[
https://issues.apache.org/jira/browse/OPENJPA-2507?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=14010860#comment-14010860
]
Ioannis Alexandrakis commented on OPENJPA-2507:
-----------------------------------------------
I have not seen the source of spring-data very closely to see if it caches
CriteriaQueryImpl (even though after your comment I suspect that that is the
case, or something of the sort, like modifying the contexts using
get/setContexts), but since this class was never intended to be accessed by
multiple threads, then there is no reason to protect that part of the class
only.
However, Stacks' methods are synchronized and there is only a setter for
contexts (these are the main differences from the other Maps/Lists inside
CriteriaQueryImpl), perhaps if there is an explanation to the problem would be
related to those. I am not saying this is worth the effort, but I am just
curious as to why we have a problem with the specific method (Context ctx()) in
two tickets which are probably related.
Just to mention it, since I tested the proposed fix myself, I haven't seen
something related to CriteriaQueryImpl again.
> Weird EmptyStackException in CriteriaQueryImpl
> ----------------------------------------------
>
> Key: OPENJPA-2507
> URL: https://issues.apache.org/jira/browse/OPENJPA-2507
> Project: OpenJPA
> Issue Type: Bug
> Components: criteria
> Affects Versions: 2.2.0, 2.3.0
> Environment: openjpa 2.2.0 & 2.3.0, spring-data-jpa up to 1.4.2
> Reporter: Ioannis Alexandrakis
> Priority: Minor
> Labels: criteria_api, openjpa
>
> Using spring-data-jpa with openjpa, I sometimes encounter a strange error
> (not all the times, but under heavy load it makes its appearance, sparsely
> though):
> ....
> Caused by: java.util.EmptyStackException
> at java.util.Stack.peek(Stack.java:<arbitrary line>)
> at
> org.apache.openjpa.persistence.criteria.CriteriaQueryImpl.ctx(CriteriaQueryImpl.java:<arbitrary
> line>
> ....
> I do not know which behaviour triggers it, however I think it would be an
> improvement to change (I did not know where exactly to file it, because it is
> both an improvement and a bug in my opinion), in
> org.apache.openjpa.persistence.criteria.CriteriaQueryImpl, method: Context
> ctx(), this:
> return _contexts == null || _contexts.isEmpty() ? null : _contexts.peek();
> to something like this:
> try {
> return _contexts == null || _contexts.isEmpty() ? null :
> _contexts.peek();
> } catch (EmptyStackException e) {
> return null;
> }
> , in order to prevent a case where multiple threads modify the "_contexts"
> between the evaluation of the inline "if".
> I am not able to reproduce it all the time, thus I can't create a useful
> test, neither have I created a patch due to the simplicity of the 'fix'.
> However I believe it is a harmless fix which could be considered a minor
> improvement.
--
This message was sent by Atlassian JIRA
(v6.2#6252)
