RFR (S) CR 8014886: @Contended fields can overrun oop maps

[Coleen Phillimore]

On 5/22/2013 7:42 PM, Daniel D. Daugherty wrote:
> On 5/22/13 4:38 PM, Aleksey Shipilev wrote:
>> Thanks for the review!
>>
>> On 05/23/2013 02:09 AM, Daniel D. Daugherty wrote:
>>> which does not include the vm.quick testsuite which is minimum
>>> testing required for RT_Baseline pushes unless the change is
>>> trivial.
>> Fair enough.
>> I'll get back with the test results as soon as I able to run vm.quick.
>>
>>> Do you have a planned time frame or time frames for these bugs?
>> Let me submit the remaining issues to the bugtracker to keep the things
>> in order tomorrow. I think all the functional bugfixes will be ready
>> early next week. What's the hotspot-rt -> hotspot-main merge window? I
>> would certainly like to have a few nightlies before those get into 
>> hs-main.
>
> RT_Baseline snapshots on Tuesday at 1900 PT. If the Tuesday nightly is
> good, then RT_Baseline usually pushes the snapshot -> Main_Baseline on
> Wednesday. On rare occasions, RT_Baseline will do two pushes in one
> week, but that has to coordinated with the current RT GK (Rickard
> Backman).
>
> In some cases, RT_Baseline will snapshot on Tuesday like normal, but
> will wait for two nights of nightly testing to get the full spectrum
> of configurations. However, this requires very close coordination
> with the RT GK and Runtime Team since we'll have to avoid pushing
> anything after the Tuesday snapshot before before the Wednesday
> nightly starts (again at 1900 PT).
>
>
>> Other issues can wait till the next integration window. (And being the
>> cleanups, they will naturally entail some bike-shedding).
>
> I have to say that the number of deferrals of issues to new bugs during
> a current bug's review cycle along with the general bug tail of 
> @Contended
> is making me very nervous. It's starting to feel like we're just 
> stumbling
> around wearing a bucket and have a very limited view of the horizon due
> to the bucket.
>
> If I was the gatekeeper this month, I would be entertaining the idea of
> backing out @Contended until it could be proven to be more stable via
> a whole lot of testing. Fortunately (for many reasons), I'm not the RT
> GK this month. :-)

I interpret this as a second request to fix all the bugs and cleanups in 
one change and let us review that change.   I don't think it needs to be 
backed out because it only affects the contended tests and that's not in 
the JDK yet (or is it?)   Is it breaking code that doesn't use it?

Thanks,
Coleen
>
> Dan
>