RFR (S) 8009829: CDS: JDK JPRT test fails crash in Symbol::equals()
Coleen Phillimore
coleen.phillimore at oracle.com
Tue Mar 12 11:24:12 PDT 2013
On 03/12/2013 01:17 PM, Ioi Lam wrote:
> This looks good to me.
>
> Could you cut-and-paste the C stack trace (if you still have it) into
> the bug report so we know the code path taken?
I cut and pasted the C stack from -Xshare:dump assert and attached the
hs_err_pid file.
thanks!
Coleen
>
> Thanks
> - Ioi
>
> On 03/11/2013 08:51 PM, Coleen Phillimore wrote:
>> On 3/11/2013 11:46 PM, David Holmes wrote:
>>> Hi Coleen,
>>>
>>> On 12/03/2013 7:51 AM, Coleen Phillimore wrote:
>>>> Summary: -Xshare:dump was creating a Symbol in C_heap. There's an
>>>> assert there that jdk jprt wasn't hitting because it was only done in
>>>> product
>>>
>>> The change seems reasonable - Reviewed.
>>>
>>> But I'm unclear what changed - why did we start creating a c_heap
>>> symbol during the dump?
>>
>> The change or C-heap symbol came from default method processing. I
>> think either a generic or method signature had a component in it that
>> hadn't been processed as a class name first which would have put it
>> in the CDS region. SignatureStream creates new symbols if they're
>> not found and I think this one used to be found (or has been
>> added). Apparently there have been a lot of changes since b80.
>>
>> Thanks!
>> Coleen
>>
>>>
>>> David
>>>
>>>> Tested with -Xshare:dump and jdk/test/sun/nio/cs tests.
>>>>
>>>> open webrev at http://cr.openjdk.java.net/~coleenp/8009829/
>>>> bug link at http://bugs.sun.com/view_bug.do?bug_id=8009829 (might not
>>>> be there yet)
>>>> local bug link https://jbs.oracle.com/bugs/browse/JDK-8009829
>>>>
>>>> thanks,
>>>> Coleen
>>>>
>>
>
More information about the hotspot-runtime-dev
mailing list