[jira] Commented: (JCR-1040) JCR2SPI: remove node operation missing in submitted SPI batch

["Julian Reschke (JIRA)" <ji...@apache.org>]

    [ https://issues.apache.org/jira/browse/JCR-1040?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel#action_12516380 ] 

Julian Reschke commented on JCR-1040:
-------------------------------------

This seems to be non trivial.

The code in o.a.j.jcr2spi.state.ChangeLog tries to sanitize the operations, but seems to go to far.

a) statusChanged EXISTING_REMOVED->REMOVED removes the Remove operation.

b) Taking out that code leads to a ConstraintViolation in checkIsSelfContained().

c) Taking out that check leads to a Batch being submitted which does contain the Remove operation, but now another issue becomes visible -- looking at the sequence of operations again:

1) remove "/a"
2) add "/a"
3) add "/a/b" 

The third operation's parent id now refers to the node "/a" that was removed in step 1, and thus will cause a failure in the SPI implementation (because that node is gone when step 3 is executed).

This requires more research...


> JCR2SPI: remove node operation missing in submitted SPI batch
> -------------------------------------------------------------
>
>                 Key: JCR-1040
>                 URL: https://issues.apache.org/jira/browse/JCR-1040
>             Project: Jackrabbit
>          Issue Type: Bug
>          Components: SPI
>            Reporter: Julian Reschke
>
> In JCR2SPI, the following sequence of operations seems to lead to an incorrect SPI batch being submitted:
> 1) remove "/a"
> 2) add "/a"
> 3) add "/a/b"
> 4) session.save()
> This seems to create an SPI batch where the first remove operation is missing.
> Note that the problem only seems to occur when step 3 is part of the sequence.
> Full Java source for test:
>     try {
>       if (session.getRepository().getDescriptor(Repository.LEVEL_2_SUPPORTED).equals("true")) {
>         Node testnode;
>         String name = "delete-test";
>           
>         Node root = session.getRootNode();
>         
>         // make sure it's there
>         if (! root.hasNode(name)) {
>           root.addNode(name, "nt:folder");
>           session.save();
>         }
>         
>         // now test remove/add in one batch
>         if (root.hasNode(name)) {
>           testnode = root.getNode(name);
>           testnode.remove();
>           // session.save(); // un-commenting this makes the test pass
>         }
>         
>         testnode = root.addNode(name, "nt:folder");
>         // add one child
>         testnode.addNode(name, "nt:folder"); // commenting this out makes the test pass
>         
>         session.save();
>       }
>     } finally {
>       session.logout();
>     }
>     
>     

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