not rmSync with { force: true }) so EPERM errors
// are NOT silently swallowed. On Windows, if the unlink fails
// transiently, the error surfaces via releaseError so the caller
// can observe and surface it rather than leaving a stale lock.
try { fs.unlinkSync(lockPath); } catch (error) { failures.push(error as Error); }
if (failures.length > 0) {
const releaseError = new Error(`failed to release installer migration lock: ${lockPath}`) as Error & { failures: Error[] };
releaseError.failures = failures;
throw releaseError;
}
};
} catch (error) {
if (fd !== null) {
try { fs.closeSync(fd); } catch { /* best-effort */ }
try { fs.unlinkSync(lockPath); } catch { /* best-effort */ }
fd = null;
} else if (lockCreatedByUs) {
// fd was closed but writeFileSync threw before we returned the release
// closure — the empty lock file is still on disk and must be removed
// so it does not orphan as an unreadable (empty/invalid JSON) stale lock.
try { fs.unlinkSync(lockPath); } catch { /* best-effort */ }
}
const err = error as NodeJS.ErrnoException;
if (err && err.code === 'EEXIST') {
// Stale-lock reclamation: read the on-disk PID and check liveness.
// If the PID is dead (ESRCH) or is our own process (same-process
// re-entry caused by rmSync silently swallowing an unlink error on
// a previous call in the same invocation — the root cause of #3670),
// reclaim the lock by removing the stale file and retrying.
const lockData = readLockFile(lockPath);
if (lockData !== null) {
const holderPid = lockData.pid;
const isSameProcess = holderPid === process.pid;
const isDeadProcess = !isPidAlive(holderPid);
if (isSameProcess || isDeadProcess) {
// Reclaim: remove the stale lock and loop back to openSync.
// Only continue (retry) when unlink actually succeeds — a silent
// continue on reclaim failure recreates the original deadlock:
// the lock stays on disk and we spin indefinitely.
let reclaimed = false;
try { fs.unlinkSync(lockPath); reclaimed = true; } catch { /* unlink failed — fall through to timeout path */ }
if (reclaimed) continue;
}
}
if (clock.now() - started >= timeoutMs) {
const holderInfo = lockData ? ` (held by pid ${lockData.pid} since ${lockData.acquiredAt})` : '';
throw new Error(`installer migration lock is held: ${lockPath}${holderInfo}`);
}
clock.sleep(Math.min(50, Math.max(1, timeoutMs - (clock.now() - started))));
continue;
}
throw error;
}
}
}
interface EnsureInsideConfigResult {
normalized: string;
fullPath: string;
}
function ensureInsideConfig(configDir: string, relPath: string): EnsureInsideConfigResult {
const normalized = normalizeRelPath(relPath);
const fullPath = path.resolve(configDir, normalized);
const root = path.resolve(configDir);
if (fullPath !== root && !fullPath.startsWith(root + path.sep)) {
throw new Error(`migration path escapes configDir: ${relPath}`);
}
return { normalized, fullPath };
}
function isStructurallyEmpty(value: unknown): boolean {
if (value === null || value === undefined) return true;
if (Array.isArray(value)) return value.length === 0;
return typeof value === 'object' && Object.keys(value).length === 0;
}
interface JournalAction extends Record<string, unknown> {
status: string;
}
function journalAction(action: MigrationAction, status: string, extras: Record<string, unknown> = {}): JournalAction {
const { value: _value, ...safeAction } = action;
return { ...safeAction, ...extras, status };
}
interface PlanContext {
configDir: string;
runtime: string | null;
scope: string | null;
manifest: InstallManifest;
state: InstallState;
baselineScan: boolean;
now: () => string;
classifyArtifact: (relPath: string) => ArtifactClassification;
readJson: (relPath: string) => ReadJsonResult;
}
interface PlannedAction extends MigrationAction {
migrationId: string;
migrationChecksum: string;
type: string;
relPath: string;
reason: string;
classification: string;
originalHash: string | null;
currentHash: string | null;
requestedType?: string;
backupRelPath?: string | null;
value?: unknown;
deleteIfEmpty?: boolean;
prompt?: unknown;
choices?: unknown[];
}
interface MigrationPlan {
generatedAt: string;
manifest: InstallManifest;
state: InstallState;
pendingMigrationIds: string[];
pendingMigrations: MigrationRecord[];
actions: PlannedAction[];
blocked: PlannedAction[];
checksumDrift: Array<{ id: string; storedChecksum: string; currentChecksum: string }>;
}
function planInstallerMigrations({
configDir,
runtime = null,
scope = null,
migrations,
baselineScan = false,
now = () => new Date().toISOString(),
}: {
configDir: string;
runtime?: string | null;
