Conversation
…chronous part ObjectInitializer deduplicated InitializeAsync with Lazy<Task> in ExecutionAndPublication mode, which runs the factory - InitializeAsync itself - under a lock. Every other caller for the same shared object blocked a thread-pool thread in Monitor.Enter until the synchronous part of InitializeAsync finished, starving the pool when that part did sync-over-async (e.g. Testcontainers' Docker probe) (thomhurst#6904). Publish a TaskCompletionSource (RunContinuationsAsynchronously) before any user code runs instead. The caller that publishes it runs InitializeAsync inline, as before, with no lock held and awaits it directly, so it takes no extra thread-pool hop; other callers await the published task and resume on their own pool threads. InitializeAsync still runs once per object, failures stay cached with the original exception object (thomhurst#4715), and a caller's cancellation only stops that caller waiting. Same pattern as ObjectLifecycleService.EnsureInitializedAsync. Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com> Claude-Session: https://claude.ai/code/session_01SZ1LwaBqCw5AJXeAU1vVo4
…er tests - Cancelling the caller that runs InitializeAsync only stops that caller: waiters still get the result, or the original exception object. - Assert the precondition of the inline-continuation test (InitializeAsync itself completed on the completing thread) and cover both a cancellable and a non-cancellable wait. - Bound every await with a timeout, release the blocking prefix in a finally, and cover a synchronously thrown OperationCanceledException. - Publish the result after the try/catch and make the comments precise. Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com> Claude-Session: https://claude.ai/code/session_01SZ1LwaBqCw5AJXeAU1vVo4
The previous version awaited a helper async method that ran InitializeAsync and published its outcome. BenchmarkDotNet showed that extra async frame cost ~0.15-0.7 us and ~86 B per first initialization of a per-test fixture whose InitializeAsync really awaits. The caller that publishes the task now awaits InitializeAsync itself and completes the TaskCompletionSource afterwards; only if it stops waiting (cancellation) or the initialization fails does a helper publish the initialization's own outcome, so one caller's cancellation is never cached as the result. First initialization now costs the same as the Lazy<Task> code and allocates less; repeat calls are unchanged. Also: a null task from InitializeAsync fails every caller with one InvalidOperationException naming the type; the published failure copy is marked observed; publishing doesn't wait for the initializing caller's context. Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com> Claude-Session: https://claude.ai/code/session_01SZ1LwaBqCw5AJXeAU1vVo4
If cancellation regressed, these awaits would hang the suite instead of failing, because the fixture is only released after them. Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com> Claude-Session: https://claude.ai/code/session_01SZ1LwaBqCw5AJXeAU1vVo4
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com> Claude-Session: https://claude.ai/code/session_01SZ1LwaBqCw5AJXeAU1vVo4
Following the review on thomhurst#6906: a helper runs InitializeAsync and completes a plain TaskCompletionSource; every caller, including the one that started it, waits the same way on the published task. - Drop RunContinuationsAsynchronously (a separate change); without it the initializing caller pays no extra thread-pool hop, so it no longer needs its own completion path, StartInitializer or PublishOutcomeAsync. - Drop the custom InvalidOperationException for a null task and the observed-exception read. - Keep InitializeCoreAsync async, so an OperationCanceledException thrown by InitializeAsync still completes callers' tasks as Canceled, as before. - Trim the tests to the five cases from the review; the failure case covers synchronous/asynchronous and ordinary/OperationCanceledException failures. Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com> Claude-Session: https://claude.ai/code/session_01SZ1LwaBqCw5AJXeAU1vVo4
Sing303
had a problem deploying
to
Pull Requests
September 28, 2026 15:05 — with 
GitHub Actions
Failure
GitHub Actions
Failure
This branch was successfully deployed
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.
This fork PR was the staging copy of thomhurst#6906, which is where review happens now. The branch is the same, so both PRs show the same commits. See thomhurst#6906 for the current description, measurements and discussion.
Current state (
92c54e2):ObjectInitializerfollows the maintainer's suggested shape. One helper runsInitializeAsyncand completes a plainTaskCompletionSource<bool>, and every caller waits on the published task.InitializeCoreAsyncstaysasync, so anOperationCanceledExceptionfrom the initializer still completes callers' tasks as Canceled, as onmain. The tests are trimmed to the five review cases (9 test cases).🤖 Generated with Claude Code
https://claude.ai/code/session_01SZ1LwaBqCw5AJXeAU1vVo4