Conversation
* Prefactor activity execution traits * Prefactor ephemeral server APIs * Add SDK test environments * Address SDK testing review feedback * self review * self review: better heartbeating * fix merge * fixup merge conflict * pr feedback * rename * add back error msg part * pr feedback * support saa
* chore(sdk): release 0.7.0 * Apply suggestions from code review Co-authored-by: Chris Olszewski <chrisdolszewski@gmail.com>
* feat(client): allow setting memo on workflow start `WorkflowStartOptions` had no memo field, so there was no way to attach a memo when starting a workflow from the client. The read side and the workflow side already exist - continue-as-new can set one, and `WorkflowExecutionDescription::memo()` reads one back - it was only the client start path that was missing. Adds `memo: Option<ProtoMemo>` and wires it into both `StartWorkflowExecutionRequest` and `SignalWithStartWorkflowExecutionRequest`. Uses the proto type directly, matching the existing `header: Option<Header>` field. Both are `map<string, Payload>` and the caller has to convert values either way. Imported as `ProtoMemo` so it doesn't collide with the re-exported `temporalio_common::Memo`, which is the read-side wrapper. Two integ tests in workflow_client_tests.rs: one that a memo set at start comes back through describe, one that it's empty when unset. * feat(client): use MemoValues for memo on workflow start Per review, the client shouldn't ask callers to build the proto Memo themselves. It now takes `MemoValues`, the same type the workflow surface already uses for continue-as-new and `upsert_memo`. `MemoValue`/`MemoValues` move out of `temporalio_workflow` and into common-wasm, next to the existing `Memo` read wrapper. That's where they have to go for both sides to reach them... `temporalio_workflow` depends on common-wasm, not common. Re-exported from `temporalio_common`, `temporalio_workflow` and `temporalio_sdk`, so existing imports still work. `MemoValue` holds its value in an `Arc` now instead of an `Rc`, and wants `Send + Sync`. `WorkflowStartOptions` travels through the interceptor chain inside a `Send` future and an `Rc` can't cross that. Same shape as `DataConverter`, which already erases its converters as `Arc<dyn ... + Send + Sync>`. Also runs the memo through the payload codec, not just the payload converter. The read side already codec-decodes it... describe and list both run `decode_payloads` over the memo. So converter-only encoding broke the round trip for anyone with a real codec. Caught this after wiring up the tests below, the two client tests fail without it. Client unit tests cover start and signal-with-start under a codec, plus the serialization failure path. Integ tests moved over to `MemoValues`. * Address review feedback on memo `MemoValue` implements `TemporalSerializable` instead of exposing a doc hidden `to_payload`, so callers serialize it through the normal converter path. It also picks up the caller's serialization context now rather than hardcoding `Workflow`. `MemoValues::encode` is gone in favour of documented `get`/`iter`. The three callers build the payload map themselves. That is a few duplicated lines each, but nothing doc hidden in the public API. Test cleanups from the review... dropped the `memo_with` helper, the `assert_ne!` on the raw bytes that the following assertion already covers, and the empty-memo integ test that the client unit test covers. * fixup changelog --------- Co-authored-by: Chris Olszewski <chrisdolszewski@gmail.com>
* fix(sdk): typed signal_with_start_workflow * fix: rebase memo fixes
…oralio#1511) * Stabilize legacy query activation order * Clarify legacy query ordering comment SDK-Sentinel-Request: temporalio#1511/comment-5398459129
* Move changelog release notes helper to its own crate * Select the changelog for release notes
* Allow including LA arguments in marker * Expose option in Rust SDK
…oralio#1483) * Add envconfig integration test mode * Add isolated Cloud namespace helper * Test envconfig harness in isolated Cloud namespace * Clarify envconfig snapshot comment * Move Cloud namespace commands into integration runner
…1525) Updates the requirements on [base64](https://github.com/marshallpierce/rust-base64) to permit the latest version. - [Changelog](https://github.com/marshallpierce/rust-base64/blob/master/RELEASE-NOTES.md) - [Commits](marshallpierce/rust-base64@v0.22.0...v0.23.1) --- updated-dependencies: - dependency-name: base64 dependency-version: 0.23.1 dependency-type: direct:production ... Signed-off-by: dependabot[bot] <support@github.com> Co-authored-by: dependabot[bot] <49699333+dependabot[bot]@users.noreply.github.com>
…io#1536) * chore(sdk): mark incoming error as non_exhaustive * chore(sdk/envconfig): add new SDK owned envconfig structs instead of reexporting from core * chore(sdk): mark schedule aspects with nonexhaustive * chore(sdk): remove legacy ActExtiValue * chore(sdk): move FailOnNondeterminismInterceptor to tests * chore(sdk): non_exhaust data converter types * chore(sdk): mark SerializationContext as non_exhaustive * chore(sdk): make TaskToken bytes private * chore(sdk): make ActivityCloseTimeouts non exhaustive * chore(sdk): mark workflow execution info with non_exhaustive * chore(sdk): WorkerCallbacks is nonexhaustive * chore(sdk): make Priority and WorkflowDeploymentVersion nonexhaustive * changelog fixups * chore(sdk): mark envconfig as non_exhaustive * fix lint * fix docs * .
