2.0.0-M5 REGRESSION: sessionClosed called before doDecode completes
-------------------------------------------------------------------
Key: DIRMINA-687
URL: https://issues.apache.org/jira/browse/DIRMINA-687
Project: MINA
Issue Type: Bug
Affects Versions: 2.0.0-M5
Reporter: Serge Baranov
Priority: Critical
The problem is described in detail at
http://markmail.org/message/3iqmkdcukhygxxva .
I did more tests by setting the CLOSED custom attribute inside sessionClosed
and by checking for this attribute in the middle of doDecode.
Under some networking conditions on the production server I can see the
following in the logs:
[2009-04-16 22:30:17,329] Session closed in the middle of decoding!
[2009-04-16 22:30:21,126] Session closed in the middle of decoding!
[2009-04-16 22:30:22,345] Session closed in the middle of decoding!
[2009-04-16 22:30:22,969] Session closed in the middle of decoding!
[2009-04-16 22:30:25,217] Session closed in the middle of decoding!
[2009-04-16 22:30:25,840] Session closed in the middle of decoding!
[2009-04-16 22:30:59,958] Session closed in the middle of decoding!
Unfortunately, I didn't manage to prepare an isolated test case for this issue,
must be some race condition which is tricky to catch.
Most likely the problem is caused by this commit:
http://svn.apache.org/viewvc?view=rev&revision=762167 .
As mentioned, this problem is new to 2.0.0-M5, worked fine for many months with
2.0.0-M4 and previous builds. Rolling back to M4 builds fixes the problem
instantly.
OrderedThreadPool must guarantee the order of events, sessionClosed must not be
called when doDecode is in progress.
--
This message is automatically generated by JIRA.
-
You can reply to this email to add a comment to the issue online.
