Compare commits

..
Author SHA1 Message Date
Garry TanandClaude Fable 5 a42352081c test(extract): pin the version-arm stamp floor with an explicit regression test
The floor in extractStaleFromDB was only covered incidentally — the #1768
test's updated_at (2026-06-02) fell below the bumped VERSION_TS, but its
comment still says the date was chosen to sit ABOVE the watermark. A future
date 'repair' there would silently drop floor coverage. This test pins it
directly: pre-watermark page clears after --stale and stays cleared.
Verified fail-without-fix against master's extract.ts.

Co-Authored-By: Claude Fable 5 <noreply@anthropic.com>
2026-07-22 11:59:09 -07:00
Garry TanandClaude Fable 5 070f4c5678 fix(extract): floor --stale stamp at LINK_EXTRACTOR_VERSION_TS so version bumps don't create permanently-stale pages
The version-ts bump in this PR exposed a latent conflict: extractStaleFromDB
stamps links_extracted_at = the page's read updated_at (#1768 µs fix), but a
page last edited BEFORE the new LINK_EXTRACTOR_VERSION_TS then lands below the
version watermark and the 'links_extracted_at < versionTs' arm re-flags it
stale on every run — an infinite re-extraction loop. Stamp
max(updated_at, versionTs); versionTs is always a past release date, so the
D4 concurrent-edit race guard still holds.

Fixes the test/extract-stale.test.ts CI failure on this branch.

Co-Authored-By: Claude Fable 5 <noreply@anthropic.com>
2026-07-22 10:56:38 -07:00
Garry TanandClaude Fable 5 4b2df89935 fix(extract): resolve path-style wikilinks outside DIR_PATTERN + add ops/ to whitelist (#2576 bug 2)
- Add ops to DIR_PATTERN so [[ops/services/...]] wikilinks and bare
  ops/... slug refs are recognized as qualified entity references.
- queryBasenameIndex (the shared basename matcher behind
  resolveBasenameMatches, the FS resolver, and the doctor check) now
  falls back to the path tail when a path-style ref misses — before,
  normalizeBasename stripped slashes into a garbage key that could
  never hit the tail-keyed index.
- Bump LINK_EXTRACTOR_VERSION_TS so extract --stale re-sweeps
  previously-stamped pages with the new extraction logic.

Bugs 1+3 of #2576 are covered by PR #2717 (--stale nullResolver).

Co-Authored-By: Claude Fable 5 <noreply@anthropic.com>
2026-07-21 14:33:45 -07:00
6 changed files with 58 additions and 311 deletions
-109
View File
@@ -151,100 +151,6 @@ export function shouldSpawnAutopilotWorker(args: string[]): boolean {
return !args.includes('--no-worker');
}
/**
* #1525 — positional subcommand translation.
*
* Pre-fix, `gbrain autopilot status` silently fell through to "start daemon"
* because `runAutopilot()` only branched on flag forms (`--status`, etc.).
* `status` was treated as a stray positional and ignored.
*
* This translator maps known positional subcommands to their flag form so
* `autopilot status` is equivalent to `autopilot --status`, then rejects
* any unrecognized positional with a fail-loud error before any side
* effect (lockfile, daemon spawn, sync dispatch) runs.
*
* Scope decisions:
* - Known aliases: `status` → `--status`, `install` → `--install`,
* `uninstall` → `--uninstall`, `start` → (drop; default daemon launch).
* - `stop` is intentionally NOT aliased here. Stopping a running daemon
* is a new behavior (read PID from lock, SIGTERM, drain) that deserves
* its own design and PR. Users typing `gbrain autopilot stop` today get
* the unknown-positional error with the canonical alternatives.
* - At most one positional allowed; multiple positionals fail loud.
*/
// Every flag that consumes the NEXT argv token. Missing one here makes the
// translator misread the flag's value as a positional subcommand and exit 2
// (e.g. `--install --target linux-cron`). Keep in sync with parseArg call sites.
const AUTOPILOT_VALUE_FLAGS = new Set(['--repo', '--interval', '--target']);
const AUTOPILOT_POSITIONAL_ALIASES: Record<string, string | null> = {
status: '--status',
install: '--install',
uninstall: '--uninstall',
start: null, // drop the positional; default behavior is daemon launch
};
export type PositionalTranslation =
| { ok: true; args: string[] }
| {
ok: false;
reason: 'unknown_subcommand' | 'multiple_subcommands';
message: string;
};
export function translatePositionalSubcommands(args: string[]): PositionalTranslation {
const out: string[] = [];
let positionalSeen = false;
let i = 0;
while (i < args.length) {
const a = args[i];
if (AUTOPILOT_VALUE_FLAGS.has(a)) {
// Pass through the flag and its value untouched. If the value is
// missing at end-of-argv, fall through so the existing parseArg
// path can report the broken usage.
