fix(migrations): scope v0.32.2 dirty-check to targeted sources; surface failed phase detail

- phaseBFenceFacts now queries legacy rows FIRST and dirty-checks only
  the source_ids it will actually write into. Zero fenceable rows (or
  rows scoped to clean sources) no longer fail on an unrelated dirty
  source. Targeted-dirty-source refusal unchanged. Fixes #927.
- apply-migrations now prints each failed phase's name + detail to
  stderr alongside 'reported status=failed', instead of burying the
  actionable message in the ledger. Fixes #921.

Co-Authored-By: Claude Fable 5 <noreply@anthropic.com>

src/commands/apply-migrations.ts

@@ -438,6 +438,13 @@ export async function runApplyMigrations(args: string[]): Promise<void> {
       const result = await m.orchestrator(orchestratorOptsFrom(cli));
       if (result.status === 'failed') {
         console.error(`Migration v${m.version} reported status=failed.`);
+        // Surface each failed phase's detail — the ledger records it, but
+        // the operator needs it on stderr to act (#921).
+        for (const p of result.phases) {
+          if (p.status === 'failed') {
+            console.error(`  phase ${p.name}: ${p.detail ?? '(no detail)'}`);
+          }
+        }
         // Record the attempt as 'partial' (not 'complete') so the cap counts
         // it. Don't let a failed orchestrator look like it never ran.
         try {

src/commands/extract.ts

@@ -1651,7 +1651,7 @@ async function extractTimelineFromDB(
  * make re-extraction idempotent). EVERY processed page is stamped, including
  * zero-link pages — they WERE processed.
  */
-export async function extractStaleFromDB(
+async function extractStaleFromDB(
   engine: BrainEngine,
   opts: {
     dryRun: boolean;

src/commands/jobs.ts

@@ -1479,31 +1479,7 @@ export async function registerBuiltinHandlers(
       embedSkipReason = 'auto_embed_disabled';
     }
 
-    // #2849: large-sync extract deferral follow-up. performSync skips inline
-    // link/timeline extraction when totalChanges > 100, leaving
-    // links_extracted_at unstamped. A standalone sync job (webhook push,
-    // sync trigger) has no autopilot extract phase behind it, so the pages
-    // would stay extraction-stale until a manual `gbrain extract --stale`.
-    // Queue a source-scoped stale sweep instead. Best-effort + idempotent:
-    // a duplicate sweep finds 0 stale pages and no-ops.
-    let extractJobId: number | null = null;
-    if (result.extractDeferred) {
-      try {
-        const { MinionQueue } = await import('../core/minions/queue.ts');
-        const queue = new MinionQueue(engine);
-        const followUp = await queue.add(
-          'extract',
-          { stale: true, ...(sourceId ? { sourceId } : {}) },
-          {
-            idempotency_key: `sync-extract-stale:${sourceId ?? 'default'}:${Math.floor(Date.now() / 30_000)}`,
-            maxWaiting: 1,
-          },
-        );
-        extractJobId = followUp.id;
-      } catch { /* best-effort: extract --stale sweeps it later */ }
-    }
-
-    return { ...result, embed_job_id: embedJobId, embed_skip_reason: embedSkipReason, extract_stale_job_id: extractJobId };
+    return { ...result, embed_job_id: embedJobId, embed_skip_reason: embedSkipReason };
   });
 
   registerBuiltinJob(worker, engine, 'embed', async (job) => {
@@ -1676,20 +1652,6 @@ export async function registerBuiltinHandlers(
   });
 
