SEPURI-SAI-KRISHNA commented on PR #13161: URL: https://github.com/apache/seatunnel/pull/13161#issuecomment-6907320249
@DanielLeens Thanks, both are done in `b6a8882d87`. **Issue 1.** `lastAddTimeMicros` now has a field comment: microseconds since the epoch; why each row of a writer must be later than the one before, also within a millisecond and when the clock steps back; that it lives only in the writer instance, is not checkpointed, and a new or restored writer starts again from the clock; that rows of different writers are ordered only by their clocks; and that `add_time` is local time of the JVM, so it repeats an hour when daylight saving time ends. The PR description lists the same limits. **Issue 2.** The step and the formatting are now static methods, `nextAddTimeMicros(nowMillis, lastMicros)` and `formatAddTime(micros)`, and the tests no longer read the wall clock: - a 10,000 event burst at one fixed clock reading increases strictly and ends exactly 9,999 microseconds after it; - when the clock steps back one second, the next value is the last one plus one microsecond, and a later clock reading is taken as is; - the rendering is pinned in `UTC` and `Asia/Shanghai`, and in `America/New_York` across the end of daylight saving time on 2026-11-01, where one second later is written as `01:00:00` after `01:59:59`. That test documents the limit; I kept local time, because a change could reorder rows that an older version left unmerged. Dropping the `+ 1`, or taking the clock alone, fails the burst and rollback tests. This also settles the optional Mockito point from your earlier review: no add_time test calls `nextAddTime` on a mock of the writer any more. Connector tests: 27 run, 0 failures (JDK 17). -- 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]