scope?: string | null;
migrations: MigrationRecord[];
baselineScan?: boolean;
now?: () => string;
}): MigrationPlan {
if (!configDir) throw new Error('configDir is required');
if (!Array.isArray(migrations)) throw new Error('migrations must be an array');
const manifest = readInstallManifest(configDir);
const state = readInstallState(configDir);
const validatedMigrations = migrations.map((migration) =>
validateInstallerMigrationRecord(migration)
);
const scopedMigrations = validatedMigrations.filter((migration) =>
migrationMatchesContext(migration, { runtime, scope })
);
const applied = appliedMigrationEntries(state);
const checksumDrift = collectAppliedChecksumDrift(applied, scopedMigrations);
const pending = scopedMigrations.filter((migration) => !applied.has(migration.id as string));
const actions: PlannedAction[] = [];
const blocked: PlannedAction[] = [];
const classifications = new Map<string, ArtifactClassification>();
const classify = (relPath: string): ArtifactClassification => {
const normalized = normalizeRelPath(relPath);
if (!classifications.has(normalized)) {
classifications.set(normalized, classifyArtifact(configDir, normalized, manifest));
}
return classifications.get(normalized)!;
};
for (const migration of pending) {
const planFn = migration.plan as (ctx: PlanContext) => unknown[];
const plannedActions = planFn({
configDir,
runtime,
scope,
manifest,
state,
baselineScan,
now,
classifyArtifact: classify,
readJson: (relPath) => readJson(configDir, relPath),
});
validateInstallerMigrationActions(plannedActions, migration);
const checksum = migrationChecksum(migration);
for (const rawAction of plannedActions as MigrationAction[]) {
const relPath = normalizeRelPath(rawAction.relPath as string);
const classification = rawAction.classification
? {
classification: rawAction.classification as string,
originalHash: rawAction.originalHash as string | null || null,
currentHash: rawAction.currentHash as string | null || null,
}
: classify(relPath);
let protectedType = rawAction.type as string;
if (rawAction.type === 'remove-managed' && classification.classification === 'managed-modified') {
protectedType = 'backup-and-remove';
}
if (rawAction.type === 'remove-managed' && classification.classification === 'unknown') {
protectedType = 'preserve-user';
}
const action: PlannedAction = {
migrationId: migration.id as string,
migrationChecksum: checksum,
type: protectedType,
relPath,
reason: rawAction.reason as string || migration.description as string || '',
classification: classification.classification,
originalHash: classification.originalHash,
currentHash: classification.currentHash,
};
if (action.type !== rawAction.type) {
action.requestedType = rawAction.type as string | undefined;
}
if (action.type === 'backup-and-remove') {
action.backupRelPath = null;
}
if (action.type === 'rewrite-json') {
action.value = rawAction.value;
action.deleteIfEmpty = rawAction.deleteIfEmpty === true;
}
if (rawAction.prompt) action.prompt = rawAction.prompt;
if (Array.isArray(rawAction.choices)) action.choices = rawAction.choices as unknown[];
if (action.type === 'prompt-user') {
blocked.push(action);
} else if (
action.classification === 'unknown' &&
action.type !== 'rewrite-json' &&
action.type !== 'record-baseline' &&
action.type !== 'baseline-preserve-user'
) {
blocked.push(action);
}
actions.push(action);
}
}
return {
generatedAt: now(),
manifest,
state,
pendingMigrationIds: pending.map((migration) => migration.id as string),
pendingMigrations: pending,
actions,
blocked,
checksumDrift,
};
}
function uniqueActionMigrationIds(actions: PlannedAction[]): string[] {
return [...new Set(actions.map((action) => action.migrationId).filter(Boolean))];
}
interface RollbackArgs {
configDir: string;
journal: { actions: JournalAction[] };
journalPath: string;
rollbackRoot: string;
backupRoot: string;
previousInstallStateBytes: string | null;
}
function rollbackAppliedMigrationResult({ configDir, journal, journalPath, rollbackRoot, backupRoot, previousInstallStateBytes }: RollbackArgs): void {
const failures: Array<{ relPath: string; error: string }> = [];
for (const action of [...journal.actions].reverse()) {
if (!action.rollbackRelPath) continue;
const rollbackPath = path.join(configDir, action.rollbackRelPath as string);
const dest = path.join(configDir, action.relPath as string);
try {
// lstat-based existence check: a snapshot of a symlinked managed path is
// itself a link, and existsSync() follows it — a link whose target is
// gone would read as "missing" and silently skip the restore.