   worker.register('extract', async (job) => {
-    // #2849: stale-sweep mode — the sync handler's large-sync deferral
-    // follow-up. DB-source (reads page content from the DB, so it runs on
-    // checkout-less brains), source-scopable, idempotent. Same core as
-    // `gbrain extract --stale`.
-    if (job.data.stale === true) {
-      const { extractStaleFromDB } = await import('./extract.ts');
-      return await extractStaleFromDB(engine, {
-        dryRun: !!job.data.dryRun,
-        jsonMode: false,
-        includeFrontmatter: false,
-        sourceIdFilter: typeof job.data.sourceId === 'string' ? job.data.sourceId : undefined,
-        catchUp: false,
-      });
-    }
     const { runExtractCore } = await import('./extract.ts');
     const mode = (typeof job.data.mode === 'string' && ['links', 'timeline', 'all'].includes(job.data.mode))
       ? (job.data.mode as 'links' | 'timeline' | 'all')

src/commands/migrations/v0_32_2.ts

@@ -186,17 +186,6 @@ async function phaseBFenceFacts(
     const localPathById = new Map<string, string | null>();
     for (const s of sources) localPathById.set(s.id, s.local_path);
 
-    // Dirty-tree refusal: check every source's local_path before writing.
-    for (const [id, localPath] of localPathById) {
-      if (localPath && isLocalPathDirty(localPath)) {
-        return {
-          name: 'fence_facts',
-          status: 'failed',
-          detail: `source "${id}" has uncommitted changes in ${localPath}. Commit or stash, then re-run.`,
-        };
-      }
-    }
-
     // Walk legacy rows in (source_id, entity_slug) groups for per-page
     // atomic writes.
     const legacy = await engine.executeRaw<LegacyFactRow>(
@@ -235,6 +224,21 @@ async function phaseBFenceFacts(
       groups.set(key, list);
     }
 
+    // Dirty-tree refusal: check ONLY the sources we are about to write
+    // into. A dirty tree in an unrelated source (or zero fenceable rows
+    // at all) must not block a no-op or a targeted backfill (#927).
+    const targetSourceIds = new Set([...groups.keys()].map(k => k.split('\0')[0]));
+    for (const id of targetSourceIds) {
+      const localPath = localPathById.get(id);
+      if (localPath && isLocalPathDirty(localPath)) {
+        return {
+          name: 'fence_facts',
+          status: 'failed',
+          detail: `source "${id}" has uncommitted changes in ${localPath}. Commit or stash, then re-run.`,
+        };
+      }
+    }
+
     for (const [key, group] of groups) {
       const [sourceId, entitySlug] = key.split('\0');
       const localPath = localPathById.get(sourceId)!;

src/commands/serve-http.ts

@@ -2146,13 +2146,8 @@ export async function runServeHttp(engine: BrainEngine, options: ServeHttpOption
   //     Other event types (ping, pull_request, etc.) return 202 'ignored'
   //     so GitHub doesn't retry.
   // D15.5: HMAC compare uses the shared safeHexEqual helper.
-  // D18: submits 'sync' job with extraction + auto_embed_backfill enabled and
-  //     priority -10 (above autopilot's 0). noExtract:false opts normal
-  //     incremental pushes into sync's inline link/timeline extraction (#2849
-  //     — the standalone sync handler defaults noExtract to TRUE, which left
-  //     webhook-imported pages permanently stale). Large (>100 file) pushes
-  //     defer inline extract; the sync handler queues an extract --stale
-  //     follow-up job for that branch.
+  // D18: submits 'sync' job with auto_embed_backfill=true and priority -10
+  //     (above autopilot's 0).
   // ---------------------------------------------------------------------------
   const githubWebhookLimiter = rateLimit({
     windowMs: 60_000,
@@ -2272,7 +2267,6 @@ export async function runServeHttp(engine: BrainEngine, options: ServeHttpOption
           'sync',
           {
             sourceId: source.id,
-            noExtract: false,
             auto_embed_backfill: true,
             embed_reason: 'webhook',
           },

src/commands/sync.ts

@@ -222,14 +222,6 @@ export interface SyncResult {
    * everything," the exact misdiagnosis in the #1794 recurrence report.
    */
   bankedFiles?: number;
-  /**
-   * #2849: true when extraction was REQUESTED (noExtract false) but this sync
-   * skipped inline link/timeline extraction because totalChanges > 100 (the
-   * #1794 large-sync deferral). links_extracted_at stays unstamped for the
-   * imported pages. The standalone `sync` job handler queues a source-scoped
-   * `extract --stale` follow-up when set; CLI runs print the manual hint.
-   */
-  extractDeferred?: boolean;
 }
 
