You signed in with another tab or window. Reload to refresh your session.You signed out in another tab or window. Reload to refresh your session.You switched accounts on another tab or window. Reload to refresh your session.Dismiss alert
* Refactor `BaseProducer::poll` to not return early, but instead
continue processing events until the passed timeout.
Refactor `BaseProducer::flush` to spend most time in `librdkafka`,
and whatever is left in `BaseProducer::poll`.
* Simplify and rely on cast truncation.
* Fix logic error so that we always poll at least once even when
timeout is `Duration::ZERO`.
* Introduce Deadline type to simplify `BaseProducer::flush` and
`BaseProducer::poll`.
* Add `From<Timeout>` impl for `Deadline`
* Ensure we always call `poll_event` at least once, even if we have
`Timeout::After<Duration::ZERO>` for a non-blocking call.
* Allow Deadline to express `Timeout::Never` losslessly.
* Refactor poll loop to be more idiomatic.
* Centralize clamping logic to Deadline.
* Remove extraneous From impl.
* Simplify `BaseProducer::poll` to rely on `From` impl.
* Don't block forever in poll when flushing.
* Remove this clamp, in favor of relying on remaining_millis_i32.
* Ensure we always poll even if we get a timeout from flush.
* Update changelog reflecting behavior change in `BaseProducer::poll`.
0 commit comments