[
https://issues.apache.org/jira/browse/UIMA-3513?page=com.atlassian.jira.plugin.system.issuetabpanels:all-tabpanel
]
Marshall Schor reopened UIMA-3513:
----------------------------------
Burn ran a test with Java "assertions" enabled ( -ea as a JVM argument). This
caused one of the assertions in the CasCopier to trip. This assertion was
checking if a Feature Structure's reference to the CAS is the same as the CAS
it's supposed to be in.
If the situation is that the FS is added to the CAS *before* the CAS is
wrapped, then the reference is not equal, of course, and the assertion trips.
This can be fixed in the CasCopier by getting a reference to the wrapped CAS.
One (unreliable?) technique that works in both cases I know of CAS wrapping is
to use:
{code}cas1 = (CAS) cas1.getLowLevelCas();{code}
This of course depends on the wrapper implementing getLowLevelCAS() as a foward
to the same method on the wrapped CAS.
> CasCoper test for views belonging to same CAS broken
> ----------------------------------------------------
>
> Key: UIMA-3513
> URL: https://issues.apache.org/jira/browse/UIMA-3513
> Project: UIMA
> Issue Type: Bug
> Components: Core Java Framework
> Affects Versions: 2.4.2SDK
> Reporter: Marshall Schor
> Assignee: Marshall Schor
> Fix For: 2.5.0SDK
>
>
> The logic is flawed. Example 1: if cas1's viewname and cas2's viewname are
> both null (could be both base CASes, I think), then it returns the result of
> null == null because it's comparing the viewnames (null strings) instead of
> the base CASes themselves.
> The tests further may fail if the CAS is wrapped. May need to use "equals"
> instead of == , but need to check the default impl of this.
--
This message was sent by Atlassian JIRA
(v6.1.5#6160)
