mikebridge commented on PR #41824: URL: https://github.com/apache/superset/pull/41824#issuecomment-4897745984
Round 3 pushed in `7affd4e624` — **every round-2 finding resolved**: | Round-2 finding | Resolution | |---|---| | Token timestamp minted before the retry loop | Minted fresh **per attempt** (`_mint_token`), so the timestamp is genuinely acquisition time | | Unfenced stale-break (6/6 lenses, one threaded repro) | Compare-before-delete fencing on the break path, immediate in-iteration add retry (a single-attempt caller that breaks an orphan gets the acquisition rather than donating it), break log at `info`, residual multi-breaker window documented next to the release fence's | | Prune O(entries) round-trips under the lock (5 lenses) | **Zero backend calls inside the lock**: prunes exactly the value keys the candidate loop observed missing (constant get + membership filter + set). `_value_exists` deleted outright — which also retires the "has() branch never executed" test gap by removing the branch | | 4 s staleness threshold vs clock skew (4 lenses) | Factor raised to 15 (30 s), with the skew trade-off documented on the constant | | `"token:nan"` never-stale wedge | `math.isfinite` guard | | Mutation gaps: fence unfalsifiable, breaker-not-fired unasserted, TTL unobserved, clean-hit-lock-free unpinned, staleness edges | **Five new kill-tests, each validated to fail under the exact mutation it guards** (blind-delete release, always-stale breaker, TTL-less add, always-prune, isfinite removal) | | Docstring/wording overclaims; function-scoped test imports; stale PR body | Invariant docstring names both best-effort windows explicitly; "may be unreachable" wording; superset imports hoisted to module top; PR body refreshed to the three-commit state | Suite: **388 passing**. CI is running fresh on this head (the previous red was an infra flake in the brew setup step, pre-hooks). -- 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]
