You are viewing a plain text version of this content. The canonical link for it is here.
Posted to dev@felix.apache.org by "Karl Pauls (JIRA)" <ji...@apache.org> on 2017/11/24 11:22:00 UTC

[jira] [Commented] (FELIX-5751) Frequent JVM crashes during Bundle.getEntry()

    [ https://issues.apache.org/jira/browse/FELIX-5751?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=16265180#comment-16265180 ] 

Karl Pauls commented on FELIX-5751:
-----------------------------------

Interesting, yes - please try to reproduce with 5.6.10. We don't deploy snapshots regularly - hence, I'm unsure how recent that framework really is and there have been some changes. 

I'll try to look if I can see an easy way to synchronize that call in a cheap way.

> Frequent JVM crashes during Bundle.getEntry()
> ---------------------------------------------
>
>                 Key: FELIX-5751
>                 URL: https://issues.apache.org/jira/browse/FELIX-5751
>             Project: Felix
>          Issue Type: Bug
>          Components: Framework
>            Reporter: Neil Bartlett
>            Assignee: Karl Pauls
>         Attachments: hs_err_pid157.log, hs_err_pid1904.log, hs_err_pid3760.log, hs_err_pid87229.log, hs_err_pid88316.log
>
>
> Unclear whether there is anything Felix can do to fix or workaround this, but raising for info anyway...
> I am developing an application with Felix 5.7.0-SNAPSHOT (pulled from the Apache Snapshots repo, JAR built on Tue Nov 07 23:42:44 CET 2017).
> The nature of the application means that frequent calls are made to {{Bundle.getEntry()}}. Also because I am developing incrementally in Bndtools, the bundle is being updated frequently. Apparently as a result, I am getting JVM crashes every 5-10 minutes. The crash always happens in {{Bundle.getEntry()}}, specifically in the native {{getEntry(long,byte[],boolean)}} method of {{java.util.zip.ZipFile}}.
> It seems likely this is happening due to {{Bundle.getEntry}} being invoked concurrently with {{Bundle.update}}, so I wonder if Felix can implement any locking in this scenario? I have attached several hs_err files generated by the JVM.
> I am running Java 1.8.0_152-b16 on macOS Sierra (10.12.6 (16G29)). I have not yet replicated the crash with released Framework 5.6.10 but I am attempting to do so and will update this issue.



--
This message was sent by Atlassian JIRA
(v6.4.14#64029)