fix(serve): boot probe uses resolvePgserveTransport (matches connection probe)

[namastex888]

Summary

Cosmetic followup to #1667. Pre-this-fix, genie serve boot called requirePgserveDaemon() (UDS-only) and printed a misleading "pgserve unreachable" warning on hosts where pgserve install registered foreground TCP mode — even though real connections then succeeded via the resolver's TCP fallback. Operators read the warning, assumed something was broken, wasted time on the wrong fix.

What changed

-async function requirePgserveReady(): Promise<void> {
-  console.log('  Probing canonical pgserve daemon...');
-  try {
-    const { requirePgserveDaemon, resolvePgserveSocketDir } = await import('../lib/db.js');
-    await requirePgserveDaemon();
-    console.log(`  pgserve daemon ready (canonical, pm2-supervised) on ${resolvePgserveSocketDir()}`);
+async function requirePgserveReady(): Promise<void> {
+  console.log('  Probing pgserve transport...');
+  try {
+    const { resolvePgserveTransport } = await import('../lib/db.js');
+    const transport = await resolvePgserveTransport();
+    if (transport.kind === 'unix') {
+      console.log(`  pgserve ready: unix socket ${transport.socketDir}/.s.PGSQL.${transport.port}`);
+    } else {
+      console.log(`  pgserve ready: tcp ${transport.host}:${transport.port}`);
+    }

Same retry-guard semantics on failure (both GENIE_PG_NO_AUTOSTART env vars set, no exit). The misleading legacy "Recovery: pm2 status / pm2 restart pgserve / pgserve install" lines drop because the resolver throws a richer message that already mentions both probe attempts and recovery steps.

After this ships

Fresh genie update + pm2 restart genie-serve on a TCP-only host will show:

  Probing pgserve transport...
  pgserve ready: tcp 127.0.0.1:8432

…instead of the noisy "unreachable" warning. No operator action required.

Diagnostic context

While the user was dogfooding the post-#1667 binary, their ~/.genie/logs/update-diagnostics-2026-05-06T15-20-52-347Z.json showed 141 scheduler errors over 30 min: 7 distinct event types (process_cycle_error, agent_resume_timer_error, mailbox_retry_error, heartbeat_error, lease_recovery_error, orphan_reconciliation_error, retention_error), each with the legacy "pgserve canonical daemon is not reachable" hint. Investigation showed those errors were all from the pre-#1667 daemon process still running pre-restart — once the daemon picks up the new code via pm2 restart genie-serve, every error path goes through _buildConnection() → resolvePgserveTransport() and succeeds via TCP fallback. No scheduler-side change needed in this PR.

Validation

• ✅ bun run typecheck
• ✅ bun run lint (2 pre-existing test-fixture symlink warnings)
• ✅ bun test src/term-commands/serve.test.ts — 19 pass / 0 fail / 71 expects

Test plan

• CI: typecheck + lint + dead-code + tests on dev
• On a TCP-only pgserve host, run genie serve → confirm pgserve ready: tcp ... banner instead of legacy "unreachable" warning
• On a UDS host (running pgserve daemon standalone), confirm pgserve ready: unix socket ... banner
• On a host with neither, confirm retry-guard still fires (GENIE_PG_NO_AUTOSTART=1 set)

Related

• #1665 — update-unify-stages G1-G5 (merged)
• #1667 — pgserve transport discovery (merged)
• #1668 — SHARED-DESIGN.md §4.6 sync (merged)
• #1671 — skills .orphaned_at regression fix
• THIS PR — boot-probe migration

🤖 Generated with Claude Code

[coderabbitai]

Important

Review skipped

Auto reviews are disabled on base/target branches other than the default branch.

Please check the settings in the CodeRabbit UI or the .coderabbit.yaml file in this repository. To trigger a single review, invoke the @coderabbitai review command.

⚙️ Run configuration

Configuration used: Path: .coderabbit.yaml

Review profile: ASSERTIVE

Plan: Pro

Run ID: 478f7ccf-cca2-4ba0-bc02-d1d8a407e66f

You can disable this status message by setting the reviews.review_status to false in the CodeRabbit configuration file.

Use the checkbox below for a quick retry:

• 🔍 Trigger review

✨ Finishing Touches

🧪 Generate unit tests (beta)

• Create PR with unit tests
• Commit unit tests in branch fix/serve-boot-probe-uses-new-resolver

---

Thanks for using CodeRabbit! It's free for OSS, and your support helps us grow. If you like it, consider giving us a shout-out.

❤️ Share

• X
• Mastodon
• Reddit
• LinkedIn

_{Comment @coderabbitai help to get the list of available commands and usage tips.}

[gemini-code-assist]

Code Review

This pull request updates the requirePgserveReady function to utilize the unified resolvePgserveTransport resolver, which supports both Unix Domain Sockets and TCP fallbacks. This change fixes a bug where TCP-only installations would trigger a misleading 'unreachable' error message. The associated tests have been updated to reflect this new transport discovery logic. A review comment suggests improving the formatting of multi-line error messages by ensuring consistent indentation when they are printed to the console.

I am having trouble creating individual review comments. Click here to see my feedback.

src/term-commands/serve.ts (635)

The error message returned by resolvePgserveTransport (via buildBothTransportsUnavailableHint) is a multi-line string containing bullet points and recovery steps. When printed directly with a single prefix, subsequent lines will lose their indentation, making the output harder to read and inconsistent with the rest of the boot logs. Indenting all lines of the message ensures the visual structure is preserved.

    console.error("  pgserve unreachable: " + msg.replace(/\n/g, "\n  "));