RFR (S): CR 8012939: @Contended doesn't work correctly with inheritance

[Daniel D. Daugherty]

Aleksey,

Christian's request is reasonable. Please reconsider.

I'm not sure which baseline you were planning to use for this change,
but RT_Baseline has already pushed for this week for the HSX-25-B33
snapshot.

Dan


On 5/15/13 2:16 PM, Aleksey Shipilev wrote:
> That assumes we would not have the bikeshed argument about the test
> names and the subfolders (which we will). Really, I think it would be a
> good move to move all three regression tests for @Contended at once.
>
> -Aleksey.
>
> On 05/16/2013 12:13 AM, Christian Tornqvist wrote:
>> Seems like a trivial thing to just change 2 lines in each of the tests and
>> do a hg rename? It would be more of an effort to do this later than right
>> away.
>>
>> Thanks,
>> Christian
>>
>> -----Original Message-----
>> From: Aleksey Shipilev [mailto:aleksey.shipilev at oracle.com]
>> Sent: den 15 maj 2013 15:40
>> To: Christian Tornqvist
>> Cc: hotspot-dev at openjdk.java.net
>> Subject: Re: RFR (S): CR 8012939: @Contended doesn't work correctly with
>> inheritance
>>
>> I appreciate the effort of renaming the tests, but I would not like to risk
>> the integration of this change before the approaching milestone.
>> Can it be made after a fact (including other @Contended regression tests).
>>
>> -Aleksey.
>>
>> 05/15/2013 11:35 PM, Christian Tornqvist wrote:
>>> Hi Aleksey,
>>>
>>> Same thing here as in 8014509, could you please rename the test and
>>> put it in a folder that describes the area of the test?
>>>
>>> Thanks,
>>> Christian
>>>
>>> -----Original Message-----
>>> From: hotspot-dev-bounces at openjdk.java.net
>>> [mailto:hotspot-dev-bounces at openjdk.java.net] On Behalf Of Aleksey
>>> Shipilev
>>> Sent: den 14 maj 2013 18:24
>>> To: hotspot-dev at openjdk.java.net
>>> Subject: RFR (S): CR 8012939: @Contended doesn't work correctly with
>>> inheritance
>>>
>>> Hi,
>>>
>>> It seems we have a plenty of failures in new SQE tests for @Contended
>>> because of this. Please see the updated webrev for the issue:
>>>    http://cr.openjdk.java.net/~shade/8012939/webrev.01/
>>>
>>> This is the minimal fix. Most of the instance_size calculations use
>>> first_nonstatic_field_offset as the boundary for the instance fields
>> block.
>>> @Contended on the class hijacks this boundary with padding, and so the
>>> instance size is incorrect (also evidenced by the wrong assert).
>>>
>>> There is also a more thorough clean up of that code:
>>>    http://cr.openjdk.java.net/~shade/8012939/webrev.02/
>>>
>>> ...which is makes the intent cleaner. I can submit another RFE for
>>> that cleanup, although I think it prunes the source of the original
>> confusion.
>>> Speak up if you want me to split these, and we push the
>>> webrev.01 alone then.
>>>
>>> Testing:
>>>    - 8003985 and 8012939 tests on Linux x86_64/fastdebug
>>>    - JPRT run for both webrevs
>>>
>>> Thanks,
>>> Aleksey.
>>>
>>
>
>