out.push(a);
if (i + 1 < args.length) {
out.push(args[i + 1]);
i += 2;
} else {
i += 1;
}
continue;
}
if (a.startsWith('-')) {
out.push(a);
i += 1;
continue;
}
// Positional subcommand.
if (positionalSeen) {
const known = Object.keys(AUTOPILOT_POSITIONAL_ALIASES).join(', ');
return {
ok: false,
reason: 'multiple_subcommands',
message: `Multiple subcommands given. Use only one of: ${known}.`,
};
}
positionalSeen = true;
if (a in AUTOPILOT_POSITIONAL_ALIASES) {
const alias = AUTOPILOT_POSITIONAL_ALIASES[a];
if (alias) out.push(alias);
i += 1;
continue;
}
const known = Object.keys(AUTOPILOT_POSITIONAL_ALIASES).join(', ');
return {
ok: false,
reason: 'unknown_subcommand',
message:
`Unknown subcommand: \`${a}\`.\n` +
`Allowed subcommands: ${known}.\n` +
`Or use the flag form: --status, --install, --uninstall.\n` +
`Run \`gbrain autopilot --help\` for full usage.`,
};
}
return { ok: true, args: out };
}
export function isPidAlive(pid: number): boolean {
if (!Number.isFinite(pid) || pid <= 0) return false;
try {
@@ -457,11 +363,6 @@ export async function runAutopilot(engine: BrainEngine, args: string[]) {
' gbrain autopilot --install [--repo <path>]\n' +
' gbrain autopilot --uninstall\n' +
' gbrain autopilot --status [--json]\n\n' +
'Subcommand aliases:\n' +
' gbrain autopilot status → --status\n' +
' gbrain autopilot install → --install\n' +
' gbrain autopilot uninstall → --uninstall\n' +
' gbrain autopilot start → (default daemon launch)\n\n' +
'Self-maintaining brain daemon. Runs the full maintenance cycle\n' +
'(lint + backlinks + sync + extract + embed + orphans) on an interval.\n\n' +
'For a one-shot cron-triggered cycle, see `gbrain dream`.',
@@ -469,16 +370,6 @@ export async function runAutopilot(engine: BrainEngine, args: string[]) {
return;
}
// #1525: translate positional subcommands to their flag form BEFORE any
// side effect (lockfile, daemon spawn, sync dispatch). Unknown positionals
// fail loud here rather than silently starting the daemon.
const translated = translatePositionalSubcommands(args);
if (!translated.ok) {
console.error(translated.message);
process.exit(2);
}
args = translated.args;
if (args.includes('--install')) {
await installDaemon(engine, args);
return;
+9 -1
View File
@@ -1743,7 +1743,15 @@ async function extractStaleFromDB(
// `page.updated_at.toISOString()` — the JS Date is ms-truncated, so the
// µs-precision DB updated_at stayed strictly greater and the page never
// cleared on Postgres. Stamping the exact value makes them equal.
processedRefs.push({ slug: page.slug, source_id: page.source_id, extractedAt: page.updated_at_iso });
//
// Version-arm floor: a page last edited BEFORE LINK_EXTRACTOR_VERSION_TS
// would otherwise be stamped below the version watermark and stay
// permanently stale (`links_extracted_at < versionTs` re-fires every run).
// Stamp max(updated_at, versionTs) — versionTs is always a past release
// date, so a concurrent edit's now() still exceeds the stamp and D4 holds.
// Tie at ms precision picks updated_at_iso (its µs ≥ versionTs's .000000).
const stampTs = new Date(page.updated_at_iso) >= new Date(versionTs) ? page.updated_at_iso : versionTs;
processedRefs.push({ slug: page.slug, source_id: page.source_id, extractedAt: stampTs });
}
// Flush NON-swallowing (CDX-4): a throw here propagates out of the sweep so
+13 -4
View File
@@ -28,7 +28,7 @@ import { ensureWellFormed } from './text-safe.ts';
* OR updated_at > links_extracted_at`. It is an ISO-8601 string (NOT a number) —
* the column is TIMESTAMPTZ and the predicate binds it as `::timestamptz`.
