harmony-commits mailing list archives

Site index · List index
Message view « Date » · « Thread »
Top « Date » · « Thread »
From "Pavel Ozhdikhin (JIRA)" <j...@apache.org>
Subject [jira] Commented: (HARMONY-4972) [drlvm][jitrino]unbalanced monenter/monexit pair in HIR
Date Mon, 29 Oct 2007 05:45:50 GMT

    [ https://issues.apache.org/jira/browse/HARMONY-4972?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel#action_12538404
] 

Pavel Ozhdikhin commented on HARMONY-4972:
------------------------------------------

In the attached example, a synchronized method has been inlined and JIT inserted a monexit
instruction with re-throw of possible exception thrown in the inlined method (nodes 102 and
103). JIT does not generate a monitor exit loop in this case like javac does for synchronized
statements. When a synchronized method is called and monitor has been entered the monitor
object is valid. Unless it is possible to write a malicious code on jasm which will invalidate
monitor object before node 102, the example is generally harmless.

This case requires additional investigation. Until there is no real test showing the problem,
I suggest to lower priority of this bug.


> [drlvm][jitrino]unbalanced monenter/monexit  pair in HIR
> --------------------------------------------------------
>
>                 Key: HARMONY-4972
>                 URL: https://issues.apache.org/jira/browse/HARMONY-4972
>             Project: Harmony
>          Issue Type: Task
>          Components: DRLVM
>         Environment: compiled by JDK5.0
>            Reporter: Chengrong Shi
>            Priority: Critical
>         Attachments: test.rar
>
>
>     I found a puzzling case of the HIR. The attachment is the picture of the HIR. In
this pictrue, the control flows along the 2 in-edges of Node 98(the node whose id is 98) seem
to have different lock state.The control flow from the left in-edge is unlocked at the end
of Node 102, but the control flow from the right in-edge is not unlocked. Because if exception
has been thrown then the last monexit of Node 102 will not been executed.
>     Is this case legal? Or I missed something else?

-- 
This message is automatically generated by JIRA.
-
You can reply to this email to add a comment to the issue online.


Mime
View raw message