Skip to content

Service update recovery tests exceed default 5-second timeout on macOS #7417

Closed

Description

@lidge-jun

Client or integration

Other

Area

Service lifecycle

Summary

Service-to-direct update recovery tests exceed the default 5-second Bun timeout during #7414 validation. These tests and the update implementation are unchanged by the account-pool fix. Record the failures without expanding #7414 into service-update changes.

Reproduction

With the repository Bun 1.4.3 runner and isolated homes, run bun test tests/cli/cli-headless-parity.test.ts tests/update/update-job.test.ts. Observed 171 pass / 4 fail in 36.23 seconds. The first test:changed run also timed out one of these cases.

Version

PR #7414 at e2ab0d7; base eb4bf96

Operating system

macOS arm64

Logs or error output

All four timed out after about 5003–5008ms in tests/update/update-job.test.ts:

  • service reinstall failure falls back to a direct proxy start
  • service reinstall exit 0 with non-viable assets falls back to direct start
  • falls through to a direct start when a viable service never serves
  • still falls back when the service is not viable at all

Screenshots and supporting files

#7414
No real proxy or app was terminated; tests ran in the repository's isolated harness. Load sensitivity versus fixed timeout budget remains to be investigated. Related CLI flake #7416 passed the focused follow-up.

Checks

  • I searched existing issues and documentation.
  • I removed secrets, tokens, account details, request credentials, and personal data.

Activity

  1. github-actions commented on Oct 19, 2026

    @github-actions
    Contributor

    Automated translation bookkeeping — detected language: English.

  2. added
    platformOS/service/tray/ACL (Windows-heavy, not Windows-only)
    serviceService lifecycle (WinSW/launchd/scheduler)
    on Oct 19, 2026
  3. Ingwannu commented on Oct 20, 2026

    @Ingwannu
    Owner

    Checked against trusted dev 12025981e5a304cdb6b1b54f525ba8554f623977. Keep the four service-to-direct recovery timeouts open. These tests exercise fallback ownership and cleanup; a larger global timeout or skipping them would conceal a regression. A focused macOS investigation should identify which mocked readiness/settle deadline remains pending and make that clock/child completion deterministic where possible, retaining the refusal and viable-service controls. The account-pool integration did not repair these cases, and no macOS execution is certified here.

  4. lidge-jun commented on Oct 22, 2026

    @lidge-jun
    OwnerAuthor

    Fixed on dev by #7468 (944647100c), which shipped in 2.91.0: the service-to-direct recovery cases in update-job.test.ts no longer run the host's real listener scan (lsof on macOS), which is what pushed them past the 5-second default. Closing manually because GitHub only auto-closes from main; reopen if it recurs on current dev.

Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment

Metadata

Metadata

Assignees

No one assigned

    Labels

    bugSomething isn't workingplatformOS/service/tray/ACL (Windows-heavy, not Windows-only)serviceService lifecycle (WinSW/launchd/scheduler)

    Projects

    No projects

      Milestone

      No milestone

      Relationships

      None yet

      Development

      No branches or pull requests

      Issue actions