[
https://issues.apache.org/jira/browse/LUCENE-2637?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=12907559#action_12907559
]
Robert Muir commented on LUCENE-2637:
-------------------------------------
bq. If somebody is able to fix the NIO code to work in all cases (with
rock-solid tests), we can reenable it again (but only in NIOFSDir).
I completely agree.
I would also propose that if we go this route (we should open a separate issue
if we want to do this!),
that we only target trunk, and not 3.x.
the backport could always happen later after we feel comfortable that the
optimization is safe,
but i don't like that this will be the second time we have to inform users
about index corruption
on our "stable" branch from this :(
> FSDirectory.copyBytes isn't safe for SimpleFSDirectory
> ------------------------------------------------------
>
> Key: LUCENE-2637
> URL: https://issues.apache.org/jira/browse/LUCENE-2637
> Project: Lucene - Java
> Issue Type: Bug
> Components: Tests
> Affects Versions: 3.1, 4.0
> Reporter: Robert Muir
> Assignee: Robert Muir
> Attachments: LUCENE-2637.patch, LUCENE-2637.patch
>
>
> the copyBytes optimization from LUCENE-2574 is not safe for
> SimpleFSDirectory, but works fine for NIOFSDirectory.
> With SimpleFSDirectory, the copyBytes optimization causes index corruption.
> see
> http://www.lucidimagination.com/search/document/36d2dbfc691909d5/bug_triggered_by_testindexwriter_testrandomstoredfields
> for background
> here are my steps to reproduce (most of the time, at least on windows):
> {noformat}
> 1. edit line 87 of TestIndexWriter to plugin the seed:
> random = newRandom(3312389322103990899L);
> 2. edit line 5138 of TestIndexWriter to force SimpleFSDirectory:
> Directory dir = new SimpleFSDirectory(index);
> 3. run this command:
> ant clean test-core -Dtestcase=TestIndexWriter
> -Dtestmethod=testRandomStoredFields -Dtests.iter=10
> -Dtests.codec="MockVariableIntBlock(29)"
> {noformat}
--
This message is automatically generated by JIRA.
-
You can reply to this email to add a comment to the issue online.
---------------------------------------------------------------------
To unsubscribe, e-mail: [email protected]
For additional commands, e-mail: [email protected]
