test: avoid TLA self-cycle in bun-main dynamic-import test

[dylan-conway]

Follow-up to #29719.

bun:main statically imports the entry file, so await import("bun:main") at the entry's top level is a TLA self-cycle: bun:main waits for entry.mjs (async dep), and entry.mjs waits for bun:main's evaluation promise. Per spec that promise never settles.

The old JSC module loader broke these cycles early, which is why #29719 passed locally (tested against 892042c2, pre-#29393). After #29393 (WebKit module-loader rewrite) the loader correctly leaves the promise unsettled, so the test now hangs and times out at 90s — see build 48023 (debian-asan, win x64, win x64-baseline).

Fix: drop the top-level await and use import("bun:main").then(...) so entry.mjs finishes synchronously, bun:main finishes, and the import resolves on the next microtask. The preload and --hot tests are unaffected (preload isn't in bun:main's dep graph).

Note: this also surfaced that Bun now hangs forever on any unsettled-TLA cycle instead of exiting 13 like Node — separate PR coming for that.

How did you verify your code works?

• bun bd test test/js/bun/resolve/bun-main-entry-point.test.ts — 3 pass, 20 consecutive runs on Windows
• USE_SYSTEM_BUN=1 bun test ... — 2 fail (still catches the alias bug on unfixed bun)

[robobun]

^{Updated 11:20 PM PT - Apr 25th, 2026}

❌ @dylan-conway, your commit ca81de4 has 3 failures in Build #48048 (All Failures):

• scripts/build/ci.ts - build failed on 🍎 aarch64 - build-cpp
• test/cli/inspect/test-reporter.test.ts - code 1 on 🐧 13 x64
• test/js/bun/util/v8-heap-snapshot.test.ts - SIGKILL on 🐧 25.04 x64-baseline

---

🧪   To try this PR locally:

bunx bun-pr 29738

That installs a local version of the PR into your bun-29738 executable, so you can run:

bun-29738 --bun

[claude]

LGTM — test-only deflake, well-reasoned fix for the TLA self-cycle.

Extended reasoning...

Overview

This PR modifies a single test fixture in test/js/bun/resolve/bun-main-entry-point.test.ts. It rewrites the embedded entry.mjs script in the "dynamic import('bun:main')" test case to use import("bun:main").then(...) instead of top-level await import("bun:main"), avoiding a TLA self-cycle deadlock that surfaced after the WebKit module-loader rewrite (#29393). The assertions are unchanged; an explicit rejection handler was added that prints the error and exits 1.

Security risks

None. This is a test-only change to an embedded fixture string. No production/runtime code, no auth/crypto/permissions, no new dependencies.

Level of scrutiny

Low. The change is small (~15 lines net), mechanical, and confined to test fixture code. The reasoning is clearly documented in both the PR description and an inline comment, and the author verified 20 consecutive passes on Windows plus confirmed the test still catches the original alias regression under USE_SYSTEM_BUN=1.

Other factors

• The assertion semantics are preserved: the same checks run inside the .then() fulfillment handler, and the outer test still asserts stdout: "OK\n", empty stderr, and exitCode: 0. If an assertion throws inside the handler it becomes an unhandled rejection, which will produce non-empty stderr / non-zero exit and correctly fail the outer expect.
• The other two tests in the file (preload and --hot) are untouched, matching the PR description's note that they aren't in bun:main's dep graph.
• No prior reviewer comments or outstanding feedback on the timeline; bug-hunting system found no issues.

[coderabbitai]

No actionable comments were generated in the recent review. 🎉

ℹ️ Recent review info

⚙️ Run configuration

Configuration used: Path: .coderabbit.yaml

Review profile: ASSERTIVE

Plan: Pro

Run ID: 21caa34f-8be7-493c-bb14-872399f29a98

📥 Commits

Reviewing files that changed from the base of the PR and between 34d86be and ca81de4.

