mikemccand commented on issue #13867: URL: https://github.com/apache/lucene/issues/13867#issuecomment-2400800507
> The PR also changed the default value of the compress constructor parameter of `Lucene99HnswScalarQuantizedVectorsFormat `from [true to false](https://github.com/apache/lucene/pull/13767/files#diff-2c2e9079edea55e44171f704b7e0ab0058d8251c7c5bd07e855e359eae84d795L70). Note that `compress=true` when `int bits` is `7` does nothing -- it silently behaves as if `compress=false` yet it does then record `compress=true` into the index (which later leads to the back-compat break / exception in 9.12.0). I am trying to understand why our bwc indices test failed to catch this. I see we have a [`TestInt8HnswBackwardsCompatibility`](https://github.com/apache/lucene/blob/e4092396d55064009de51d00a9b0c9f03a07ad83/lucene/backward-codecs/src/test/org/apache/lucene/backward_index/TestInt8HnswBackwardsCompatibility.java) which (I think) was run in 9.11.1 to generate the bwc indices e.g. [`int8_hnsw.9.11.zip`](https://github.com/apache/lucene/blob/2da5a867d7ac4db133800b1fae73bae5c4b7cf29/lucene/backward-codecs/src/test/org/apache/lucene/backward_index/int8_hnsw.9.11.1.zip). Confusingly, it is named `int8` but I think it actually generated default (7 bit quantization) indices (still trying to verify that)... I'd like to get the bwc test to fail properly, first, then make @benwtrent's above proposed change, and see it pass... -- 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
