fix(hir): continue in for-await drivers skipped the iterator advance (spin)

[proggeramlug]

Problem

Every iterator-driver desugar that modeled for await (x of src) as

let __result = await __iter.next();
while (!__result.done) {
    <bind x>; <user body>;
    __result = await __iter.next();   // advance at the body TAIL
}

broke continue: it jumps to the while condition, skips the tail advance, and re-processes the same result forever — an infinite spin observed as a hang. Six copies of the shape existed (lower/stmt_loops.rs ×3: generic for-await, recognized-generator-call, Web-ReadableStream reader; lower_decl/body_stmt.rs ×2; lower_decl/body_stmt/for_await.rs ×1).

The array/lazy sync path (lazy_iter_for_stmt) already documents this exact footgun — "a while with the advance at the body tail would skip it on continue and spin" — and dodges it with Stmt::For's update clause, which the await-capable drivers can't use (no await in a for update).

Ten-line repro (hangs pre-fix, node prints both values + done):

async function* events() { yield "a"; yield "ping"; yield "b"; }
async function* filtered() {
  for await (const w of events()) {
    if (w === "ping") continue;   // ← spins forever on the same result
    yield w;
  }
}
for await (const x of filtered()) console.log(x);

Canonical real-world failure: an SSE consumer's for await (const ev of stream) { if (ev.event === "ping") continue; … } — a large esbuild-bundled CLI app hung on the first real server ping event. Local mocks that never sent pings masked it completely; ping is the only SSE event the client skips via continue.

Fix

All six drivers now advance at the TOP:

let __result = undefined;
while (true) {
    __result = await __iter.next();
    if (__result.done) break;
    <bind x>; <user body>;
}

continue falls to the while (true) condition and re-runs the advance. break/return/throw behavior is unchanged — the synthetic if done break is appended after the abrupt-close rewrite over the user body, so normal completion never runs a spurious IteratorClose. The three stmt_loops.rs sites share a new iter_driver_while_stmt helper; the body_stmt copies mirror it.

Tests

New crates/perry/tests/for_await_continue.rs (5 tests, node-oracle-verified, 30s hang guard since pre-fix they spin forever):

• continue inside for-await in an async generator (the SSE shape)
• in a plain async fn
• over a custom [Symbol.asyncIterator] iterable (runtime GetAsyncIterator driver)
• over a Web ReadableStream (getReader()/read() driver)
• break + nested-loop-continue behavior guard

(The pre-existing c262_parity::logical_property_assignment_short_circuits_the_store_4586 failure reproduces identically on pristine main — unrelated.)

Summary by CodeRabbit

• Bug Fixes

  • Improved for await and iterator-based loop handling so continue now advances to the next item correctly instead of reusing the same completed result.
  • Fixed async loops over generators, custom async iterators, and readable streams to avoid hangs and infinite spinning.
  • Preserved correct behavior for break and nested-loop continue cases.

• Tests

  • Added regression coverage for for await loops to verify correct output and prevent future loop-control regressions.

[coderabbitai]

📝 Walkthrough

Walkthrough

This PR fixes an infinite-spin bug in for await / for...of loop lowering by restructuring iterator-driven loops (async generators, custom async iterators, and ReadableStream) to advance the iterator (next()/read()) at the top of a while (true) loop rather than at the tail, ensuring continue correctly re-drives iteration. Adds regression tests.

Changes

Advance-at-top Iterator Driver Fix

Layer / File(s)	Summary
New iter_driver_while_stmt helper and call sites crates/perry-hir/src/lower/stmt_loops.rs	Adds iter_driver_while_stmt that sets result from next_call at loop top, breaks on done, then runs body; applies it in for-await iterator, for-of iterator-protocol, and ReadableStream lowering, changing result-local init to undefined.
body_stmt.rs advance-at-top rewrite crates/perry-hir/src/lower_decl/body_stmt.rs	Rewrites ReadableStream and generic iterator-protocol loops in lower_body_stmt to while (true) with result assignment and done/break check inside the body instead of the loop condition.
for_await.rs synthesized loop update crates/perry-hir/src/lower_decl/body_stmt/for_await.rs	Changes __result init to undefined and restructures the synthesized driver loop to advance next() first, break on done, then bind/execute user body.
Regression tests crates/perry/tests/for_await_continue.rs	Adds compile-and-run test helper with a 30s hang guard, and tests for continue/break across async generators, custom async iterators, and ReadableStream for-await loops, including nested-loop continue semantics.

Estimated code review effort: 4 (Complex) | ~50 minutes

Possibly related PRs

• PerryTS/perry#5810: Both PRs change how for await/async-iterator consumption drives next() and handles done completion.

Suggested labels: run-extended-tests

🚥 Pre-merge checks | ✅ 5

✅ Passed checks (5 passed)

Check name	Status	Explanation
Title check	✅ Passed	The title clearly matches the main fix: preventing skipped advances and spins in for-await drivers.
Description check	✅ Passed	The description covers the problem, fix, and tests, but it does not follow the template headings and omits Related issue and Checklist sections.
Docstring Coverage	✅ Passed	No functions found in the changed files to evaluate docstring coverage. Skipping docstring coverage check.
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

---

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.}

[coderabbitai]

🧹 Nitpick comments (1)

crates/perry-hir/src/lower/stmt_loops.rs (1)

361-393: 📐 Maintainability & Code Quality | 🔵 Trivial | ⚡ Quick win

Consider exposing iter_driver_while_stmt (or a shared equivalent) so the other two files reuse it.

The advance-at-top shape here is re-inlined verbatim in lower_decl/body_stmt.rs (ReadableStream + iterator-protocol) and lower_decl/body_stmt/for_await.rs. Those copies even reference this helper in their comments. Since it was exactly this shape diverging across the "six copies" that produced the hang, promoting this to a pub(crate) helper and calling it from all sites would prevent future drift. The implementation itself is correct.

🤖 Prompt for 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.

In `@crates/perry-hir/src/lower/stmt_loops.rs` around lines 361 - 393, The
advance-at-top iterator-driver loop is duplicated in other lowering paths, which
risks the copies drifting again; expose iter_driver_while_stmt as a shared
pub(crate) helper (or equivalent common helper) and update
lower_decl/body_stmt.rs and lower_decl/body_stmt/for_await.rs to call it instead
of inlining the shape. Keep the implementation in stmt_loops.rs as the single
source of truth so the iterator-driver behavior stays consistent across all
sites.

🤖 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.

Nitpick comments:
In `@crates/perry-hir/src/lower/stmt_loops.rs`:
- Around line 361-393: The advance-at-top iterator-driver loop is duplicated in
other lowering paths, which risks the copies drifting again; expose
iter_driver_while_stmt as a shared pub(crate) helper (or equivalent common
helper) and update lower_decl/body_stmt.rs and lower_decl/body_stmt/for_await.rs
to call it instead of inlining the shape. Keep the implementation in
stmt_loops.rs as the single source of truth so the iterator-driver behavior
stays consistent across all sites.

---

ℹ️ Review info

⚙️ Run configuration

Configuration used: defaults

Review profile: CHILL

Plan: Pro Plus

Run ID: bd80e267-7432-4997-aad7-fc91be287be5

📥 Commits

Reviewing files that changed from the base of the PR and between 95059b8 and 4bf4e9f.

📒 Files selected for processing (4)

• crates/perry-hir/src/lower/stmt_loops.rs
• crates/perry-hir/src/lower_decl/body_stmt.rs
• crates/perry-hir/src/lower_decl/body_stmt/for_await.rs
• crates/perry/tests/for_await_continue.rs