Cancelled and deadline-exceeded polls already retain their slots during backoff. Keep their targets unchanged so transient failures do not leave polling concurrency reduced until a later server scale-up.
Picks up the per-runner `timeout-minutes` fix so `macos-intel` integ tests get the 40 minutes the matrix asks for.
The same fix already arrived here with the Core merge, so this only keeps the branch a descendant of the repairs branch.
The unified branch carries the same vendored WIT, so consumers that run the newer nexgen need the placeholder declarations here too.
The Python bridge builds against temporalio_common::WorkerDeploymentVersion::builder, which only exists in the newer upstream Core.
…o#1608) `temporalio-common`'s build script emits `payload_visitor_impl.rs` by iterating `PayloadModel::payload_containing`, a `HashSet<String>`, and orders each message's fields by iterating `oneof_fields`, a `HashMap<i32, _>`. Both use the default `RandomState`, which is seeded per process, so the generated file holds the same implementations in a different order on every build. The file is a compile input, so its content is part of what a compilation cache keys on. sccache therefore misses `temporalio-common` on every build, and — because each miss changes the crate's own output — misses every crate downstream of it too. In a workspace using the Rust SDK that is `temporalio-client`, `temporalio-sdk-core` and `temporalio-sdk`, plus the user's own crates that link them. Collect both into a `Vec` and sort before emitting, matching what the payload-limits generator in the same file already does for its roots, generated names and oneof indices. Validation: five consecutive build-script runs against the patched tree produce byte-identical `payload_visitor_impl.rs` and `payload_limits_impl.rs`; three runs against the unpatched tree produce three different files. The generated content is unchanged — sorting the unpatched output's lines yields the patched output.
* Update API definitions * Fix API definition update * Classify callback context as a blob * Order callback context limit classification
* SAA samples
Pages carry the same command list in order, only the intermediate ones are buffered, and the final page merges them, so the completion still reaches History as one commit.
# Conflicts: # crates/sdk-core/src/worker/workflow/managed_run.rs
# Conflicts: # crates/sdk-core/src/protosext/mod.rs
2 of 3 tasks
# Conflicts: # crates/sdk-core/src/protosext/mod.rs
# Conflicts: # crates/sdk-core/src/worker/workflow/machines/workflow_machines.rs
moedash
added a commit
that referenced
this pull request
Oct 3, 2026
Carries the vendored api change 4304fd8..071feb8: the five channel requests' execution fields and both linked_to fields reuse the old numbers and nothing is reserved. Core-3 fbc064b applies the same patch.
moedash
added a commit
that referenced
this pull request
Oct 3, 2026
Owner
Author
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
19 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 PR merges current upstream
maininto the external stream branch and fixes the call sites it broke.What changed?
managed_run.rsfollows upstream'sWftFailureKindreporting and its renamedauto_reply_failfield, and keeps the external stream state.add_cmd_to_wf_tasktakes its annotations by value, for the lang-issued channel commands and the ones Core issues on its own, and the arm that rejects a stray external stream command names the command in itsfatal!.WorkflowStreamProgressordering rule says why workflow command pagination can't move the marker. Pages carry the same command list in order, the server buffers only the intermediate pages, and the final page merges them into one commit.WorkflowStreamChannelsreport, so lang names the channels a run listens on and Core defers the subscribe to the completion that ends the Workflow Task.UnsubscribeNotificationChannelcommand and its machine, issued for a channel that left the reported set ahead of any subscribe on the same completion.ChannelSubscriptionInfowithDescribeWorkflowExecutionResponse.channel_subscriptions.This layer holds the external lineage plus upstream only. #4 adds the native stream commits.
Part of AI-198 (epic AI-37).
Why?
The two prototype Cores sat weeks apart on
main, so combining them meant reconciling everything at once. After this merge, stream code is the only difference left between them. The Python bridge also needsWorkerDeploymentVersion::builderfrom the newer base to compile.How did you test it?
The full cargo suite ran on the branch, and the sdk-core unit suite,
cargo fmt --checkandcargo test-lintran again on the head that carries the deferred subscribe and the unsubscribe. The bridge compile check ran from sdk-python against the descendant branch. No live lane runs against this head on its own, since the Python SDK pins #4.