if (fs.lstatSync(rollbackPath, { throwIfNoEntry: false })) {
fs.mkdirSync(path.dirname(dest), { recursive: true });
copyPreservingSymlink(rollbackPath, dest);
}
} catch (error) {
failures.push({ relPath: action.relPath as string, error: (error as Error).message });
}
if (action.backupRelPath) {
try {
fs.rmSync(path.join(configDir, action.backupRelPath as string), { force: true });
} catch {
// backup cleanup is best-effort; preserve restore failures above
}
}
}
try {
if (previousInstallStateBytes === null) {
fs.rmSync(path.join(configDir, INSTALL_STATE_NAME), { force: true });
} else {
atomicWriteInstallState(configDir, previousInstallStateBytes);
}
} catch (error) {
failures.push({ relPath: INSTALL_STATE_NAME, error: (error as Error).message });
}
try {
fs.rmSync(journalPath, { force: true });
fs.rmSync(rollbackRoot, { recursive: true, force: true });
fs.rmSync(backupRoot, { recursive: true, force: true });
} catch {
// journal cleanup is best-effort; the rollback above is the safety-critical part
}
if (failures.length > 0) {
const error = new Error('migration rollback incomplete') as Error & { rollbackFailures: typeof failures };
error.rollbackFailures = failures;
throw error;
}
}
function cleanupMigrationRunArtifacts(journalPath: string, rollbackRoot: string, backupRoot: string): void {
try { fs.rmSync(journalPath, { force: true }); } catch { /* best-effort */ }
try { fs.rmSync(rollbackRoot, { recursive: true, force: true }); } catch { /* best-effort */ }
try { fs.rmSync(backupRoot, { recursive: true, force: true }); } catch { /* best-effort */ }
}
interface ApplyResult {
appliedMigrationIds: string[];
journalRelPath: string;
rollback: () => void;
}
function applyInstallerMigrationPlan({
configDir,
plan,
now = () => new Date().toISOString(),
}: {
configDir: string;
plan: MigrationPlan;
now?: () => string;
}): ApplyResult {
if (!configDir) throw new Error('configDir is required');
if (!plan || !Array.isArray(plan.actions)) throw new Error('plan with actions is required');
if (Array.isArray(plan.blocked) && plan.blocked.length > 0) {
throw new Error(`migration plan has ${plan.blocked.length} blocked action(s)`);
}
const appliedAt = now();
const runId = migrationRunId(appliedAt);
const journalRelPath = path.posix.join('gsd-migration-journal', `${runId}.json`);
const journalPath = path.join(configDir, journalRelPath);
const rollbackRootRelPath = path.posix.join('gsd-migration-journal', `${runId}-rollback`);
const rollbackRoot = path.join(configDir, rollbackRootRelPath);
const backupRootRelPath = path.posix.join('gsd-migration-journal', `${runId}-backups`);
const backupRoot = path.join(configDir, backupRootRelPath);
const journal: { schemaVersion: number; appliedAt: string; appliedMigrationIds: string[]; actions: JournalAction[] } = {
schemaVersion: 1,
appliedAt,
appliedMigrationIds: uniqueActionMigrationIds(plan.actions),
actions: [],
};
const rollback: Array<{ relPath: string; rollbackPath: string }> = [];
const installStatePath = path.join(configDir, INSTALL_STATE_NAME);
const previousInstallStateBytes = fs.existsSync(installStatePath)
? fs.readFileSync(installStatePath, 'utf8')
: null;
try {
fs.mkdirSync(path.dirname(journalPath), { recursive: true });
platformWriteSync(journalPath, JSON.stringify(journal, null, 2) + '\n');
for (const action of plan.actions) {
if (
action.type !== 'remove-managed' &&
action.type !== 'backup-and-remove' &&
action.type !== 'rewrite-json' &&
action.type !== 'record-baseline' &&
action.type !== 'baseline-preserve-user' &&
action.type !== 'remove-empty-dir'
) {
throw new Error(`unsupported migration action type: ${action.type}`);
}
const { normalized, fullPath } = ensureInsideConfig(configDir, action.relPath);
if (!fs.existsSync(fullPath)) {
journal.actions.push(journalAction(action, 'missing'));
continue;
}
if (action.type === 'record-baseline' || action.type === 'baseline-preserve-user') {
journal.actions.push(journalAction(action, action.type === 'record-baseline' ? 'recorded' : 'preserved'));
continue;
}
if (action.type === 'remove-empty-dir') {
// Directory actions never enter the file-copy/rollback machinery below:
// there is nothing to snapshot-and-restore for a directory node itself
// (its former CONTENTS were already snapshotted by their own file-level
// actions before this one runs), and rollback of a removed empty
// directory is simply re-creating it, which the rollback path below
// does not model. Non-recursive by construction (evaluateRemoveEmptyDir
// only ever calls fs.rmdirSync), so there is nothing destructive to undo
// beyond an mkdir the next install/migration run will happily redo.