*/
export const LINK_EXTRACTOR_VERSION_TS = '2026-05-31T00:00:00Z';
export const LINK_EXTRACTOR_VERSION_TS = '2026-07-21T00:00:00Z';
// ─── Entity references ──────────────────────────────────────────
@@ -80,10 +80,10 @@ export type LinkResolutionType = 'qualified' | 'unqualified';
* Directory prefix whitelist. These are the top-level slug dirs the extractor
* recognizes as entity references. Upstream canonical + our extensions:
* - Gbrain canonical: people, companies, meetings, concepts, deal, civic, project, source, media, yc, projects
* - Our domain extensions: tech, finance, personal, openclaw (domain-organized wikis)
* - Our domain extensions: tech, finance, personal, openclaw, ops (domain-organized wikis)
* - Our entity prefix: entities (we kept some legacy entities/projects/ pages)
*/
const DIR_PATTERN = '(?:people|companies|meetings|concepts|deal|civic|project|projects|source|media|yc|tech|finance|personal|openclaw|entities)';
const DIR_PATTERN = '(?:people|companies|meetings|concepts|deal|civic|project|projects|source|media|yc|tech|finance|personal|openclaw|entities|ops)';
/**
* Match `[Name](path)` markdown links pointing to entity directories.
@@ -865,7 +865,16 @@ export function queryBasenameIndex(idx: Map<string, string[]>, name: string): st
if (!name || typeof name !== 'string') return [];
const trimmed = name.trim();
if (!trimmed) return [];
const hit = idx.get(trimmed) ?? idx.get(trimmed.toLowerCase()) ?? idx.get(normalizeBasename(trimmed));
let hit = idx.get(trimmed) ?? idx.get(trimmed.toLowerCase()) ?? idx.get(normalizeBasename(trimmed));
// Issue #2576 bug 2: path-style refs (`runbooks/2026-05-01-x`) from dirs
// outside DIR_PATTERN reach here, but normalizeBasename strips slashes
// into a garbage key (`runbooks2026-05-01-x`) that can never hit the
// tail-keyed index. Fall back to the path tail so qualified refs resolve
// by basename like everything else.
if (!hit && trimmed.includes('/')) {
const tail = trimmed.slice(trimmed.lastIndexOf('/') + 1).trim();
if (tail) hit = idx.get(tail) ?? idx.get(tail.toLowerCase()) ?? idx.get(normalizeBasename(tail));
}
return hit ? [...hit].sort(basenameSort) : [];
}
@@ -1,197 +0,0 @@
/**
* Tests for translatePositionalSubcommands() — the v0.41.x #1525 fix that
* prevents `gbrain autopilot status` from silently starting the daemon.
*
* IRON RULE regression guard: the exact ticket repro (`gbrain autopilot
* status`) MUST translate to `--status`, not fall through to the default
* daemon launch. Verified by the "ticket-exact repro" case below.
*/
import { describe, test, expect } from 'bun:test';
import { translatePositionalSubcommands } from '../src/commands/autopilot.ts';
describe('translatePositionalSubcommands — known aliases', () => {
test('IRON RULE — `autopilot status` translates to `--status` (ticket #1525 repro)', () => {
const r = translatePositionalSubcommands(['status']);
expect(r.ok).toBe(true);
if (r.ok) expect(r.args).toEqual(['--status']);
});
test('`install` translates to `--install`', () => {
const r = translatePositionalSubcommands(['install']);
expect(r.ok).toBe(true);
if (r.ok) expect(r.args).toEqual(['--install']);
});
test('`uninstall` translates to `--uninstall`', () => {
const r = translatePositionalSubcommands(['uninstall']);
expect(r.ok).toBe(true);
if (r.ok) expect(r.args).toEqual(['--uninstall']);
});
test('`start` drops the positional (default daemon launch)', () => {
const r = translatePositionalSubcommands(['start']);
expect(r.ok).toBe(true);
if (r.ok) expect(r.args).toEqual([]);
});
test('`start --json` drops only the positional, keeps the flag', () => {
const r = translatePositionalSubcommands(['start', '--json']);
expect(r.ok).toBe(true);
if (r.ok) expect(r.args).toEqual(['--json']);
});
});
describe('translatePositionalSubcommands — flag/positional interleaving', () => {
test('`status --json` preserves the trailing flag', () => {
const r = translatePositionalSubcommands(['status', '--json']);
expect(r.ok).toBe(true);
if (r.ok) expect(r.args).toEqual(['--status', '--json']);
});
test('`--json status` preserves the leading flag', () => {
const r = translatePositionalSubcommands(['--json', 'status']);
expect(r.ok).toBe(true);
if (r.ok) expect(r.args).toEqual(['--json', '--status']);
});
test('`--repo /foo status` does not mis-classify the path as positional', () => {
const r = translatePositionalSubcommands(['--repo', '/foo', 'status']);
expect(r.ok).toBe(true);
if (r.ok) expect(r.args).toEqual(['--repo', '/foo', '--status']);
});
test('`--interval 300 install` does not mis-classify the number as positional', () => {
const r = translatePositionalSubcommands(['--interval', '300', 'install']);
expect(r.ok).toBe(true);
if (r.ok) expect(r.args).toEqual(['--interval', '300', '--install']);
});
test('`--install --target linux-cron` does not mis-classify the target as positional', () => {
// --target is installDaemon's value flag; its value must never be read
// as a positional subcommand (regression guard for the review fix).
