[
https://issues.apache.org/jira/browse/LUCENE-5377?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=13878709#comment-13878709
]
Robert Muir commented on LUCENE-5377:
-------------------------------------
{quote}
And now it fails, because writing with old Codecs is not supported. Maybe it
cannot be done like that for 4x because we need to support 3x indexes, but I
haven't tried (because in 4x we catch and ignore the UOE, so I don't know yet
how it will affect the test).
{quote}
Oh, thanks for investigating. Now i understand how it got thru the tests.
Impersonation masked the bug!!!!
{quote}
That's a good idea - if we strip the segment name from the files .si records
(which is redundant anyway!), we won't need to rewrite .si. I believe though
that we should fix this in 4.6.1?
{quote}
But thats the long-term fix. For the short term i think we should just allow
the old .si format to be written. Very simple, very clean and conservative
solution, just not elegant at all.
Yes, this is a corruption issue. I really think we should respin 4.6.1 for this.
> IW.addIndexes(Dir[]) causes silent index corruption
> ---------------------------------------------------
>
> Key: LUCENE-5377
> URL: https://issues.apache.org/jira/browse/LUCENE-5377
> Project: Lucene - Core
> Issue Type: Bug
> Components: core/index
> Affects Versions: 4.6
> Environment: windows/linux
> Reporter: Littlestar
> Attachments: TestCore_45.java, TestCore_46.java
>
>
> my old facet index create by Lucene version=4.2
> use indexChecker ok.
> now I upgrade to Lucene 4.6 and put some new records to index.
> then reopen index, some files in indexdir missing....
> no .si files.
> I debug into it, new version format of segments.gen(segments_N) record bad
> segments info.
--
This message was sent by Atlassian JIRA
(v6.1.5#6160)
---------------------------------------------------------------------
To unsubscribe, e-mail: [email protected]
For additional commands, e-mail: [email protected]
