Repository navigation
Conversation
An empty activation still ran one drain, so the marker has to record it or replay fires wait_condition predicates a different number of times. The marker also goes in with the first set that drains, rather than after the signal and update jobs have already published.
Behind a patch-only job set the install added a second drain, so every wait_condition predicate fired once more than the recorded task fired it. The dispatch moved to its own method so the schedule can be asserted.
1 of 3 tasks
Owner
Author
|
Replaced by #103 in the v5 series (phase 1, Signal wake only). The branch stays as a pin. |
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Outdated suggestions cannot be applied.This suggestion has been applied or marked resolved.Suggestions cannot be applied from pending reviews.Suggestions cannot be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.
This PR makes external stream replay take the same drains the live run took.
What changed?
An activation with no stream activity still records its segment once a subscription exists, so input and output share one replay schedule. A prerelease marker that left those segments out is rejected with a
NondeterminismErrorrather than guessed at. A replay marker is applied right before the first drain that can publish, and it no longer earns a drain of its own. One handoff case reads the marker right before the step it checks, because every closing task now writes its own marker. The changelog gets a Fixed line.Part of AI-198 (epic AI-37).
Why?
Replays of runs that read input, publish and wait on an activity diverged. Without the empty segments,
wait_conditionpredicates ran a different number of times on replay. A marker installed after the Signal jobs had already published wiped those records and failed the manifest check, and its extra drain put the activation one drain ahead of the task the marker came from. These are two fixes, but they share the activation path andtest_output_runtime.py, so they ship as one PR with a commit each.How did you test it?
Link to a test plan if any -
The lint set is clean. The new schedule and drain cases, the whole external stream suite and
tests/streamspass on the dev server the fixtures start and with-Eagainst a local channel server.