const r = translatePositionalSubcommands(['--install', '--target', 'linux-cron']);
expect(r.ok).toBe(true);
if (r.ok) expect(r.args).toEqual(['--install', '--target', 'linux-cron']);
});
test('`install --target macos` keeps the alias translation and the target value', () => {
const r = translatePositionalSubcommands(['install', '--target', 'macos']);
expect(r.ok).toBe(true);
if (r.ok) expect(r.args).toEqual(['--install', '--target', 'macos']);
});
test('value-flag at end of argv with missing value passes through (so parseArg can report it)', () => {
const r = translatePositionalSubcommands(['--repo']);
expect(r.ok).toBe(true);
if (r.ok) expect(r.args).toEqual(['--repo']);
});
test('value-flag whose value looks like an alias is NOT translated', () => {
// `--repo status` means "use repo path 'status'", not "show status".
// Translator must not destructure the value of --repo.
const r = translatePositionalSubcommands(['--repo', 'status']);
expect(r.ok).toBe(true);
if (r.ok) expect(r.args).toEqual(['--repo', 'status']);
});
});
describe('translatePositionalSubcommands — pass-through cases', () => {
test('empty args returns empty args', () => {
const r = translatePositionalSubcommands([]);
expect(r.ok).toBe(true);
if (r.ok) expect(r.args).toEqual([]);
});
test('flag-only invocation passes through unchanged', () => {
const r = translatePositionalSubcommands(['--status', '--json']);
expect(r.ok).toBe(true);
if (r.ok) expect(r.args).toEqual(['--status', '--json']);
});
test('short flag `-h` passes through unchanged', () => {
const r = translatePositionalSubcommands(['-h']);
expect(r.ok).toBe(true);
if (r.ok) expect(r.args).toEqual(['-h']);
});
test('all known bare flags pass through unchanged', () => {
const flags = ['--help', '--install', '--uninstall', '--status', '--json', '--inline', '--no-worker'];
const r = translatePositionalSubcommands(flags);
expect(r.ok).toBe(true);
if (r.ok) expect(r.args).toEqual(flags);
});
});
describe('translatePositionalSubcommands — rejection of unknown positionals', () => {
test('unknown positional `foo` fails with reason=unknown_subcommand + structured message', () => {
const r = translatePositionalSubcommands(['foo']);
expect(r.ok).toBe(false);
if (!r.ok) {
expect(r.reason).toBe('unknown_subcommand');
expect(r.message).toContain('Unknown subcommand: `foo`');
expect(r.message).toContain('status');
expect(r.message).toContain('install');
expect(r.message).toContain('uninstall');
expect(r.message).toContain('--help');
}
});
test('unknown positional `stop` fails with reason=unknown_subcommand (NOT silently aliased)', () => {
// Stop is mentioned in the ticket but deliberately NOT aliased in this
// PR — stopping a running daemon is a new behavior, not just an alias.
// Until that feature lands separately, `stop` must fail loud rather
// than starting the daemon (the bug we're fixing).
const r = translatePositionalSubcommands(['stop']);
expect(r.ok).toBe(false);
if (!r.ok) {
expect(r.reason).toBe('unknown_subcommand');
expect(r.message).toContain('Unknown subcommand: `stop`');
}
});
test('unknown positional `status-detail` (close-but-not-matching) fails', () => {
const r = translatePositionalSubcommands(['status-detail']);
expect(r.ok).toBe(false);
if (!r.ok) {
expect(r.reason).toBe('unknown_subcommand');
expect(r.message).toContain('Unknown subcommand: `status-detail`');
}
});
test('multiple positionals fail with reason=multiple_subcommands (`start install`)', () => {
const r = translatePositionalSubcommands(['start', 'install']);
expect(r.ok).toBe(false);
if (!r.ok) {
expect(r.reason).toBe('multiple_subcommands');
expect(r.message).toContain('Multiple subcommands');
}
});
test('multiple positionals fail even when both are known aliases (`status install`)', () => {
const r = translatePositionalSubcommands(['status', 'install']);
expect(r.ok).toBe(false);
if (!r.ok) {
expect(r.reason).toBe('multiple_subcommands');
expect(r.message).toContain('Multiple subcommands');
}
});
test('known-then-unknown rejects with multiple_subcommands (first-positional-wins)', () => {
// First positional is known, second is not. Rejection comes from the
// multiple-positional rule, which fires before the unknown check; the
// intent is "only one subcommand allowed."
