lsm / HyperNeo / 32543924661

80%

push

github

fix(daemon): route recovery handler's blocked write through casStatus (task #1195) (#2684)

* fix(daemon): route recovery handler's blocked write through casStatus (task #1195)

handleAliveStuckExecutions now flips the canonical task to blocked via
taskRepo.casStatus, with the expected set derived from the pilot-5
transition table (open/in_progress/rate_limited/usage_limited; a
user-parked 'stopped' is excluded). On superseded the handler logs,
skips the remaining work for that execution, and does not retry or
throw, so a park landing during the run-blocked emit wins. Flips the
phase-0 park-during-recovery characterization pin accordingly; the
park-during-spawn pin stays pinned for the staged phases.

* fix(daemon): terminate the tick when the recovery CAS is superseded (task #1195)

Review follow-up on #2684: returning 'none' after a superseded task CAS
let processRunTick continue the remaining recovery handlers with the
stale snapshot even though the run was already transitioned to blocked;
handleWaitingRebindExecutions then attempted the invalid blocked ->
blocked run transition and rejected the tick. The superseded path now
returns 'blocked' so the caller stops the tick for the run. Adds a
regression test with a waiting_rebind sibling execution.

* test(daemon): cover the casStatus won path for the stuck-agent block (task #1195)

Adds the third test in the park-during-recovery CAS suite: same seeded
exhausted agentStuckRecovery state, no concurrent park, asserting the
full normal blocked outcome (status, blockReason, recovery reason, run
and execution blocked, both notifications). Pins the won branch the
flipped superseded tests no longer exercise.

44785 of 60184 branches covered (74.41%)

Branch coverage included in aggregate %.

5 of 5 new or added lines in 1 file covered. (100.0%)

60823 of 72590 relevant lines covered (83.79%)

4043.77 hits per line