test: nested object literal must not corrupt an adjacent native (WS) handle

[machineloop]

Summary

Adds a regression test for a codegen bug — present on v0.5.1206, fixed since — where a deeply nested object/array literal at module scope corrupted an adjacent native handle: it silently broke node:http WebSocket inbound dispatch globally (the upgrade handler's wsId.on("message", …) stopped firing) while outbound wsId.send(...) kept working. The literal itself read back correctly — the corruption hit a neighbouring allocation (the WS client handle), so the only observable was the dropped inbound frame. A flat, single-level literal was unaffected.

Changes

• crates/perry/tests/nested_object_literal_ws_inbound.rs: a self-testing compile+run regression test, modeled on the existing issue_5174_headers_http_pump_hang.rs. The fixture defines a nested literal of the shape { a: [..], b: [{ kind, child: { a: [..], b: [{ kind }] } }] }, stands up a node:http WS echo server, connects a ws client to it in-process, and asserts the ping → echo:ping round-trip. A healthy build prints WSOK and exits 0; a miscompile drops the inbound frame and hangs — caught by both the Rust wall-clock timeout and the fixture's own 8s exit(1).

Related issue

n/a — standalone regression guard (the underlying bug is already fixed on main).

Test plan

Verified the fixture compiles and completes the WS round-trip on a current main build:

$ perry run nested_literal_ws.ts
...
Running ...
WSOK          # exit 0

• cargo build --release clean (-p perry, off main)
• cargo test --workspace … — the new #[test] follows the established issue_5174 compile+run+timeout pattern; I verified the fixture's behavior directly (above) but left the full workspace test run to CI
• (user-facing) Added a #[test] in the affected crate (crates/perry/tests/)
• (no CLI / stdlib / runtime API changed — docs n/a)
• (no platform UI backend touched)

Screenshots / output

Fixture stdout on a healthy build:

WSOK

Checklist

• I have NOT bumped the workspace version or edited CLAUDE.md / CHANGELOG.md (maintainer handles these at merge)
• My commit follows the loose feat: / fix: / docs: / chore: prefix convention (test:)
• I've read CONTRIBUTING.md and agree to the Code of Conduct

Summary by CodeRabbit

• Tests
  • Added a regression test covering WebSocket inbound message handling with deeply nested object and array literals.
  • Ensures the compiled app echoes received messages correctly and fails safely if execution hangs, using a timeout watchdog.

[coderabbitai]

No actionable comments were generated in the recent review. 🎉

ℹ️ Recent review info

⚙️ Run configuration

Configuration used: defaults

Review profile: CHILL

Plan: Pro Plus

Run ID: 49b9aa5f-e8cd-4bd5-9068-16020d593dd0

📥 Commits

Reviewing files that changed from the base of the PR and between 78f02a1 and 3e7bbc1.

📒 Files selected for processing (1)

• crates/perry/tests/nested_object_literal_ws_inbound.rs

🚧 Files skipped from review as they are similar to previous changes (1)

• crates/perry/tests/nested_object_literal_ws_inbound.rs

---

📝 Walkthrough

Walkthrough

Adds a Rust regression test that compiles an inline TypeScript WebSocket echo program, runs the resulting binary with a timeout watchdog, and asserts the echoed output.

Changes

Nested WebSocket inbound regression test

Layer / File(s)	Summary
Test setup and compile helper crates/perry/tests/nested_object_literal_ws_inbound.rs	Adds the regression-test comment and helper functions for locating perry, writing inline TypeScript into a temp directory, and compiling it to a binary.
Timeout runner crates/perry/tests/nested_object_literal_ws_inbound.rs	Adds a watchdog runner that spawns the compiled binary, captures stdout and stderr, polls for exit, and kills the child on timeout with diagnostic output.
WS echo regression test crates/perry/tests/nested_object_literal_ws_inbound.rs	Adds the test that compiles the nested-literal WebSocket fixture, runs it through the timeout harness, and asserts the echoed stdout.

Estimated code review effort

🎯 2 (Simple) | ⏱️ ~10 minutes

Possibly related PRs

• PerryTS/perry#5534: Touches the HTTP upgrade/WebSocket path that this regression test exercises end to end.

Poem

A bunny hopped through braces deep,
And woke the WebSocket from its sleep.
ping went in, echo came back,
WSOK lit up the rabbit track. 🐇

🚥 Pre-merge checks | ✅ 5

✅ Passed checks (5 passed)

Check name	Status	Explanation
Title check	✅ Passed	The title is specific and accurately summarizes the regression test added for nested object literal and WebSocket handle corruption.
Description check	✅ Passed	The description follows the template well, covering summary, changes, related issue, test plan, output, and checklist.
Docstring Coverage	✅ Passed	Docstring coverage is 100.00% which is sufficient. The required threshold is 80.00%.
Linked Issues check	✅ Passed	Check skipped because no linked issues were found for this pull request.
Out of Scope Changes check	✅ Passed	Check skipped because no linked issues were found for this pull request.

✨ Finishing Touches

🧪 Generate unit tests (beta)

• Create PR with unit tests

---

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

[coderabbitai]

Actionable comments posted: 2

🤖 Prompt for all review comments with AI agents

Verify each finding against current code. Fix only still-valid issues, skip the
rest with a brief reason, keep changes minimal, and validate.

Inline comments:
In `@crates/perry/tests/nested_object_literal_ws_inbound.rs`:
- Around line 123-127: The success path in the inbound WebSocket test relies on
`ws.close()` and `server.close()` to naturally drain the event loop, which can
stall after `WSOK` is logged. Update the cleanup in the `message` handler to
make completion deterministic by explicitly finishing the test once the expected
`"echo:ping"` message is received, using the existing `timer`, `ws`, and
`server` cleanup path without depending on asynchronous close completion.
- Around line 55-57: The test fixture in nested_object_literal_ws_inbound.rs is
discarding stderr from the spawned Command, which hides the fixture’s own
failure message on both error and timeout paths. Update the child process setup
for Command::new(bin) so stderr is captured instead of sent to Stdio::null(),
and then read or surface that stderr in the test flow that uses child so
failures remain diagnosable.

🪄 Autofix (Beta)

Fix all unresolved CodeRabbit comments on this PR:

• Push a commit to this branch (recommended)
• Create a new PR with the fixes

---

ℹ️ Review info

⚙️ Run configuration

Configuration used: defaults

Review profile: CHILL

Plan: Pro Plus

Run ID: 502030e8-e8f7-4ad2-bfce-6401d8532914

📥 Commits

Reviewing files that changed from the base of the PR and between 240f131 and de5f5f9.

📒 Files selected for processing (1)

• crates/perry/tests/nested_object_literal_ws_inbound.rs

[machineloop]

Addressed the two review comments:

• Capture stderr — run_with_timeout now pipes the fixture's stderr and includes it in both the non-zero-exit assertion and the timeout panic, so the fixture's own failure signal (console.error("no echo …")) is visible on CI failures.
• Deterministic exit — the fixture now process.exit(0)s once it receives echo:ping instead of relying on ws.close()/server.close() to drain the loop (confirmed console.log flushes before exit).

Also ran cargo fmt to clear the lint failure (the multi-line assert!/assert_eq! weren't rustfmt-clean). Re-verified the fixture round-trips 3/3 against a clean main build (WSOK, exit 0).

The security-audit failure looks unrelated — it's a cargo audit advisory in the dependency tree, and this PR adds only a test file (no dependency changes), so it should reproduce on main.