TianHengZhuang commented on PR #12384: URL: https://github.com/apache/seatunnel/pull/12384#issuecomment-5902282884
Thanks for such a detailed round, this was genuinely useful. **Issue 1 (compile)**: fixed in `08efdd66`. The map is now built per case inside an `assertRejected(key, value)` helper, so nothing gets reassigned after a lambda has captured it. The module compiles and the tests in `InfluxDBFactoryTest` actually run now. **Issue 2 (`0` for the two timeouts)**: went with `greaterOrEqual(0)`. I double-checked OkHttp: `0` means "no timeout" for both connect and read, so it is a real setting and not a broken value like `batch_size = 0`. Rejecting it would also have made the same key behave differently on the source and the sink, since the source side is unchanged. Behaviour is preserved, so these two need no incompatible-changes entry. The en and zh pages now say a negative value is rejected and that `0` means no timeout, and `zeroTimeoutsStayValid` pins it. My earlier comment saying `0` stays valid was describing the intent; the code disagreed with it at the time, and it does not anymore. **Issue 3 (pin tests)**: `usernameWithoutPasswordIsRejected` is the negative case for `bundled`, it goes red without `.bundled(USERNAME, PASSWORD)` in the rule. `multiTableSinkReplicaPassesValidation` now also runs `ConfigValidator.validateUnknownKeys` with the factory rule, so removing `MULTI_TABLE_SINK_REPLICA` from the optional list turns it red too. Merged `dev` in to clear the conflict. That side added the Helm chart and FakeSource entries to the incompatible-changes pages; both are kept alongside the InfluxDB one. -- 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]
