You are viewing a plain text version of this content. The canonical link for it is here.
Posted to dev@uima.apache.org by "Jim Challenger (JIRA)" <de...@uima.apache.org> on 2015/04/20 19:53:58 UTC

[jira] [Closed] (UIMA-4336) DUCC CLI: Different results with different JRE

     [ https://issues.apache.org/jira/browse/UIMA-4336?page=com.atlassian.jira.plugin.system.issuetabpanels:all-tabpanel ]

Jim Challenger closed UIMA-4336.
--------------------------------
    Resolution: Fixed

> DUCC CLI: Different results with different JRE
> ----------------------------------------------
>
>                 Key: UIMA-4336
>                 URL: https://issues.apache.org/jira/browse/UIMA-4336
>             Project: UIMA
>          Issue Type: Bug
>          Components: DUCC
>    Affects Versions: 1.1.0-Ducc
>            Reporter: Jim Challenger
>            Assignee: Jim Challenger
>             Fix For: 2.0.0-Ducc
>
>
> So far this only affects the service API/CLI, perhaps because it is more complex than the others.
> The net is the API produces ClassNotFound exceptions under Oracle 1.7 JREs and later, but work in all other (IBM and Apple-supplied JREs of all tested vintages).
> Everything works under IBM JREs but the fields that are initialized by initializes in the declarations in the returned objects are initialized a bit differently; in particular, at least one string field is initialized to null despite being declared like this:
> String foo = "N/A";  If the field is updated as part of executing the CLI in the SM it is returned correctly.
> Only the query was working in the Oracle JREs.  Other SM APIs returned ClassNotFound for the base class returned by the query that works!
> An older JRE produced by Sun works the same as the IBM JRE.
> I believe this is a discrepancy and possibly a bug in the Oracle JREs.
> The "solution" was not to change the DUCC classloader but instead to declare the returned API objects to look fully like Java Beans with no-parameter constructors and getter/setter methods.  It's unclear why this matters but any port in a storm ...



--
This message was sent by Atlassian JIRA
(v6.3.4#6332)