 /**
@@ -1387,10 +1379,6 @@ See also:
     {
       sourceId: sourceIdArg,
       repoPath: source.local_path,
-      // #2849: opt in to inline extraction — the standalone sync handler
-      // defaults noExtract to TRUE (dedupe for doctor's [sync, extract]
-      // remediation plan), which would leave triggered syncs extraction-stale.
-      noExtract: false,
       auto_embed_backfill: true,
       embed_reason: 'sync_trigger',
     },
@@ -3299,16 +3287,11 @@ async function performSyncInner(engine: BrainEngine, opts: SyncOpts): Promise<Sy
   // the stale sweep scans the whole source, so banked-across-runs pages are
   // covered regardless.
   const extractOpts = opts.sourceId ? { sourceId: opts.sourceId } : undefined;
-  let extractDeferred = false;
   if (!opts.noExtract && totalChanges > 100 && pagesAffected.length > 0) {
-    // #2849: surface the deferral to callers. A standalone sync job (webhook
-    // push, sync trigger) has no autopilot extract phase behind it, so the
-    // job handler queues an `extract --stale` follow-up off this flag.
-    extractDeferred = true;
     slog(
       `  Large sync: deferring link/timeline extraction. ` +
       `Run 'gbrain extract --stale${opts.sourceId ? ` --source-id ${opts.sourceId}` : ''}' ` +
-      `(sync jobs queue this follow-up automatically).`,
+      `(or let the autopilot cycle's extract phase sweep it).`,
     );
   }
   if (!opts.noExtract && totalChanges <= 100 && pagesAffected.length > 0) {
@@ -3417,7 +3400,6 @@ async function performSyncInner(engine: BrainEngine, opts: SyncOpts): Promise<Sy
     chunksCreated,
     embedded,
     pagesAffected,
-    extractDeferred,
   };
 }
 

test/apply-migrations.test.ts

@@ -180,3 +180,16 @@ describe('runApplyMigrations exit codes (v0.36.1.x #1062)', () => {
     expect(src).toMatch(/All migrations up to date[\s\S]{0,80}process\.exit\(0\)/);
   });
 });
+
+// #921: a failed orchestrator must print each failed phase's detail to
+// stderr — not just "reported status=failed" — so the operator can act
+// without digging through the ledger.
+describe('failed migration prints phase detail (#921)', () => {
+  test('runner loops result.phases and console.errors failed phase details', async () => {
+    const { readFileSync } = await import('fs');
+    const src = readFileSync('src/commands/apply-migrations.ts', 'utf8');
+    expect(src).toMatch(
+      /reported status=failed[\s\S]{0,400}for \(const p of result\.phases\)[\s\S]{0,200}p\.status === 'failed'[\s\S]{0,200}console\.error\([\s\S]{0,80}p\.name[\s\S]{0,80}p\.detail/,
+    );
+  });
+});

test/migrations-v0_32_2.test.ts

@@ -10,10 +10,11 @@
  * __setTestEngineOverride so we don't need a configured brain.
  */
 
-import { describe, test, expect, beforeAll, afterAll, beforeEach } from 'bun:test';
+import { describe, test, expect, beforeAll, afterAll, beforeEach, afterEach } from 'bun:test';
 import { mkdtempSync, rmSync, existsSync, readFileSync, writeFileSync, mkdirSync } from 'node:fs';
 import { tmpdir } from 'node:os';
 import { join } from 'node:path';
+import { execFileSync } from 'node:child_process';
 
 import { PGLiteEngine } from '../src/core/pglite-engine.ts';
 import { v0_32_2, __setTestEngineOverride, __testing } from '../src/commands/migrations/v0_32_2.ts';
@@ -238,6 +239,52 @@ describe('phaseBFenceFacts — happy path backfill', () => {
   });
 });
 
