RFR (s): 8009531: Crash when redefining class with annotated method

[serguei.spitsyn at oracle.com]

Copy usually means copying values by pointers.
Not clear what's wrong with the "set_annotations" as there is no such 
function yet.
Probably, you find it confusing too.
But I'm not insisting on the copy_annotation_pointers() either. :)
Let's keep it as it is.

Thanks,
Serguei


On 3/26/13 3:58 PM, Coleen Phillimore wrote:
>
> It copies the pointers.  I can't change it to set_annotations because 
> there are already set_ functions.  O could change to 
> copy_annotation_pointers() if you insist.
>
> Coleen
>
> On 03/26/2013 07:00 PM, serguei.spitsyn at oracle.com wrote:
>> Coleen,
>>
>> This does not look like a clone or copy, it just sets the value?
>>   366 void ConstMethod::copy_annotations(ConstMethod* cm) {
>>   367   if (cm->has_method_annotations()) {
>>   368     assert(has_method_annotations(), "should be allocated already");
>>   369     set_method_annotations(cm->method_annotations());
>>   370   }
>>   ...
>>
>> Do we have to actually clone the annotations?
>> If not, then the name "copy_annotations" is wrong.
>> It must be "set_annotations".
>>
>> The test fixes look Ok.
>>
>> Thanks,
>> Serguei
>>
>>
>> On 3/26/13 3:11 PM, Coleen Phillimore wrote:
>>> Summary: Neglected to copy the annotations in clone_with_new_data 
>>> when they were moved to ConstMethod.
>>>
>>> open webrev at http://cr.openjdk.java.net/~coleenp/8009531/
>>> bug link at http://bugs.sun.com/view_bug.do?bug_id=8009531
>>>
>>> Also. please review JDK test modified to test that this crash is 
>>> fixed (will check in in two weeks).
>>>
>>> open webrev at http://cr.openjdk.java.net/~coleenp/8009531_jdk
>>>
>>> Thanks,
>>> Coleen
>>>
>>>
>>>
>>>
>>>
>>>
>>
>

-------------- next part --------------
An HTML attachment was scrubbed...
URL: http://mail.openjdk.java.net/pipermail/serviceability-dev/attachments/20130326/15a3b919/attachment.html