merge-tracker concurrent write test flaked

Error message

merge-tracker concurrent write test flaked (${e.message}). Retrying once...

What it means

test-all.mjs exercises merge-tracker.mjs under concurrent writes, asserting that no tracker row is lost when two merges race. The first transient failure (e.g. a lost row on a loaded machine) logs this warning and retries once; exhausting retries turns it into a hard failure for merge-tracker's write path.

Solutions

One warning then a pass is the intended tolerance for timing noise — no action needed.
If it fails both attempts, reproduce with two manual merges against the same tracker and inspect merge-tracker.mjs's lock acquisition and atomic rename.
Run on a local disk (not NFS/SMB) where file locking behaves predictably.
Check for stray concurrent test processes (parallel CI shards sharing a working tree) interfering with the lock file.

Example fix

// before (merge-tracker write)
fs.writeFileSync(trackerPath, content);
// after
const lock = acquireLock(trackerPath);
try { fs.writeFileSync(tmpPath, content); fs.renameSync(tmpPath, trackerPath); } finally { lock.release(); }

Defensive patterns

Strategy: retry

Validate before calling

// sanity: single-writer merge must succeed before testing concurrency
if (run('node', ['merge-tracker.mjs']) === null) fail('merge-tracker broken; skip concurrency test');

Try / catch

try { assertRowsPresent(merged, expectedRows); } catch (e) { if (retries-- > 0) warn(`flaked (${e.message}), retrying`); else fail(e.message); }

Prevention

Verify merge-tracker uses a lock file plus atomic rename for writes
Run tests on local disks; network mounts weaken locking
Avoid parallel CI jobs sharing the same repository checkout
Escalate to fail() only after retries are exhausted — one warning is expected noise

When it happens

Trigger: The merged tracker after two concurrent merge runs is missing a row (`concurrent tracker merge lost a row`) or the merge process crashes, on an attempt where retries > 0 — warning is emitted and retries decremented.

Common situations: Slow CI runners widening the race window between read-modify-write cycles, shared/network filesystems with unreliable locking, or a real bug in merge-tracker.mjs's locking or atomic-write implementation (fails every attempt).

Related errors

AI-assisted analysis of santifer/career-ops@e7abd431fc (2026-09-16). Data as JSON: /api/errors/2a397ee0e0aded59. Report an issue: GitHub.

Appendix: source

Thrown at test-all.mjs:13011

      const second = spawnMerge(additionsB, 0);
      const [firstResult, secondResult] = await Promise.all([first.result, second.result]);

      if (firstResult.code === 0 && secondResult.code === 0) {
        pass('concurrent merge processes both exited successfully');
      } else {
        throw new Error(`concurrent merge process failed: first=${firstResult.code} second=${secondResult.code} stderr=${firstResult.stderr || secondResult.stderr}`);
      }

      const merged = readFileSync(join(mergeTmp, 'data', 'applications.md'), 'utf-8');
      if (merged.includes('Alpha') && merged.includes('Beta')) {
        pass('concurrent tracker merges preserve rows from both processes');
      } else {
        throw new Error(`concurrent tracker merge lost a row: ${merged}`);
      }
      break;
    } catch (e) {
      if (retries > 0) {
        warn(`merge-tracker concurrent write test flaked (${e.message}). Retrying once...`);
        retries -= 1;
      } else {
        fail(`merge-tracker concurrent write test crashed: ${e.message}`);
        break;
      }
    } finally {
      rmSync(mergeTmp, { recursive: true, force: true });
    }
  }
} catch (e) {
  fail(`merge-tracker concurrent write test crashed: ${e.message}`);
}

// ── 12. COLD-START TRIGGER ──────────────────────────────────────

console.log('\n12. Cold-start trigger (deterministic onboarding state)');

try {

View on GitHub (pinned to e7abd431fc)