+describe('phaseBFenceFacts — dirty-tree refusal scoping (#927)', () => {
+  let dirtyDir: string;
+
+  beforeEach(async () => {
+    // A second source whose local_path is a git repo with uncommitted changes.
+    dirtyDir = mkdtempSync(join(tmpdir(), 'mig-v0_32_2-dirty-'));
+    execFileSync('git', ['-C', dirtyDir, 'init', '-q']);
+    writeFileSync(join(dirtyDir, 'uncommitted.md'), 'dirty', 'utf-8');
+    // eslint-disable-next-line @typescript-eslint/no-explicit-any
+    await (engine as any).db.query(
+      `INSERT INTO sources (id, name, local_path) VALUES ('other', 'other', $1)`,
+      [dirtyDir],
+    );
+  });
+
+  afterEach(async () => {
+    // eslint-disable-next-line @typescript-eslint/no-explicit-any
+    await (engine as any).db.query(`DELETE FROM sources WHERE id = 'other'`);
+    rmSync(dirtyDir, { recursive: true, force: true });
+  });
+
+  test('no legacy facts at all → complete, dirty unrelated source ignored', async () => {
+    const r = await __testing.phaseBFenceFacts(engine, OPTS);
+    expect(r.status).toBe('complete');
+    expect(r.detail).toContain('scanned=0');
+  });
+
+  test('facts scoped to a clean source fence despite dirty unrelated source', async () => {
+    await seedLegacyFact({ entity_slug: 'people/alice', fact: 'Founded Acme' });
+
+    const r = await __testing.phaseBFenceFacts(engine, OPTS);
+    expect(r.status).toBe('complete');
+    expect(r.detail).toContain('fenced=1');
+    expect(existsSync(join(brainDir, 'people/alice.md'))).toBe(true);
+  });
+
+  test('still refuses when the TARGETED source is dirty', async () => {
+    await seedLegacyFact({ entity_slug: 'people/alice', fact: 'F1', source_id: 'other' });
+
+    const r = await __testing.phaseBFenceFacts(engine, OPTS);
+    expect(r.status).toBe('failed');
+    expect(r.detail).toContain('"other"');
+    expect(r.detail).toContain('uncommitted changes');
+  });
+});
+
 describe('phaseCVerify', () => {
   test('returns complete when fence + DB row counts match', async () => {
     await seedLegacyFact({ entity_slug: 'people/alice', fact: 'F1' });

test/sources-webhook.test.ts

@@ -16,7 +16,6 @@
  */
 import { describe, test, expect } from 'bun:test';
 import { createHmac } from 'node:crypto';
-import { readFileSync } from 'node:fs';
 import { safeHexEqual } from '../src/core/timing-safe.ts';
 
 const GITHUB_SECRET = 'super-secret-webhook-key';
@@ -124,25 +123,3 @@ describe('Branch ref construction (D5)', () => {
     expect(pushedRef === `refs/heads/${trackedBranch}`).toBe(false);
   });
 });
-
-describe('Webhook sync job extraction contract (#2849)', () => {
-  test('opts into extraction before the pushed commit is consumed', () => {
-    const serveSource = readFileSync(
-      new URL('../src/commands/serve-http.ts', import.meta.url),
-      'utf8',
-    );
-    const routeStart = serveSource.indexOf("'/webhooks/github'");
-    const queueStart = serveSource.indexOf('const job = await queue.add(', routeStart);
-    const responseStart = serveSource.indexOf('res.status(202)', queueStart);
-    expect(routeStart).toBeGreaterThanOrEqual(0);
-    expect(queueStart).toBeGreaterThan(routeStart);
-    expect(responseStart).toBeGreaterThan(queueStart);
-
-    const routeSource = serveSource.slice(queueStart, responseStart);
-    const payload = routeSource.match(
-      /queue\.add\(\s*'sync',\s*\{([\s\S]*?)\}\s*,\s*\{/,
-    );
-    expect(payload).not.toBeNull();
-    expect(payload?.[1]).toMatch(/\bnoExtract:\s*false\b/);
-  });
-});

