The GitHub Actions job "Binary Compatibility" on 
pekko.git/optimize-flatten-merge-avoid-materialization has failed.
Run started by GitHub user He-Pin (triggered by He-Pin).

Head commit for run:
5f731d5d6e2a78e86acad536cd5d368f2220d7fe / He-Pin <[email protected]>
fix: address PR #2977 review feedback for value-presented sources

Motivation:
Copilot review on PR #2977 flagged two issues: (1) Java streams obtained
via Source.fromJavaStream were converted to Scala iterators and fed into
InflightIteratorSource, dropping the BaseStream close contract and leaking
onClose handlers and underlying resources; (2) the test purporting to
verify "no pre-materialization for value-presented sources" actually
counted lazySingle materializations and was misleading because Source.lazySingle
is itself non-VP.

Modification:
- Add InflightJavaStreamSource in InflightSources.scala that wraps the
  BaseStream directly, eagerly closes empty streams, closes on exhaustion,
  and closes on cancel.
- Wire the new wrapper through FlattenConcat.addJavaStreamSource and
  FlattenMerge.addInflightJavaStreamSource so both stages honor the
  close contract on the value-presented fast path.
- Cancel queued inflight sources in FlattenMerge.postStop so JavaStream
  resources held in the queue (not yet promoted to active SubSinkInlets)
  are released on stage termination.
- Replace the misleading test with one that mixes value-presented and
  non-VP inner sources via lazySingle().buffer() and asserts the counter
  equals only the non-VP count, proving the VP fast path skips
  materialization.
- Add two regression tests for the close contract: exhaustion of
  finite Java streams and downstream cancel against infinite ones.

Result:
Java streams routed through the VP fast path now close deterministically
on exhaustion, cancel, and stage termination. The materialization-skip
property is demonstrated by a meaningful counter test rather than a
tautology. All 39 FlowFlattenMergeSpec tests pass.

Report URL: https://github.com/apache/pekko/actions/runs/26012711694

With regards,
GitHub Actions via GitBox


---------------------------------------------------------------------
To unsubscribe, e-mail: [email protected]
For additional commands, e-mail: [email protected]

Reply via email to