RFR (XXS) CR 8006997: ContendedPaddingWidth should be uintx
Aleksey Shipilev
aleksey.shipilev at oracle.com
Fri Feb 8 02:39:36 PST 2013
On 02/07/2013 12:55 PM, David Holmes wrote:
> On 7/02/2013 5:52 PM, Aleksey Shipilev wrote:
>> On 02/07/2013 03:43 AM, David Holmes wrote:
>>> I have to wonder whether leaving ContendedPaddingWidth as signed is not
>>> the simplest thing to do as it may be impractical, if not impossible, to
>>> get uniformity of the signed-ness of expressions across all platforms.
>>
>> Well, Vladimir I. suggested we booby-trap against negative values with
>> guarantee(...). Will that be sufficient?
>
> Sure - a number of flags get range checked.
Ok then. This is the new webrev:
http://cr.openjdk.java.net/~shade/8006997/webrev.01/
I haven't found another good place to stick the guarantee() to, so opted
to stick it before the actual use. I would be happy to stick it
somewhere else where the options are generally checked, but found no
non-platform/non-OS specific location.
Testing:
- Ad-hoc: fastdebug/release builds with negative value
- JPRT: full-cycle HotSpot testing
-Aleksey.
More information about the hotspot-dev
mailing list