[ https://issues.apache.org/jira/browse/LUCENE-7325?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=15324584#comment-15324584 ]
Michael McCandless commented on LUCENE-7325: -------------------------------------------- bq. The attached patch fixes a bug in GeoPointPrefixTermsEnum and removes GeoPointTestUtil from TestGeoPointQuery. I'm a little confused: the patch fixes a different bug, that was uncovered by improving testing here by removing {{GeoPointTestUtil}}? Can you add an explicit test case showing that bug? bq. Specifically, This change which Robert's comment explains quite nicely. I see no comment there? Maybe the link isn't working for me ... the link seems to take me to a local diff where we no longer quantize the values at query time. bq. The fix (mentioned in the comment) is to move to the stable 64 bit encoding in LUCENE-7186. But GeoPointField will need bwc with the 62 bit encoding But, today, {{GeoPointField}} is still using 62 bit encoding right? (And, yes, switching to 64 bit will need back compat!). +1 for the test improvement (removing {{GeoPointTestUtil}}), but it's maybe good news it doesn't provoke new failures ;) And it looks like my test is invalid! So just remove it! And I agree this should not be a blocker anymore. > GeoPointInBBoxQuery no longer includes boundaries? > -------------------------------------------------- > > Key: LUCENE-7325 > URL: https://issues.apache.org/jira/browse/LUCENE-7325 > Project: Lucene - Core > Issue Type: Bug > Affects Versions: 6.1 > Reporter: Michael McCandless > Priority: Blocker > Attachments: LUCENE-7325.patch, LUCENE-7325.patch > > > {{GeoPointInBBoxQuery}} is supposed to include boundaries, and it does in 5.x > and 6.0, but in 6.1 something broke. -- This message was sent by Atlassian JIRA (v6.3.4#6332) --------------------------------------------------------------------- To unsubscribe, e-mail: dev-unsubscr...@lucene.apache.org For additional commands, e-mail: dev-h...@lucene.apache.org