mikemccand commented on pull request #2088:
URL: https://github.com/apache/lucene-solr/pull/2088#issuecomment-731222134
> > As long as FieldNumbers.clear() is only called from
IndexWriter.deleteAll(), my understanding is that the safety is provided by the
try (Closeable finalizer = docWriter.lockAndAbortAll()) { block, which (I
think) guarantees that any concurrent indexing is blocked until the lock is
released.
>
> Thanks, maybe there is a way to improve the testing of `deleteAll` to
better enforce this? Lucene is not testing this method much, but I know some
users (e.g. Solr) are using it often. My concern is some race condition that
ultimately creates segments with unaligned field numbers. This would be a
disaster and definitely result in corruption (think, stored fields merging etc
which copies binary/compressed data directly).
Maybe a new unit test that randomly calls `deleteAll()` and other IW APIs
across threads?
Hmm, actually, it looks like `TestIndexWriter.testThreadInterruptDeadlock`
is doing just that already.
`TestIndexWriterDelete` tests specific interesting cases.
`TestIndexWriterMergePolicy` seems to test the interaction of `deleteAll` and
merge-on-commit.
----------------------------------------------------------------
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.
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
