[
https://issues.apache.org/jira/browse/SOLR-8742?page=com.atlassian.jira.plugin.system.issuetabpanels:all-tabpanel
]
Steve Rowe reassigned SOLR-8742:
--------------------------------
Assignee: Steve Rowe
Attachment: SOLR-8742.patch
Patch switching out RAMDirectory for the new ByteBuffersDirectory, which
apparently doesn't have the read-past-EOF bug. See LUCENE-8438 and sub-issues,
where RAMDirectory has been deprecated on 7.X and will be removed in 8.0.
Andrzej's seed from 2017 ({{18BE1C093E093813}}) is the only one reported on
this issue that still reproduces on master. That seed, and a new failure's
seed {{DAAD1021FC4DC399}} from
[https://jenkins.thetaphi.de/job/Lucene-Solr-master-Linux/22808], reproduce
without the patch, and do not reproduce with it. I also beasted the suite with
the patch 100 times using Miller's beasting script, and there were no failures.
I'll commit shortly after precommit finishes.
> HdfsDirectoryTest fails reliably after changes in LUCENE-6932
> -------------------------------------------------------------
>
> Key: SOLR-8742
> URL: https://issues.apache.org/jira/browse/SOLR-8742
> Project: Solr
> Issue Type: Bug
> Reporter: Hoss Man
> Assignee: Steve Rowe
> Priority: Major
> Attachments: SOLR-8742.patch
>
>
> the following seed fails reliably for me on master...
> {noformat}
> [junit4] 2> 1370568 INFO
> (TEST-HdfsDirectoryTest.testEOF-seed#[A0D22782D87E1CE2]) [ ]
> o.a.s.SolrTestCaseJ4 ###Ending testEOF
> [junit4] 2> NOTE: reproduce with: ant test -Dtestcase=HdfsDirectoryTest
> -Dtests.method=testEOF -Dtests.seed=A0D22782D87E1CE2 -Dtests.slow=true
> -Dtests.locale=es-PR -Dtests.timezone=Indian/Mauritius -Dtests.asserts=true
> -Dtests.file.encoding=ISO-8859-1
> [junit4] ERROR 0.13s J0 | HdfsDirectoryTest.testEOF <<<
> [junit4] > Throwable #1: java.lang.NullPointerException
> [junit4] > at
> __randomizedtesting.SeedInfo.seed([A0D22782D87E1CE2:31B9658A9A5ABA9E]:0)
> [junit4] > at
> org.apache.lucene.store.RAMInputStream.readByte(RAMInputStream.java:69)
> [junit4] > at
> org.apache.solr.store.hdfs.HdfsDirectoryTest.testEof(HdfsDirectoryTest.java:159)
> [junit4] > at
> org.apache.solr.store.hdfs.HdfsDirectoryTest.testEOF(HdfsDirectoryTest.java:151)
> [junit4] > at java.lang.Thread.run(Thread.java:745)
> {noformat}
> git bisect says this is the first commit where it started failing..
> {noformat}
> ddc65d977f920013c5fca16c8ac75ae2c6895f9d is the first bad commit
> commit ddc65d977f920013c5fca16c8ac75ae2c6895f9d
> Author: Michael McCandless <[email protected]>
> Date: Thu Jan 21 17:50:28 2016 +0000
> LUCENE-6932: RAMInputStream now throws EOFException if you seek beyond
> the end of the file
>
> git-svn-id: https://svn.apache.org/repos/asf/lucene/dev/trunk@1726039
> 13f79535-47bb-0310-9956-ffa450edef68
> {noformat}
> ...which seems remarkable relevant and likely to indicate a problem that
> needs fixed in the HdfsDirectory code (or perhaps just the test)
--
This message was sent by Atlassian JIRA
(v7.6.3#76005)
---------------------------------------------------------------------
To unsubscribe, e-mail: [email protected]
For additional commands, e-mail: [email protected]
