RFR 8058927: ATG throws ClassNotFoundException

[Coleen Phillimore]

Hi David,
I already checked this in.  Sorry to not wait until you weren't 
traveling (but I was traveling myself).

On 9/26/14, 9:22 PM, David Holmes wrote:
> Sorry Coleen but I don't quite get the changes.
>
> javaClasses.cpp:
>
>  639     assert(class_loader() == k->class_loader(), "should be same");
>
> I can't quite get my head around the different things that k may be by 
> the time we get to the assert, and its relation to the current class.

k is the current class.  I added the assert because I could have taken 
out the class_loader parameter to create_mirror, because we can get 
class_loader out of the current class, but I didn't want to make the 
change harder to merge since there have been a few changes in this area 
lately.

> ---
>
> arrayKlass.cpp
>
>   java_lang_Class::create_mirror(k, Handle(THREAD, k->class_loader()), 
> Handle(NULL), CHECK);
>
> I expected to see something extracting the classloader of the element 
> type here - but perhaps I'm misunderstanding what k represents.

The current class's class loader is already set to the classloader of 
the element type.   An assert to that effect here might have been useful.

>
> ---
>
> test/runtime/LoadClass/ShowClassLoader.java
>
> Sorry, my little 'test' program was really a demo. The actual test 
> should be checking that for an array 'a':
>
> a.getClass().getClassLoader() ==
>    a.getClass().getComponentType().getClassLoader()
>
> and for completeness it should test a primitive array, object array 
> for a boot class, and object array for a non-boot class. So the test 
> name should then be CheckArrayClassLoader.

Yes, that would have been a better test...  and worth filing a bug to 
improve it.   I couldn't find a way to complain about the lack of this 
test in the JCK tests, which I think would be better since my motivation 
was to make this simply a regression test.

Thanks,
Coleen
>
> Thanks,
> David
>
> On 26/09/2014 12:05 PM, Coleen Phillimore wrote:
>>
>> Thanks for pointing out I cut/pasted the wrong webrevs.  Here they are:
>>
>> open webrev at http://cr.openjdk.java.net/~coleenp/8058927/
>> bug link https://bugs.openjdk.java.net/browse/JDK-8058927
>>
>> Thanks,
>> Coleen
>>
>> On 9/25/14, 7:33 PM, Coleen Phillimore wrote:
>>> Summary: ClassLoader for array klass was set to null and not the class
>>> loader of the component type.
>>>
>>> open webrev at http://cr.openjdk.java.net/~coleenp/8058843
>>> bug link https://bugs.openjdk.java.net/browse/JDK-8058843
>>>
>>> See bug for more details.
>>>
>>> Ran testbase, jtreg and jck tests, not that they found this bug in the
>>> first place.  Boris verified the fix with ATG.
>>>
>>> Thank you to David Holmes for finding it through visual inspection of
>>> the original change.
>>>
>>> Thanks,
>>> Coleen
>>