andygrove commented on PR #5847: URL: https://github.com/apache/datafusion-comet/pull/5847#issuecomment-5804788962
A correction to item 2 in my earlier review. #5613 changed shape in `c2d5a2836`. `shrink` now releases to Spark first and only then takes the bytes off `used` in `settle_release`. So it no longer decrements before the release. After the rebase, a failed release would still leave `consumer_usage` non-zero when `unregister` runs, and the abort path is still there. Could you either decrement `consumer_usage` before the JVM release or make the `unregister` check a `debug_assert` with a warning? Decrementing before the release is safe for the pool total because `used` stays charged until the release settles. -- 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]
