thecoop opened a new issue, #14930: URL: https://github.com/apache/lucene/issues/14930
### Description Following on from https://github.com/apache/lucene/pull/14844#discussion_r2168818779, there are cases where `IndexWriter.merge` does not close merge instances before closing the 'root' readers, leading to the necessity of the extra check in `AssertingKnnVectorsReader.close`. In particular, [SegmentMerger](https://github.com/apache/lucene/blob/f7e7c9fb271a009f9da3e3b40f2b50e421d92a2d/lucene/core/src/java/org/apache/lucene/index/IndexWriter.java#L5213) (which creates the merge reader instances) is not cleaned up if an [exception is thrown later](https://github.com/apache/lucene/blob/f7e7c9fb271a009f9da3e3b40f2b50e421d92a2d/lucene/core/src/java/org/apache/lucene/index/IndexWriter.java#L5222), which is where the original problem was found. There may of course be more instances elsewhere. Currently, there's some ambiguity in the role of merge reader instances - are they clones? Do they have their own independent lifecycle? Are they dependent on their 'parent' reader at all? They were just clones, but the introduction of `finishMerge` introduces a lifecycle that is not quite independent - add on top of the restrictions around how [ReadAdvice needs to be managed](https://github.com/apache/lucene/pull/14844). I think the way to solve this is to make merge instances properly closeable by replacing `finishMerge` with a full `close` implementation, with their own independent lifecycle. `SegmentMerger` would then be closeable, as part of tidying up after an exception during merging. Most readers wouldn't actually need the independence given to them, but it's better to be more permissive in what merge readers are, which can be ignored by most implementations, than less permissive, leading to lifecycle issues when a merge reader can't do what it needs to by its implementation. ### Version and environment details _No response_ -- 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: issues-unsubscr...@lucene.apache.org.apache.org For queries about this service, please contact Infrastructure at: us...@infra.apache.org --------------------------------------------------------------------- To unsubscribe, e-mail: issues-unsubscr...@lucene.apache.org For additional commands, e-mail: issues-h...@lucene.apache.org
