[
https://issues.apache.org/jira/browse/LUCENE-3255?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=13056619#comment-13056619
]
Mark Harwood commented on LUCENE-3255:
--------------------------------------
Thanks for the quick turnaround Mike/Simon.
Greg is on 2.9.x and so the suggestion I have is that he adds some checking
code in the app to remove the latest segments_n file if it looks to have
anything other than -9 as a format value given he knows that his Lucene indexes
should always be of that version. Maybe that could be a utility class that can
be posted here on this issue for others who might share this issue. I'm
guessing it's a quirk of the file system to leave that all-zeros file in place
prior to a flush of some kind?
> Corrupted segment file not detected and wipes index contents
> ------------------------------------------------------------
>
> Key: LUCENE-3255
> URL: https://issues.apache.org/jira/browse/LUCENE-3255
> Project: Lucene - Java
> Issue Type: Bug
> Affects Versions: 2.9.4, 3.2
> Reporter: Mark Harwood
> Assignee: Michael McCandless
> Fix For: 3.4
>
> Attachments: AllZerosSegmentFile, BadSegmentsFileTest.java,
> LUCENE-3255.patch, LUCENE-3255_testcase.patch
>
>
> Lucene will happily wipe an existing index if presented with a latest
> generation segments_n file of all zeros. File format documentation says
> segments_N files should start with a format of -9 but SegmentInfos.read
> accepts >=0 as valid for backward compatibility reasons.
--
This message is automatically generated by JIRA.
For more information on JIRA, see: http://www.atlassian.com/software/jira
---------------------------------------------------------------------
To unsubscribe, e-mail: dev-unsubscr...@lucene.apache.org
For additional commands, e-mail: dev-h...@lucene.apache.org