journal.actions.push(journalAction(action, evaluateRemoveEmptyDir(configDir, fullPath)));
continue;
}
const rollbackPath = path.join(rollbackRoot, normalized);
fs.mkdirSync(path.dirname(rollbackPath), { recursive: true });
copyPreservingSymlink(fullPath, rollbackPath);
rollback.push({ relPath: normalized, rollbackPath });
if (action.type === 'rewrite-json') {
if (action.deleteIfEmpty && isStructurallyEmpty(action.value)) {
fs.rmSync(fullPath, { force: true });
journal.actions.push(journalAction(action, 'removed', {
rollbackRelPath: path.posix.join(rollbackRootRelPath, normalized),
}));
} else {
platformWriteSync(fullPath, JSON.stringify(action.value, null, 2) + '\n');
journal.actions.push(journalAction(action, 'rewritten', {
rollbackRelPath: path.posix.join(rollbackRootRelPath, normalized),
}));
}
continue;
}
if (action.type === 'backup-and-remove') {
const backupRelPath = action.backupRelPath || path.posix.join(backupRootRelPath, normalized);
const backupPath = path.join(configDir, backupRelPath);
fs.mkdirSync(path.dirname(backupPath), { recursive: true });
copyPreservingSymlink(fullPath, backupPath);
journal.actions.push(journalAction(action, 'removed', {
backupRelPath,
rollbackRelPath: path.posix.join(rollbackRootRelPath, normalized),
}));
} else {
journal.actions.push(journalAction(action, 'removed', {
rollbackRelPath: path.posix.join(rollbackRootRelPath, normalized),
}));
}
fs.rmSync(fullPath, { force: true });
}
platformWriteSync(journalPath, JSON.stringify(journal, null, 2) + '\n');
const state = readInstallState(configDir);
const applied = appliedMigrationIds(state);
const nextApplied = [...state.appliedMigrations];
reconcileDriftedChecksums(nextApplied, plan.checksumDrift);
const actionsByMigrationId = new Map<string, PlannedAction>();
for (const action of plan.actions) {
if (action.migrationId && !actionsByMigrationId.has(action.migrationId)) {
actionsByMigrationId.set(action.migrationId, action);
}
}
for (const id of journal.appliedMigrationIds) {
if (!applied.has(id)) {
const action = actionsByMigrationId.get(id);
nextApplied.push({
id,
appliedAt,
journal: journalRelPath,
checksum: action && action.migrationChecksum ? action.migrationChecksum : null,
});
}
}
writeInstallState(configDir, {
schemaVersion: 1,
appliedMigrations: nextApplied,
});
return {
appliedMigrationIds: journal.appliedMigrationIds,
journalRelPath,
rollback: () => rollbackAppliedMigrationResult({ configDir, journal, journalPath, rollbackRoot, backupRoot, previousInstallStateBytes }),
};
} catch (error) {
const rollbackFailures: Array<{ relPath: string; rollbackPath: string; error: string }> = [];
for (const entry of rollback.reverse()) {
const dest = path.join(configDir, entry.relPath);
try {
fs.mkdirSync(path.dirname(dest), { recursive: true });
// Symlink-preserving, same as the forward path: `entry.rollbackPath` is
// itself a link whenever the managed path was one, so a raw copy here
// would dereference it and write the referent's bytes back to the LIVE
// install path — a worse leak than the journal-tree one, since it is
// user-visible and at a predictable location.
