[ 
https://issues.apache.org/jira/browse/LUCENE-5212?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=13769281#comment-13769281
 ] 

Dawid Weiss commented on LUCENE-5212:
-------------------------------------

-XX:-DoEscapeAnalysis 

Yes, the bug is somewhere in the escape analysis/ ideal graph optimization 
chain so disabling it does the job. The first JVM revision that had this bug is 
jdk7u04, versions before that one are not affected.

The good news is that this problem seems to affect *only* 32-bit JVMs, I 
couldn't reproduce it on a 64-bit JVM. It may be due to a larger number of CPU 
registers (so spill stack variables are not needed and thus there are no missed 
updates...). If this is the case then the problem may still occur somewhere 
else, depending on the method/ inlines/ switches/, you name it...

The current status of LUCENE-5168 is that one of hotspot core developers 
(Vladimir Kozlov) has collected the information needed for debugging, we know 
after which commit to hotpot the problem starts appearing and we know it's 
still present both in jdk7 and in jdk8. I'd have to devote myself full time to 
getting up to speed with internal hotspot code to figure out anything more -- 
even the logs get really hairy (and huge) at the detail level we're looking at 
the problem. Hopefully Vladimir will be able to look into it at some point.



                
> java 7u40 causes sigsegv and corrupt term vectors
> -------------------------------------------------
>
>                 Key: LUCENE-5212
>                 URL: https://issues.apache.org/jira/browse/LUCENE-5212
>             Project: Lucene - Core
>          Issue Type: Bug
>            Reporter: Robert Muir
>         Attachments: hs_err_pid32714.log, jenkins.txt
>
>


--
This message is automatically generated by JIRA.
If you think it was sent incorrectly, please contact your JIRA administrators
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

Reply via email to