Scope redriven signal waits and search-attribute types out for Python

CHANGELOG.md

@@ -16,7 +16,7 @@ and this project adheres to [Semantic Versioning](https://semver.org/spec/v2.0.0
 
 ### Added
 
-- Replay of Workflow 2.19 redrive reuse is qualified for `MemoUpserted` and for `SearchAttributesUpserted` value identity. Neither event is a replay clock source: `ctx.now()` stays the persisted start time, which on a redriven run is the failed run's start, and the cancellation budget does not move on an upsert. Search-attribute type identity is not qualified, because the Server records namespace-registered types the SDK cannot see. Signal waits (`SignalWaitOpened`, `SignalReceived`, `SignalApplied`) are not qualified for Python: the SDK has no signal-wait primitive. Replaying such reused events through a signal handler applies the signal once and takes no time from them, but that does not qualify signal waits. (#235)
+- Replay of Workflow 2.19 redrive reuse is qualified for `MemoUpserted` and for `SearchAttributesUpserted` value identity. Neither event is a replay clock source: `ctx.now()` stays the persisted start time, which on a redriven run is the failed run's start, and the cancellation budget does not move on an upsert. Search-attribute type identity is not qualified. The SDK's upsert declares no `attribute_types`, and the Server records namespace-registered types the SDK cannot see, so recorded types are not compared with authored ones. They are only checked to be well formed and able to hold the recorded value. Signal waits (`SignalWaitOpened`, `SignalReceived`, `SignalApplied`) are scoped out for Python, because the SDK cannot author a Core signal wait. Reused signal-wait events replay for history validity only: a cold replay applies the recorded signal once, takes no time from these events and opens no wait again. This does not qualify Python workers for Core signal waits. (#235)
 
 ## [2.13.1] - 2026-11-03
 

tests/test_replay_clock.py

@@ -11,7 +11,7 @@
 
 from durable_workflow import serializer, workflow
 from durable_workflow.errors import NonDeterministicReplayError, WorkflowCancelled
-from durable_workflow.workflow import CompleteWorkflow, WorkflowContext, query_state, replay
+from durable_workflow.workflow import CompleteWorkflow, WaitCondition, WorkflowContext, query_state, replay
 from tests.redrive_fixture_registry import VALID_SUCCESSORS
 
 
@@ -344,6 +344,28 @@
     assert serializer.decode(received["arguments"], codec="avro") == ["Grace"]
 
 
+def test_a_cold_worker_replays_reused_signal_wait_events_without_reopening_the_wait() -> None:
+    # #235 scope: the SDK cannot author a Core signal wait, so these reused
+    # events replay for history validity only. A cold replay applies the
+    # recorded signal once and opens no wait again. This does not qualify
+    # Python workers for Core signal waits: the counterpart's next call is
+    # numbered from 1, while Core's wait occupies step 1.
+    history = redriven("redrive-reused-signal-wait-clock")
+    for _ in range(2):
+        RedriveSignalWorkflow.applied = []
+        commands = replay_redriven(RedriveSignalWorkflow, history)
+        assert not any(isinstance(command, WaitCondition) for command in commands)
+        assert RedriveSignalWorkflow.applied == ["Grace"]
+    # Control: without the recorded signal the same replay opens the wait.
+    unsignalled = [deepcopy(event) for event in history
+                   if event["event_type"] not in ("SignalReceived", "SignalApplied")]
+    for number, event in enumerate(unsignalled, start=1):
+        event["sequence"] = number
+    RedriveSignalWorkflow.applied = []
+    assert [type(command) for command in replay_redriven(RedriveSignalWorkflow, unsignalled)] == [WaitCondition]
+    assert RedriveSignalWorkflow.applied == []
+
+
 def _retimed(history: list[dict[str, Any]], event_type: str, mode: str) -> list[dict[str, Any]]:
     changed = deepcopy(history)
     event = only_event(changed, event_type)