RFR: 8140584: nmethod::oops_do_marking_epilogue always runs verification code

[Stefan Karlsson]

Thanks, Vladimir.

StefanK

On 2015-11-11 02:26, Vladimir Kozlov wrote:
> Good.
>
> Thanks,
> Vladimir
>
> On 11/10/15 3:14 AM, Stefan Karlsson wrote:
>> Hi Thomas,
>>
>> Thanks for reviewing. Here's the updated patches:
>>   http://cr.openjdk.java.net/~stefank/8140584/webrev.03
>>   http://cr.openjdk.java.net/~stefank/8140584/webrev.03.delta
>>
>> Thanks,
>> StefanK
>>
>> On 2015-11-10 11:29, Thomas Schatzl wrote:
>>> On Mon, 2015-11-09 at 10:34 +0100, Stefan Karlsson wrote:
>>>> Hi Vladimir,
>>>>
>>>> What do you think about this version:
>>>>    http://cr.openjdk.java.net/~stefank/8140584/webrev.02/
>>>> http://cr.openjdk.java.net/~stefank/8140584/webrev.02.delta/
>>>>
>>>> Thanks,
>>>> StefanK
>>>   in relocInfo_x86.cpp, line 70 and 77: shouldn't these asserts be
>>> guarantees too?
>>>
>>> Also, the assert in NativeInstruction::verify_data64_sethi() in
>>> relocInfo_sparc.cpp needs to be a guarantee, as it is called by
>>> Relocation::pd_set_data_value().
>>>
>>> I also grepped the code for more "verify_only" parameters, and they 
>>> also
>>> use asserts instead of guarantees.
>>> Like Relocation::pd_set_data_value() in relocInfo_aarch64.cpp,
>>> relocInfo_ppc.cpp, in PhaseIdealLoop::build_and_optimize() in
>>> loopnode.cpp.
>>>
>>> Although I think the one in PhaseIdealLoop::build_and_optimize() is not
>>> executed via nmethod::oops_do_marking_epilogue().
>>>
>>> Thanks,
>>>    Thomas
>>>
>>>
>>