copyPreservingSymlink(entry.rollbackPath, dest);
} catch (rollbackError) {
rollbackFailures.push({
relPath: entry.relPath,
rollbackPath: entry.rollbackPath,
error: (rollbackError as Error).message,
});
}
}
if (rollbackFailures.length > 0) {
const rollbackError = new Error(`migration apply failed and rollback incomplete: ${(error as Error).message}`) as Error & { cause: unknown; rollbackFailures: typeof rollbackFailures };
rollbackError.cause = error;
rollbackError.rollbackFailures = rollbackFailures;
throw rollbackError;
}
cleanupMigrationRunArtifacts(journalPath, rollbackRoot, backupRoot);
throw error;
}
}
function markPendingMigrationsApplied({
configDir,
plan,
now = () => new Date().toISOString(),
}: {
configDir: string;
plan: MigrationPlan;
now?: () => string;
}): string[] {
if (!plan) return [];
const hasPending = Array.isArray(plan.pendingMigrationIds) && plan.pendingMigrationIds.length > 0;
const hasDrift = Array.isArray(plan.checksumDrift) && plan.checksumDrift.length > 0;
if (!hasPending && !hasDrift) return [];
const appliedAt = now();
const state = readInstallState(configDir);
const applied = appliedMigrationIds(state);
const nextApplied = [...state.appliedMigrations];
const reconciledCount = reconcileDriftedChecksums(nextApplied, plan.checksumDrift);
const newlyApplied: string[] = [];
if (hasPending) {
const checksumsByMigrationId = new Map<string, string>();
for (const migration of plan.pendingMigrations || []) {
checksumsByMigrationId.set(migration.id as string, migrationChecksum(migration));
}
for (const id of plan.pendingMigrationIds) {
if (applied.has(id)) continue;
nextApplied.push({
id,
appliedAt,
journal: null,
checksum: checksumsByMigrationId.get(id) || null,
});
newlyApplied.push(id);
}
}
if (newlyApplied.length > 0 || reconciledCount > 0) {
writeInstallState(configDir, {
schemaVersion: 1,
appliedMigrations: nextApplied,
});
}
return newlyApplied;
}
interface RunResult {
appliedMigrationIds: string[];
journalRelPath: string | null;
plan: MigrationPlan;
blocked?: PlannedAction[];
rollback?: () => void;
}
function runInstallerMigrations({
configDir,
runtime = null,
scope = null,
migrationsDir = DEFAULT_MIGRATIONS_DIR,
migrations = discoverInstallerMigrations({ migrationsDir }),
baselineScan = false,
now = () => new Date().toISOString(),
lockTimeoutMs = DEFAULT_LOCK_TIMEOUT_MS,
}: {
configDir: string;
runtime?: string | null;
scope?: string | null;
migrationsDir?: string;
migrations?: MigrationRecord[];
baselineScan?: boolean;
now?: () => string;
lockTimeoutMs?: number;
} = { configDir: '' }): RunResult {
const releaseLock = acquireInstallMigrationLock(configDir, { timeoutMs: lockTimeoutMs });
let primaryError: (Error & { suppressed?: Error[] }) | null = null;
let completed = false;
try {
const plan = planInstallerMigrations({ configDir, runtime, scope, migrations, baselineScan, now });
if (plan.actions.length === 0) {
const newlyApplied = markPendingMigrationsApplied({ configDir, plan, now });
completed = true;
return {
appliedMigrationIds: newlyApplied,
journalRelPath: null,
plan,
};
}
if (plan.blocked.length > 0) {
completed = true;
return {
appliedMigrationIds: [],
journalRelPath: null,
plan,
blocked: plan.blocked,
};
}
const result = applyInstallerMigrationPlan({ configDir, plan, now });
completed = true;
return { ...result, plan };
} catch (error) {
primaryError = error as Error & { suppressed?: Error[] };
throw error;
} finally {
try {
releaseLock();
} catch (releaseError) {
if (primaryError) {
primaryError.suppressed = [...(primaryError.suppressed || []), releaseError as Error];
} else if (completed) {
throw releaseError;
} else {
throw releaseError;
}
}
}
}
// Unused but kept to satisfy eslint — sleepSync is referenced in the original
// and may be used by test code that patches this module.
void sleepSync;
export = {
DEFAULT_MIGRATIONS_DIR,
INSTALL_MIGRATION_LOCK_NAME,
INSTALL_STATE_NAME,
MANIFEST_NAME,
acquireInstallMigrationLock,
applyInstallerMigrationPlan,
classifyArtifact,
discoverInstallerMigrations,
evaluateRemoveEmptyDir,
migrationChecksum,
planInstallerMigrations,
readInstallManifest,
readInstallState,
runInstallerMigrations,
writeInstallState,
};