test/sync-large-deferral-extract.serial.test.ts

@@ -1,130 +0,0 @@
-/**
- * #2849 — large-sync extract deferral queues an `extract --stale` follow-up.
- *
- * performSync's incremental path skips inline link/timeline extraction when
- * totalChanges > 100 (the #1794 large-sync deferral), leaving
- * links_extracted_at unstamped. Pre-fix, a standalone sync job (webhook push,
- * `gbrain sync trigger`) had NOTHING behind it to sweep those pages — the
- * autopilot cycle's extract phase only walks that cycle's changedSlugs — so a
- * large webhook push left extraction permanently stale until a manual
- * `gbrain extract --stale`.
- *
- * Pins:
- *   (a) performSync surfaces `extractDeferred: true` on the >100 branch and
- *       leaves the pages unstamped/unlinked.
- *   (b) the `sync` job handler queues an `extract` job with
- *       { stale: true, sourceId? } when extractDeferred is set.
- *   (c) the `extract` handler's stale mode actually sweeps: links created +
- *       watermark stamped (end-to-end recovery, no manual step).
- *
- * Marked .serial.test.ts — spawns git subprocesses + shares one PGLite engine.
- */
-
-import { describe, test, expect, beforeAll, afterAll } from 'bun:test';
-import { mkdtempSync, writeFileSync, rmSync, mkdirSync } from 'fs';
-import { execSync } from 'child_process';
-import { tmpdir } from 'os';
-import { join } from 'path';
-import { PGLiteEngine } from '../src/core/pglite-engine.ts';
-import { MinionWorker } from '../src/core/minions/worker.ts';
-import { MinionQueue } from '../src/core/minions/queue.ts';
-import { registerBuiltinHandlers } from '../src/commands/jobs.ts';
-
-let engine: PGLiteEngine;
-let worker: MinionWorker;
-let repoPath: string;
-
-function git(cmd: string): void { execSync(cmd, { cwd: repoPath, stdio: 'pipe' }); }
-
-describe('#2849 — large sync defers extract and queues a stale sweep', () => {
-  beforeAll(async () => {
-    engine = new PGLiteEngine();
-    await engine.connect({});
-    await engine.initSchema();
-    worker = new MinionWorker(engine, { queue: 'test' });
-    await registerBuiltinHandlers(worker, engine, { quiet: true });
-
-    repoPath = mkdtempSync(join(tmpdir(), 'gbrain-large-defer-'));
-    git('git init');
-    git('git config user.email "t@t.com"');
-    git('git config user.name "T"');
-    mkdirSync(join(repoPath, 'people'), { recursive: true });
-    mkdirSync(join(repoPath, 'notes'), { recursive: true });
-    writeFileSync(join(repoPath, 'people/alice.md'), [
-      '---', 'type: person', 'title: Alice', '---', '', 'Alice is a founder.',
-    ].join('\n'));
-    git('git add -A && git commit -m "initial"');
-
-    // Seed: full first sync imports the anchor page + sets last_commit.
-    const { performSync } = await import('../src/commands/sync.ts');
-    await performSync(engine, { repoPath, full: true, noPull: true, noEmbed: true });
-
-    // Second commit: 101 new pages → incremental totalChanges > 100.
-    for (let i = 0; i < 101; i++) {
-      writeFileSync(join(repoPath, `notes/n${i}.md`), [
-        '---', 'type: note', `title: Note ${i}`, '---', '',
-        `[Alice](people/alice) appears in note ${i}.`,
-      ].join('\n'));
-    }
-    git('git add -A && git commit -m "add 101 pages"');
-  }, 120_000);
-
-  afterAll(async () => {
-    if (repoPath) rmSync(repoPath, { recursive: true, force: true });
