[
https://issues.apache.org/jira/browse/LUCENE-3037?page=com.atlassian.jira.plugin.system.issuetabpanels:all-tabpanel
]
Robert Muir updated LUCENE-3037:
--------------------------------
Attachment: LUCENE-3037.patch
the problem is in our computation of number of skip Levels:
in this case: Math.log(243) / Math.log(3) == 0.499999999999, which is floored
to 4, but there are really 5 skipLevels here.
Attached is a patch to use StrictMath, which fixes the bug. Maybe there is a
cleaner way though... I don't like the floating point math here?
> TestFSTs.testRealTerms produces a corrupt index
> -----------------------------------------------
>
> Key: LUCENE-3037
> URL: https://issues.apache.org/jira/browse/LUCENE-3037
> Project: Lucene - Java
> Issue Type: Bug
> Reporter: Robert Muir
> Attachments: LUCENE-3037.patch, LUCENE-3037_test.patch, index.7z.001,
> index.7z.002, index.7z.003
>
>
> seems to be prox/skip related: the test passes, but the checkindex upon
> closing fails.
> ant test-core -Dtestcase=TestFSTs -Dtests.seed=-4012305283315171209:0
> -Dtests.multiplier=3 -Dtests.nightly=true
> -Dtests.linedocsfile=c:/data/enwiki.random.lines.txt.gz
> Note: to get the enwiki.random.lines.txt.gz you have to fetch it from hudson
> (warning 1 gigabyte file).
> you also have to run the test a few times to trigger it.
> ill upload the index this thing makes to this issue.
--
This message is automatically generated by JIRA.
For more information on JIRA, see: http://www.atlassian.com/software/jira
---------------------------------------------------------------------
To unsubscribe, e-mail: [email protected]
For additional commands, e-mail: [email protected]
