🤖 tests: gate the fg-to-bg migration tests on a release file, not sleep 2

[ThomasK33]

Summary

The three foreground-to-background migration tests in tests/ipc/runtime/backgroundBashDirect.test.ts relied on a fixed sleep 2: the command had to keep running for 2 s while the bash tool claimed a record name and migrated it. On Windows CI that migration can take longer, so the command finished first and the tool correctly took the normal completion path (#4878). Now each script waits for a release file that the test creates after it has checked the migration. The test then waits for the process to exit instead of sleeping 2.5 s. This is a test-only change.

Fixes #5593

Not a regression from #5579 or #5592

The same test failed on Windows before #5579 (merged 2026-10-03):

• 2026-09-26, run 36272619908 (🤖 fix: render and send in the VS Code webview after workspace selection #4740, merge queue), logged on 🤖 tests: ModelsSection discovery and Bedrock endpoint-pinning tests flake in the merge queue #4463.
• 2026-09-27, run 36284062610 (🤖 fix: guarded synthetic wakes skip on-send compaction #4794), logged on 🤖 tests: ModelsSection discovery and Bedrock endpoint-pinning tests flake in the merge queue #4463. In this one, the AFTER_BG line was already in the output when the migration returned, which is the same timing problem.

#5579 and #5592 only change the refused and failed migration paths. The successful migration path that these tests use is unchanged.

Root cause

In the merge-queue failure (run 37154303115), result.success was true but result.backgroundProcessId was undefined (backgroundBashDirect.test.ts:502). So the tool returned a normal completion. bash.ts first awaits the migration's name claim, then checks whether the command already exited, and if so completes normally. With a 2 s command, any claim slower than about 2 s produces exactly this result. Jest reported the failure location as afterAll, line 439, but the failing assertion is the one at line 502.

Validation

• Reproduced on Linux with a probe (not committed) that delays claimMigrationProcessId by 2.5 s, which matches a slow Windows claim. On main, all three tests fail with the CI symptom:

  ● Foreground to Background Migration › should migrate foreground bash to background and continue running
    Expected: "fg_to_bg_…"
    Received: undefined
  ● … › should preserve output across stream boundaries        (Received: undefined)
  ● … › should not kill backgrounded process when abort signal fires (Received: undefined)
  

  With this change and the same probe, all four tests in that describe pass.

• Without the probe, the whole file passes (11 tests, bun x jest, Bun 1.3.12). make static-check passes.

• Windows: the gate path uses forward slashes (C:/…), which Git Bash accepts. The scripts never escape backslashes.

---

Generated with xum • Model: anthropic:claude-opus-5-5 • Thinking: high • Cost: $47.22

[chatgpt-codex-connector]

Codex Review Summary

This comment shows the latest Codex review activity on this pull request.

Review	Status	Commit	Review trigger
📝 Code Review	✅ Completed 2026-10-03T22:33:50.801034Z	e6db164	PR opened
🔒 Security Review	✅ Completed 2026-10-03T22:34:11.807785Z	e6db164	PR opened

ℹ️ About Codex in GitHub

Your team has set up Codex to review pull requests in this repo. Reviews are triggered when you

• Open a pull request for review
• Mark a draft as ready
• Comment "@codex review" or "@codex security review".

Codex reacts with 👀 while any review is running, comments if it has suggestions, and reacts with 👍 once all reviews finish with no findings.

[ThomasK33]

Readiness record for head e6db164:

• CI: all required checks pass on this head (26 pass, 4 skipped). Test / Windows ran backgroundBashDirect.test.ts with the release gate and passed. The only open item is the optional Pixel / Review. Unresolved threads: 0.
• Local: make static-check passes, and the file passes (11 tests). A probe that delays the name claim by 2.5 s makes all three old tests fail with the CI symptom (Received: undefined), and the new tests pass under it.
• Review budget: 3 of 6 assessments (1 normal and 1 security automatic review, both with no findings, plus the final check).
• Final independent check (clean context, on exactly this head): "ready with tracked follow-ups". No blocker. Its follow-up, the slow migration name claim on Windows, is tracked in 🤖 perf: the migration name claim can take over 2 s on Windows #5596.
• Decision: ready. Merging with squash, pinned to this head.

---

Generated with xum • Model: anthropic:claude-opus-5-5 • Thinking: high • Cost: $47.22