📒 Files selected for processing (1)

• test/js/bun/resolve/bun-main-entry-point.test.ts

---

Walkthrough

Replaced a top-level await import("bun:main") in the entry.mjs test fixture with a promise-based import("bun:main").then(...) wrapper to avoid a TLA self-cycle deadlock. Moved module-namespace and empty-exports assertions into the then callback and added .catch(...) to log errors and exit with code 1.

Changes

Cohort / File(s)	Summary
Test fixture refactor test/js/bun/resolve/bun-main-entry-point.test.ts	Updated the entry.mjs test fixture: replaced top-level await import("bun:main") with import("bun:main").then(...); moved Symbol.toStringTag and empty-export assertions into the then callback; added .catch(...) to print errors and call process.exit(1).

🚥 Pre-merge checks | ✅ 4

✅ Passed checks (4 passed)

Check name	Status	Explanation
Title check	✅ Passed	The title 'test: avoid TLA self-cycle in bun-main dynamic-import test' directly describes the main change: fixing a test hang caused by a top-level-await self-cycle.
Description check	✅ Passed	The description provides detailed context about the TLA self-cycle issue, references related PRs, explains the fix, and includes verification steps. Both required template sections are addressed.
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.

_{✏️ Tip: You can configure your own custom pre-merge checks in the settings.}

---

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

[coderabbitai]

Actionable comments posted: 1

🤖 Prompt for all review comments with AI agents

Verify each finding against the current code and only fix it if needed.

Inline comments:
In `@test/js/bun/resolve/bun-main-entry-point.test.ts`:
- Around line 29-42: The import("bun:main").then(...) call currently passes a
rejection handler as the second .then argument but assertion errors thrown
inside the fulfillment callback (the function that inspects m using
Symbol.toStringTag and Object.keys) will bypass that and won't trigger the
process.exit(1) path; fix by chaining a .catch() after the .then to log the
error with console.error(String(e)) and call process.exit(1) so any thrown
errors from the fulfillment handler are handled and cause the test to fail.

🪄 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: Path: .coderabbit.yaml

Review profile: ASSERTIVE

Plan: Pro

Run ID: 3097c256-f9ef-4f30-9575-3b097705514e

📥 Commits

Reviewing files that changed from the base of the PR and between 30478d2 and 34d86be.

📒 Files selected for processing (1)

• test/js/bun/resolve/bun-main-entry-point.test.ts

[claude]

LGTM — straightforward test-fixture deflake, and the .catch() feedback is already addressed.

Extended reasoning...

Overview

This PR touches a single test file, test/js/bun/resolve/bun-main-entry-point.test.ts, and only modifies the inline entry.mjs fixture string for one test case. It replaces const m = await import("bun:main") at the entry's top level with import("bun:main").then(...).catch(...), plus an explanatory comment. The assertion logic inside is byte-for-byte identical to before; only the control-flow wrapper changed. The other two tests in the file (preload and --hot) are untouched.

Security risks

None. This is a test-only change to a fixture string that runs in a spawned subprocess inside a temp dir. No runtime, resolver, or production code is modified.

Level of scrutiny

Low. The PR description gives a precise and correct explanation: bun:main statically imports the entry, so top-level await import("bun:main") from the entry creates a TLA self-cycle that per spec never settles, and the post-#29393 module loader now (correctly) honors that. Dropping the top-level await lets entry.mjs finish synchronously so bun:main can settle and the dynamic import resolves on the next microtask. The .then().catch() shape ensures both import rejection and assertion throws inside the fulfillment handler hit process.exit(1), so regressions still fail the outer exitCode: 0 / stdout: "OK\n" assertion. The change is mechanical and self-contained.

Other factors

• No bugs flagged by the bug-hunting system.
• CodeRabbit's one nit (use chained .catch() instead of the second .then arg so assertion throws are caught) was addressed in ca81de4 and the final diff reflects it; that thread is resolved.
• Author verified 3/3 pass over 20 consecutive Windows runs, and confirmed the test still fails on an unfixed bun (so it hasn't been neutered).
• The robobun ❌ on build #48033 is the generic CI status comment; given this PR only edits one test fixture string, any unrelated failures there are not attributable to this change.