Conversation
✅ Deploy Preview for golemcloud canceled.
|
- Pin wasmtime, wasmtime-wasi and wasmtime-wasi-http to =46.0.1 (component-model-async-bytes feature renamed to component-model-bytes) - Update all WIT trees from wasi 0.3.0-rc-2026-03-15 to the final 0.3.0 interfaces (root wit/, golem-common, golem-cli, and the SDK WIT copies) - Adapt to the v46 host APIs: trait-generic *WithStore<U> impls, BytesMut stream buffers, ComponentExtern.ty in agent extraction - golem-rust SDK: wasip3 0.7.0 and 0.3.0 WIT remaps - http-tests test component: wasi-fetch 0.2.0 (targets final wasi:http@0.3.0 and wit-bindgen 0.58, fixing the never-polled request-body upload task of wasi-fetch 0.1.3 / wit-bindgen#1305) Amp-Thread-ID: https://ampcode.com/threads/T-019f22a8-4db2-75fb-8967-5d612dccffe0 Co-authored-by: Amp <amp@ampcode.com>
Wasmtime runs sync-lowered host calls and async resource limiters through block_on on the store's fiber; while such a call awaits, the store event loop cannot poll any store-polled host future (wasmtime#11869, wasmtime#11870). Any async lock that such a host future holds across an await - or is merely queued on, since tokio's fair mutex hands ownership to woken-but-unpolled waiters - deadlocks the whole store as soon as fiber-side host code waits on the same lock. This intermittently hung http_client_using_reqwest_async_parallel. Shared-state paths reachable from both contexts no longer use await-crossing locks: - PrimaryOplog, ForwardingOplog and EphemeralOplog are actors now: the state is exclusively owned by an independent task processing a FIFO job channel, and callers only await oneshot replies - DurableWorkerCtx::increase_memory no longer awaits in the memory.grow limiter; the GrowMemory hint and global memory admission run on a detached task, and on admission failure the worker is restarted, reacquiring its memory through the startup admission path - commit_oplog_and_update_state sends the status-change notification from a detached task instead of queueing on the instance lock - the blob and compressed oplog archives no longer hold their internal locks across storage IO - the replay seen-log set moved out of the replay cursor lock into a synchronous mutex, so stdio host calls never queue behind cursor-advance transactions - LazyPromiseService clones the implementation Arc out instead of holding its RwLock across the delegated call Amp-Thread-ID: https://ampcode.com/threads/T-019f22a8-4db2-75fb-8967-5d612dccffe0 Co-authored-by: Amp <amp@ampcode.com>
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
1 participant
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.
Long-running migration branch, do not merge yet