RFR (S): 8135025: Error message is repeated for large value at G1ConcRefinementThreads

[sangheon.kim]

Hi all,

After some discussion, I decided to propose this webrev.01 version which 
has different style from webrev.00.

The difference is to add public factory function and privatize the 
constructor.
So that newly added 'public bool initialize_threads() (from webrev.00)' 
is not needed.
With this approach we can minimize public function and handle the error 
situation correctly.

Can I get some reviews for this?
http://cr.openjdk.java.net/~sangheki/8135025/webrev.01/

Thanks,
Sangheon


On 09/04/2015 10:18 AM, sangheon.kim wrote:
> Hi all,
>
> Can I get some reviews for handling thread creation failure on 
> ConcurrentG1Refine?
>
> This pre-existing problem is revealed during commandline test for GC 
> flags(8078555) which tries to create for 'max_jint-1' threads.
> The reason of repeating error message when thread creation failed is 
> that we call 'vm_shutdown_during_initialization' when fails.
> And this function just returns and then 'for loop' is continued for 
> 'max_jint-1'.
>
> I separated thread initialization part from the constructor so that we 
> can return when fails to create a new native thread.
> And it will be handled from G1CollectedHeap::initialize().
>
> I didn't add test because it is not stable.
> From windows 64, it takes really long time to fail a thread creation 
> failure and hangs the system.
> And 8134995(REDO: GC flag validation) will cover this situation.
>
> CR: https://bugs.openjdk.java.net/browse/JDK-8135025
> Webrev: cr.openjdk.java.net/~sangheki/8135025/webrev.00
> Testing: JPRT
>
> Thanks,
> Sangheon