RFR (XXS): 8134856: Incorrect use of PLAB::min_size() in MaxPLABSizeBounds

Thomas Schatzl thomas.schatzl at oracle.com
Tue Sep 1 16:52:58 UTC 2015


Hi Sangheon,

On Tue, 2015-09-01 at 09:47 -0700, sangheon.kim wrote:
> Hi Thomas,
> 
> Thank you for fixing this!
> 
> One question for the test.
> If we set 'G1HeapRegionSize=1M' and 'YoungPLABSize=2M', error message 
> comes from the constructor of G1CollectorPolicy.
>      jio_snprintf(buffer, sizeof(buffer), "%sPLABSize should be at most 
> " SIZE_FORMAT,
>                   OldPLABSize > region_size ? "Old" : "Young", region_size);

No. This "validation" has already been removed in an earlier RFR
(JDK-8134858).

> So VM exits before going to the command-line validation (i.e. 
> MaxPLABSizeBounds() ) routines.

No, it ends up in the new command line validation that has this bug.

> Am I missing something?

Thanks,
  Thomas





More information about the hotspot-gc-dev mailing list