villebro commented on PR #43407: URL: https://github.com/apache/superset/pull/43407#issuecomment-5396705166
Thanks for the review — all three findings are addressed in follow-up PRs against this branch: - **P1 (async chart-data hang):** [#43461](https://github.com/apache/superset/pull/43461). The `202` is returned when tasks are *scheduled*, not finished, so the shared poll cursor is `<= now < any task's future terminal` — the bug was the `await baselineReady` sitting between the `202` and waiter registration, which let a fast task finish + a concurrent chart's poll advance the cursor past it (socket event dropped, no waiter yet). Fix: register the waiter **synchronously** in the same tick as the `202`; no `task_ids` lookup needed. - **P1 (guest_key too long):** [#43461](https://github.com/apache/superset/pull/43461). `guest-` + 64-hex = 70 chars; widened `task_subscribers.guest_key` to `String(128)` in the model + the folded migration. - **P2 (ownership-less lock release):** [#43462](https://github.com/apache/superset/pull/43462). Per-acquisition token + compare-and-delete on both the Redis and KV paths, threaded through the `DistributedLock` context manager; `token=None` preserves the cross-process Excel-export caller unchanged. Both PRs are green (unit tests + mypy/ruff/pre-commit); each carries regression tests. -- 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]
