Review request: 8024990: JT_JDK: 11 failures with SIGSEGV on arm-sflt platforms in nightly fastdebug build

[Vladimir Kozlov]

I will wait the result of further investigation as Dean suggested.

Thanks,
Vladimir

On 9/26/13 12:22 PM, Jiangli Zhou wrote:
> Hi Vladimir,
>
> Here is updated webrev:
> http://cr.openjdk.java.net/~jiangli/8024990/webrev.01/.
>
> Thanks,
> Jiangli
>
> On 09/26/2013 11:50 AM, Jiangli Zhou wrote:
>> Hi Vladimir,
>>
>> On 09/26/2013 11:26 AM, Vladimir Kozlov wrote:
>>> Should it be done only for (stub_id ==
>>> Runtime1::load_appendix_patching_id)?
>>
>> That sounds safer. I'll add that.
>>
>>> Neither bug report or your description say that it needs to be
>>> removed for load_klass_or_mirror_patch_id too.
>>
>> I should have included the info when sending the review request. Sorry
>> about that.
>>
>> Just so I understand it more, why 'copy_buff -= *byte_count' was there
>> for load_klass_or_mirror_patch_id? Was there a case where instructions
>> need to be re-winded?
>>
>> Thanks,
>> Jiangli
>>
>>>
>>> Thanks,
>>> Vladimir
>>>
>>> On 9/26/13 11:12 AM, Jiangli Zhou wrote:
>>>> Hi Vladimir,
>>>>
>>>> The copy_buff is at the 'ldr' instruction already which is the one we
>>>> want to patch. Rewinding the copy_buff by *byte_count causes the wrong
>>>> instruction being patched. I hit an assertion after enabling the patch
>>>> code is enabled for load_appendix_patching_id. That's why I removed the
>>>> line.
>>>>
>>>> Thanks,
>>>> Jiangli
>>>>
>>>> On 09/26/2013 10:12 AM, Vladimir Kozlov wrote:
>>>>> Why next line is removed?:
>>>>>
>>>>> -            copy_buff -= *byte_count;
>>>>>
>>>>> Thanks,
>>>>> Vladimir
>>>>>
>>>>> On 9/26/13 9:42 AM, Jiangli Zhou wrote:
>>>>>> Hi,
>>>>>>
>>>>>> Please review the fix for 8024990:
>>>>>>
>>>>>>    http://cr.openjdk.java.net/~jiangli/8024990/webrev.00/
>>>>>>
>>>>>> Needs to enable instruction patching for
>>>>>> Runtime1::load_appendix_patching_id.
>>>>>>
>>>>>> Thanks,
>>>>>> Jiangli
>>>>
>>
>