Why on JDK6 only and not on JDK 5? I suspect it has to do with
java.io.getFiles() or String.hasCode behaving differently. Why does that
matter? We go through the list of bundles and install it. The order in
the list is dependent on implementation of java.io.File.listFiles() and
String.hashCode(), etc, as we searches files in a directory and put the
names in a Map. The order finally plays a role in resolution process.
Ideally it should not matter, because the OSGi platform carefully picks
up an exporter assuming that there are no split-packages. The exact
rules are specified in section #3.7 of the R4 spec:
/
The following list defines the preferences, if multiple choices are
possible, in order of decreasing priority:
" A resolved exporter must be preferred over an unresolved exporter. "
An exporter with a higher version is preferred over an exporter with a
lower version. "
An exporter with a lower bundle ID is preferred over a bundle with a
higher ID.
/
We are experiencing the problem because the assumption is broken.
Thanks,
Sahoo
Mitesh Meswani wrote:
> The root cause turns out to be because of split packages in
> EclipseLink. Recent changes to Glassfish uncovered the issue. A bug
> has been filed against EclipseLink
> (https://bugs.eclipse.org/bugs/show_bug.cgi?id=243265)to track the issue.
>
> Thanks,
> Mitesh
>
> Mitesh Meswani wrote:
>> I am investigating issue with JDK 6...
>>
>> Hong Zhang wrote:
>>> Yes, you are right! The problem went away after I switched to JDK 5.
>>>
>>> Sahoo wrote:
>>>> Kedar observed this as well and he was telling that he saw the
>>>> failures while running on JDK 6, where as JDK 5 was fine. What JDK
>>>> are you using? Try JDK 5 if applicable.
>>>>
>>>> Thanks,
>>>> Sahoo
>>>>
>>>> Hong Zhang wrote:
>>>>> I checked out a fresh workspace just now and I got one failure and
>>>>> one skip when running QL.
>>>>>
>>>>> stng] url=http://localhost:8080/jpainjectemf/jpa?testcase=llinit
>>>>> [testng] FAILED: test1
>>>>> [testng] java.lang.AssertionError: Unexpected Results
>>>>> expected:<true> but was:<false>
>>>>> [testng] at
>>>>> test.jpa.jpainjectemf.JpaInjectEMFTestNG.test1(JpaInjectEMFTestNG.java:71)
>>>>>
>>>>> [testng] ... Removed 26 stack frames
>>>>> [testng] SKIPPED: test2
>>>>>
>>>>> [testng] ===============================================
>>>>> [testng] jpainjectemf
>>>>> [testng] Tests run: 2, Failures: 1, Skips: 1
>>>>>
>>>>> However, Hudson reported all QL tests passed, so I am not sure
>>>>> why. Anyone seeing the same thing?
>>>>>
>>>>> Thanks,
>>>>>
>>>>> - Hong
>>>>>
>>>>> ---------------------------------------------------------------------
>>>>> To unsubscribe, e-mail: dev-unsubscribe_at_glassfish.dev.java.net
>>>>> For additional commands, e-mail: dev-help_at_glassfish.dev.java.net
>>>>>
>>>>
>>>> ---------------------------------------------------------------------
>>>> To unsubscribe, e-mail: dev-unsubscribe_at_glassfish.dev.java.net
>>>> For additional commands, e-mail: dev-help_at_glassfish.dev.java.net
>>>>
>>>
>>>
>>> ---------------------------------------------------------------------
>>> To unsubscribe, e-mail: dev-unsubscribe_at_glassfish.dev.java.net
>>> For additional commands, e-mail: dev-help_at_glassfish.dev.java.net
>>>
>>
>
> ---------------------------------------------------------------------
> To unsubscribe, e-mail: dev-unsubscribe_at_glassfish.dev.java.net
> For additional commands, e-mail: dev-help_at_glassfish.dev.java.net
>