Re: [I] Test failure in TestHnswFloatVectorGraph [lucene]
msokolov commented on issue #12945: URL: https://github.com/apache/lucene/issues/12945#issuecomment-1858117628 I spent a while staring at the previous refactor commit, and: (1) I am convinced the refactor you did is way cleaner.(2) it was a lot to do in a refactor and it is hard to mentally convince myself there were no functional changes. My best guess is that something changed about the order in which the segments are being merged, but I don't see how that is the case. -- 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 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
Re: [I] Test failure in TestHnswFloatVectorGraph [lucene]
msokolov commented on issue #12945: URL: https://github.com/apache/lucene/issues/12945#issuecomment-1858052541 one weird thing I noticed is the failing test seed only fails when `-Ptests.directory=MMapDirectory` is present -- 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 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
Re: [I] Test failure in TestHnswFloatVectorGraph [lucene]
benwtrent commented on issue #12945: URL: https://github.com/apache/lucene/issues/12945#issuecomment-1856557362 Bumping the searched vectors to 70 from 60 makes the test pass, but this still bugs be a bit as that commit shouldn't have changed any behavior... -- 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 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
Re: [I] Test failure in TestHnswFloatVectorGraph [lucene]
benwtrent commented on issue #12945: URL: https://github.com/apache/lucene/issues/12945#issuecomment-1856480501 This is interesting, that commit shouldn't have changed anything, just a refactor. I have confirmed I can repeat it (after several attempts), but cannot when going to the commit before. One suspect place is the `oldToNewOrdinal` mapping, but I just added a test like this: ``` Map originalOrdMap = getOldToNewOrdinalMap(mergedVectorIterator); assert oldToNewOrdinalMap.length == originalOrdMap.size(); for (int i = 0; i < oldToNewOrdinalMap.length; i++) { assert oldToNewOrdinalMap[i] == originalOrdMap.get(i): "ordinal maps should be equal " + oldToNewOrdinalMap + " vs " + Arrays.toString(oldToNewOrdinalMap); } ``` Where `getOldToNewOrdinalMap` is effectively copy-pasted from how it was before and this does not trip when this test fails. So, back to digging. -- 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 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
Re: [I] Test failure in TestHnswFloatVectorGraph [lucene]
msokolov commented on issue #12945: URL: https://github.com/apache/lucene/issues/12945#issuecomment-1856057317 Here, `git bisect` identifies [18bb826564bb16fde70bab3c06a167280b6cc632] Extract the hnsw graph merging from being part of the vector writer (#12657) as the commit where this test case seed started failing -- 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 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