lsm / HyperNeo / 32330821170

80%

Build Type

push

github

Committed by web-flow

Commit Message

Superpipe pilot: functional-core external-event delivery (ADR 0004) (#2597)

* refactor(space): remove dead external-event reactivate/hold machinery with admission parity harness (#2578)

prepareExternalEventTask is the only producer of ExternalEventTaskDecision and
returns only deliver/fail since #886 removed runtime-owned PR-event
auto-subscription, making every reactivate/hold branch unreachable. Add a
transcript-based parity harness that pins the current admission behavior of
the external-event delivery chain (24 scenarios: live/busy/interrupted
sessions, activation, ownership and subscription gates, cooldown, digest,
in-flight claims, pending-node queue flush), then delete the dead machinery:

- deliverReactivatedExternalEvent and its two call sites
- resolveExternalEventDelivery and revalidateRecoveredTarget
- reactivation pre-scan in handleExternalEventImpl
- hold/reactivate decision branches in delivery, enqueue, and digest paths
- recoveryInFlight field (all writers removed)

Zero behavior change: parity harness green before and after deletion;
252 lines removed.

* refactor(space): extract external-event admission gates into pure functions (#2582)

* refactor(space): extract external-event delivery flow into superpipe pipeline + interpreter

Replace the 193-line branch cascade in deliverExternalEventToWorkflowTarget
with a pure decision pipeline (external-event-delivery-pipeline.ts) and a thin
effect-executing interpreter. Gates thread a ctx bound whole under a
single-name output spec; '!hasDecision' injected boolean pipes halt on the
first decision, preserving gate precedence (terminal > claim conflict >
subscription > task admission > session routing > execution routing).

Post-activation routing is a second pipeline covering activation failure,
activated session liveness, and not-active deferral. Both runners use the
sync .end() executor so the decision adds no microtask boundaries — the
timing-sensitive retry tests keep their baseline in... (continued)

Coverage Stats

43663 of 59091 branches covered (73.89%)

Branch coverage included in aggregate %.

146 of 150 new or added lines in 4 files covered. (97.33%)

2 existing lines in 1 file now uncovered.

59522 of 71289 relevant lines covered (83.49%)

3854.79 hits per line