Fix EventLoop async API: signal every concurrent WaitForEventAsync wa…#2427
Merged
Merged
Conversation
…iter Repro: any caller pattern that issues two concurrent WaitForEventAsync calls against a single EventLoop drops the wake signal to the second caller. Root cause: the loop tracked the wake TCS in a single volatile field (_eventAvailable). When a second caller arrived while the first was still pending, the new TCS was rejected by CompareExchange and the caller fell into the "Active waiter exists — events may already be pending" branch, returning Task.CompletedTask without registering any wake signal. The caller's outer loop would then re-enter WaitForEventAsync on the next IsEmpty=true check and immediately get Task.CompletedTask again — a CPU-burning spin until the queue happened to drain by some other path. Fix: hold a list of TCS waiters under a Lock, append per call, and broadcast TrySetResult to every entry on Enqueue. The list is cleared on each broadcast so subsequent Enqueues don't re-signal stale entries; spurious wakes (e.g. self-signaling on the post-registration double-check) are harmless because the caller's outer loop re-checks its own state. Added Jint.Tests/Runtime/AsyncTests.cs::EventLoopShouldSignalAllConcurrentWaiters which fails on the previous code (waiter1 is signaled, waiter2 stays pending after Enqueue) and passes with the fix. The single-waiter cancellation path (CancellationTokenRegistration disposal via tcs.Task.ContinueWith) is preserved unchanged for the common case where only one waiter is registered.
lahma
force-pushed
the
fix/eventloop-multi-waiter
branch
from
May 14, 2026 08:14
9ed6540 to
0962d38
Compare
lahma
enabled auto-merge (squash)
May 14, 2026 08:16
lahma
added a commit
to lahma/jint
that referenced
this pull request
May 14, 2026
Bumps the test262 suite to commit 673e9bac, which brings stage-4 spec updates for the immutable-arraybuffer proposal and a stricter ArrayBuffer.prototype.sliceToImmutable algorithm. Twelve test cases now pass with the following fixes: - ArrayBufferPrototype.SliceToImmutable: rewrite to the new spec algorithm. first/final are still computed from the original len, but the detach check now precedes the bounds check and a RangeError is thrown when the buffer has shrunk below final during coercion. - TypeConverter.ToNumber(string): guard against empty input after TrimEx. The pre-trim IsNullOrWhiteSpace check misses BOM and other characters that the spec-compliant TrimEx removes, which previously caused an IndexOutOfRangeException on the whitespace-padded empty string used by the new argument-coercion.js test. - Uint8Array.prototype.setFromBase64 / setFromHex: assert the viewed buffer is not immutable before writing, per the new spec step. - %TypedArray%.from / .of: assert the buffer of a custom constructor's result is not immutable before populating it (the "write" mutability intent in TypedArrayCreateFromConstructor). - TypedArrayConstructor.IterableToList: pass the already-fetched iterator method through to GetIterator instead of re-reading Symbol.iterator on the source. Also fixes a CI-host crash on PR sebastienros#2427: a FinalizationRegistry cleanup callback can throw arbitrary exceptions (e.g. TimeoutException when engine constraints trip), and they were escaping the GC finalizer thread and aborting the entire test process. The callback invocation is now wrapped in a swallow-all try/catch, matching the spec's intent that these callbacks run in an isolated Job. Co-Authored-By: Claude Opus 4.7 (1M context) <noreply@anthropic.com>
lahma
added a commit
that referenced
this pull request
May 14, 2026
Bumps the test262 suite to commit 673e9bac, which brings stage-4 spec updates for the immutable-arraybuffer proposal and a stricter ArrayBuffer.prototype.sliceToImmutable algorithm. Twelve test cases now pass with the following fixes: - ArrayBufferPrototype.SliceToImmutable: rewrite to the new spec algorithm. first/final are still computed from the original len, but the detach check now precedes the bounds check and a RangeError is thrown when the buffer has shrunk below final during coercion. - TypeConverter.ToNumber(string): guard against empty input after TrimEx. The pre-trim IsNullOrWhiteSpace check misses BOM and other characters that the spec-compliant TrimEx removes, which previously caused an IndexOutOfRangeException on the whitespace-padded empty string used by the new argument-coercion.js test. - Uint8Array.prototype.setFromBase64 / setFromHex: assert the viewed buffer is not immutable before writing, per the new spec step. - %TypedArray%.from / .of: assert the buffer of a custom constructor's result is not immutable before populating it (the "write" mutability intent in TypedArrayCreateFromConstructor). - TypedArrayConstructor.IterableToList: pass the already-fetched iterator method through to GetIterator instead of re-reading Symbol.iterator on the source. Also fixes a CI-host crash on PR #2427: a FinalizationRegistry cleanup callback can throw arbitrary exceptions (e.g. TimeoutException when engine constraints trip), and they were escaping the GC finalizer thread and aborting the entire test process. The callback invocation is now wrapped in a swallow-all try/catch, matching the spec's intent that these callbacks run in an isolated Job. Co-authored-by: Claude Opus 4.7 (1M context) <noreply@anthropic.com>
This was referenced May 18, 2026
This was referenced Jun 4, 2026
This was referenced Jun 29, 2026
This was referenced Jul 7, 2026
This file contains hidden or bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
2 participants
Add this suggestion to a batch that can be applied as a single commit.This suggestion is invalid because no changes were made to the code.Suggestions cannot be applied while the pull request is closed.Suggestions cannot be applied while viewing a subset of changes.Only one suggestion per line can be applied in a batch.Add this suggestion to a batch that can be applied as a single commit.Applying suggestions on deleted lines is not supported.You must change the existing code in this line in order to create a valid suggestion.Outdated suggestions cannot be applied.This suggestion has been applied or marked resolved.Suggestions cannot be applied from pending reviews.Suggestions cannot be applied on multi-line comments.Suggestions cannot be applied while the pull request is queued to merge.Suggestion cannot be applied right now. Please check back later.
…iter
Repro: any caller pattern that issues two concurrent WaitForEventAsync calls against a single EventLoop drops the wake signal to the second caller.
Root cause: the loop tracked the wake TCS in a single volatile field (_eventAvailable). When a second caller arrived while the first was still pending, the new TCS was rejected by CompareExchange and the caller fell into the "Active waiter exists — events may already be pending" branch, returning Task.CompletedTask without registering any wake signal. The caller's outer loop would then re-enter WaitForEventAsync on the next IsEmpty=true check and immediately get Task.CompletedTask again — a CPU-burning spin until the queue happened to drain by some other path.
Fix: hold a list of TCS waiters under a Lock, append per call, and broadcast TrySetResult to every entry on Enqueue. The list is cleared on each broadcast so subsequent Enqueues don't re-signal stale entries; spurious wakes (e.g. self-signaling on the post-registration double-check) are harmless because the caller's outer loop re-checks its own state.
Added Jint.Tests/Runtime/AsyncTests.cs::EventLoopShouldSignalAllConcurrentWaiters which fails on the previous code (waiter1 is signaled, waiter2 stays pending after Enqueue) and passes with the fix.
The single-waiter cancellation path (CancellationTokenRegistration disposal via tcs.Task.ContinueWith) is preserved unchanged for the common case where only one waiter is registered.