geronimo-dev mailing list archives

Site index · List index
Message view « Date » · « Thread »
Top « Date » · « Thread »
From "David Jencks (JIRA)" <j...@apache.org>
Subject [jira] [Commented] (YOKO-467) Fix a bunch of deadlocks by using executor services instead of thread groups
Date Sun, 19 Feb 2017 04:31:44 GMT

    [ https://issues.apache.org/jira/browse/YOKO-467?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=15873473#comment-15873473
] 

David Jencks commented on YOKO-467:
-----------------------------------

commit a1cab3bca0fbe660f99dde2bbea46504f1c53b02
Author: Neil Richards <ngmr@linux.vnet.ibm.com>
Date:   Tue May 19 00:01:46 2015 +0100

    Use ReadWriteLock for receiver threads / shutdown

commit 07e7ccf08e7f7993dc8d306c00b95cb9947fc165
Author: David Jencks <djencks@apache.org>
Date:   Sun May 17 00:23:20 2015 -0400

    Fix deadlock on some jdks between write and close by synchronizing shutdown using the
sendMutex_ already used for send

commit 52e1b45736e8b939f6286d3833aa37c5690a4b39
Author: David Jencks <djencks@apache.org>
Date:   Sat May 16 18:43:14 2015 -0400

    use internal ExecutorServices instead of thread groups in ORBInstance to remove client
deadlock due to expecting threads ending to notify their thread group



> Fix a bunch of deadlocks by using executor services instead of thread groups
> ----------------------------------------------------------------------------
>
>                 Key: YOKO-467
>                 URL: https://issues.apache.org/jira/browse/YOKO-467
>             Project: Yoko - CORBA Server
>          Issue Type: Bug
>      Security Level: public(Regular issues) 
>          Components: orb core
>    Affects Versions: v1.1.0
>            Reporter: David Jencks
>            Assignee: David Jencks
>
> Yoko internal thread usage relies on thread groups and some jdk-implementation specific
assumptions about notifications to detect thread shutdown. These assumptions are false on
(at least) ibm jdks.  Use of Executors is a lot clearer and avoids these deadlocks, as well
as promoting thread reuse.



--
This message was sent by Atlassian JIRA
(v6.3.15#6346)

Mime
View raw message