const r = translatePositionalSubcommands(['status', 'garbage']);
expect(r.ok).toBe(false);
if (!r.ok) expect(r.reason).toBe('multiple_subcommands');
});
test('unknown-then-known rejects on the unknown (unknown fires before second-positional check)', () => {
const r = translatePositionalSubcommands(['garbage', 'status']);
expect(r.ok).toBe(false);
if (!r.ok) {
expect(r.reason).toBe('unknown_subcommand');
expect(r.message).toContain('garbage');
}
});
});
+18
View File
@@ -209,6 +209,24 @@ describe('gbrain extract --stale', () => {
expect(usRows[0]?.eq).toBe(true);
});
test('version-arm floor: page edited BEFORE LINK_EXTRACTOR_VERSION_TS clears after --stale (issue #2576 bug 3)', async () => {
// A page whose updated_at predates the version watermark used to be
// stamped at its updated_at (< versionTs), so the version arm re-fired
// every run — permanently stale. The sweep now floors the stamp at
// versionTs. (The #1768 test above also covers this since the v0.42.x
// VERSION_TS bump moved its date below the watermark, but this pins the
// behavior explicitly so a date "repair" there can't drop coverage.)
await engine.putPage('people/alice', personPage('Alice'));
await engine.executeRaw(`UPDATE pages SET updated_at = '2000-01-01T00:00:00Z' WHERE slug = 'people/alice'`);
expect(await engine.countStalePagesForExtraction({ versionTs: LINK_EXTRACTOR_VERSION_TS })).toBe(1);
await runExtract(engine, ['--stale']);
// Pre-floor this stayed 1 forever (stamp < versionTs → version arm re-fires).
expect(await engine.countStalePagesForExtraction({ versionTs: LINK_EXTRACTOR_VERSION_TS })).toBe(0);
await runExtract(engine, ['--stale']);
expect(await engine.countStalePagesForExtraction({ versionTs: LINK_EXTRACTOR_VERSION_TS })).toBe(0);
});
test('CDX-4 (D2): a link-flush throw aborts the sweep and leaves pages UNSTAMPED', async () => {
await engine.putPage('people/alice', personPage('Alice'));
await engine.putPage('companies/acme', companyPage('Acme', '[Alice](people/alice) founded [Acme](companies/acme).'));
+18
View File
@@ -140,6 +140,15 @@ describe('extractEntityRefs', () => {
expect(wikiRefs[0].needsResolution).toBe(true);
});
test('recognizes ops/ qualified wikilinks (issue #2576 bug 2)', () => {
// `ops` was missing from DIR_PATTERN, so [[ops/...]] fell through to
// the generic 2c pass (needsResolution) instead of being a real ref.
const refs = extractEntityRefs('Deployed via [[ops/services/pointer-agent]].');
expect(refs.length).toBe(1);
expect(refs[0].slug).toBe('ops/services/pointer-agent');
expect(refs[0].needsResolution).toBeUndefined();
});
test('skips qualified-syntax tokens (those belong to 2a)', () => {
// [[wiki:topics/ai]] looks like 2a's qualified shape — even though
// it wouldn't satisfy DIR_PATTERN, 2c must not claim it either
@@ -1069,6 +1078,15 @@ describe('makeResolver — fallback chain', () => {
]);
});
test('resolveBasenameMatches: path-style ref falls back to the tail (issue #2576 bug 2)', async () => {
// normalizeBasename strips slashes, so `runbooks/2026-05-01-pointer-agent`
// used to normalize to a garbage key that never hit the tail-keyed index.
const engine = makeFakeEngineWithSlugs(['ops/changes/2026-05-01-pointer-agent']);
const r = makeResolver(engine);
expect(await r.resolveBasenameMatches!('runbooks/2026-05-01-pointer-agent'))
.toEqual(['ops/changes/2026-05-01-pointer-agent']);
});
test('resolveBasenameMatches: case-insensitive fallback', async () => {
const engine = makeFakeEngineWithSlugs(['companies/fast-weigh']);
const r = makeResolver(engine);