request for review (XS), 6458402: 3 jvmti tests fail with CMS and +ExplicitGCInvokesConcurrent

[Igor Veresov]

On 12/3/10 7:08 PM, David Holmes wrote:
> Keith McGuigan said the following on 12/03/10 23:19:
>> On Dec 3, 2010, at 3:30 AM, Igor Veresov wrote:
>>
>>> You're calling notify_gc_end() while holding the FullGCCount_lock. Is
>>> that ok? Would it be more prudent to divert control to the user
>>> without holding it?
>>
>> The notify_gc_end() call does not divert control to the user -- it
>> contains only the code that will trigger a dtrace event (which is a
>> no-op as far as the VM is concerned; the OS may notice it though).
>>
>> That said, I think it would be a good idea to release the lock before
>> calling it and I'll change the code to do that. That way if someone
>> adds code to the notify_gc_end() method in the future there won't be
>> any surprises.
>
> But if you release the lock before firing the probe the two events can
> become distant in time and someone else can see a second "start" probe
> fire before they've seen the previous "end" probe fire. That makes it
> harder to use the probes to analyse what is happening.
>

But it's all in a VM thread, so it's going to be serialized.
Anyway, like Ramki noted, it doesn't report what it is supposed to...

igor