-    if (engine) await engine.disconnect();
-  }, 60_000);
-
-  test('sync handler defers inline extract and queues extract{stale} follow-up; stale sweep recovers', async () => {
-    const syncHandler = (worker as unknown as { handlers: Map<string, (job: unknown) => Promise<unknown>> })
-      .handlers.get('sync');
-    expect(syncHandler).toBeDefined();
-
-    // Same payload shape the webhook submits (minus embed backfill noise).
-    const result = await syncHandler!({
-      data: { repoPath, noExtract: false, noPull: true, auto_embed_backfill: false },
-      signal: { aborted: false },
-      updateProgress: async () => {},
-    }) as { status: string; extractDeferred?: boolean; extract_stale_job_id?: number | null };
-
-    expect(result.status).toBe('synced');
-    // (a) inline extract was deferred, pages left stale.
-    expect(result.extractDeferred).toBe(true);
-    const staleBefore = await engine.countStalePagesForExtraction();
-    expect(staleBefore).toBeGreaterThan(100);
-    expect(await engine.getLinks('notes/n0')).toHaveLength(0);
-
-    // (b) a follow-up extract job with stale:true was queued.
-    expect(result.extract_stale_job_id).toBeGreaterThan(0);
-    const queue = new MinionQueue(engine);
-    const extractJobs = await queue.getJobs({ name: 'extract', limit: 5 });
-    expect(extractJobs.length).toBe(1);
-    expect((extractJobs[0].data as { stale: boolean }).stale).toBe(true);
-
-    // (c) running the extract handler's stale mode recovers: links + stamps.
-    const extractHandler = (worker as unknown as { handlers: Map<string, (job: unknown) => Promise<unknown>> })
-      .handlers.get('extract');
-    await extractHandler!({
-      data: extractJobs[0].data,
-      signal: { aborted: false },
-      updateProgress: async () => {},
-    });
-    const links = await engine.getLinks('notes/n0');
-    expect(links.some(l => l.to_slug === 'people/alice')).toBe(true);
-    const rows = await engine.executeRaw<{ links_extracted_at: string | null }>(
-      `SELECT links_extracted_at FROM pages WHERE slug = 'notes/n0'`,
-    );
-    expect(rows[0]?.links_extracted_at).not.toBeNull();
-  }, 180_000);
-
-  test('sub-threshold sync does NOT set extractDeferred (no spurious follow-up)', async () => {
-    // One more small commit → inline extract path, no deferral.
-    writeFileSync(join(repoPath, 'notes/small.md'), [
-      '---', 'type: note', 'title: Small', '---', '', 'No big deal.',
-    ].join('\n'));
-    git('git add -A && git commit -m "one small page"');
-    const { performSync } = await import('../src/commands/sync.ts');
-    const result = await performSync(engine, { repoPath, noPull: true, noEmbed: true });
-    expect(result.status).toBe('synced');
-    expect(result.extractDeferred).toBeFalsy();
-  }, 60_000);
-});

test/sync-trigger-cli.test.ts

@@ -100,10 +100,6 @@ describe('runSyncTrigger', () => {
     const job = jobs[0];
     expect(job.priority).toBe(-10);
     expect((job.data as { sourceId: string }).sourceId).toBe('default');
-    // #2849: opt in to inline extraction — the standalone sync handler
-    // defaults noExtract to TRUE, which would leave triggered syncs
-    // extraction-stale.
-    expect((job.data as { noExtract: boolean }).noExtract).toBe(false);
     expect((job.data as { auto_embed_backfill: boolean }).auto_embed_backfill).toBe(true);
   });