mikebridge commented on PR #41824:
URL: https://github.com/apache/superset/pull/41824#issuecomment-4897581120

   Follow-up pushed in `47c50d169c` — **every panel finding above is 
addressed**:
   
   | Finding | Resolution |
   |---|---|
   | **H1** read-path locking (6/6) | Prune gated on *observed* eviction (clean 
hits never touch the lock), single non-blocking attempt, `cache.has()` with 
`get()` fallback, extracted to `_prune_evicted` and isolated so housekeeping 
can never turn a computed hit into a miss |
   | **H2** unfenced release (6/6) | Owner-token lock values; release deletes 
only on token match; `has()` collapses the critical section so TTL expiry is no 
longer plausible |
   | **H3** crash-orphaned lock (verified vs cachelib 0.13.0) | Token embeds 
its acquisition timestamp; waiters break locks older than 2×TTL with a warning |
   | MED prune-exception → lost hit | Own try/except in `_prune_evicted` |
   | MED untested invariants | Three new mutation-validated tests: timeout-skip 
(lock busy with a *fresh foreign* token → index never written unlocked), 
abandoned-TTL-less-lock breaking, and the prune-vs-store race (reader parked in 
its unlocked read while a writer stores → the locked re-read keeps the new 
entry, drops the dead one). Each fails under the exact mutation it guards — 
including one test the mutation check itself caught mis-armed and fixed before 
it shipped |
   | MED observability | Skipped store registrations log at `warning` with the 
index key; stale-lock breaks log at `warning` |
   | LOWs | Honest docstring (invariant scoped to atomic-`add` backends; 
`SimpleCache`/`NullCache` best-effort degradation named), orphaned-value 
trade-off comment, `cache: Cache` typing, module-level test imports, typed 
events, harness errors surfaced instead of swallowed, in-test retry budget for 
slow CI |
   
   Full `semantic_layers` suite: **377 passing**.


-- 
This is an automated message from the Apache Git Service.
To respond to the message, please log on to GitHub and use the
URL above to go to the specific comment.

To unsubscribe, e-mail: [email protected]

For queries about this service, please contact Infrastructure at:
[email protected]


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

Reply via email to