review request -- JDK-8011064 Some tests have failed with SIGSEGV on arm-hflt on build b82

[David Holmes]

Just for the benefit of others watching this, the undef approach is not 
ideal and there's another alternative in the pipeline.

David

On 22/05/2013 5:55 AM, JOSEPH PROVINO wrote:
> New webrev is here:
>
> http://cr.openjdk.java.net/~jprovino/8011064/webrev.01/
>
> joe
>
> On 5/21/2013 1:57 PM, JOSEPH PROVINO wrote:
>> That's a good solution.  Is there any reason NMT_detail_supported
>> can't be made const or
>> better yet eliminate the variable and just use if
>> (PLATFORM_NMT_DETAIL_SUPPORTED) ?
>>
>> joe
>>
>> On 5/20/2013 6:02 PM, BILL PITTORE wrote:
>>>  Maybe something like I did for SUPPORTS_NATIVE_CX8. This gets
>>> defined in globalDefinitions_<cpu>.hpp.
>>> In this situation maybe something like:
>>>
>>> memTracker.cpp
>>>   bool NMT_detail_supported = PLATFORM_NMT_DETAIL_SUPPORTED;
>>>
>>> then in globalDefinitions_<cpu>.hpp
>>>
>>> #define PLATFORM_NMT_DETAIL_SUPPORTED {true | false} as the case may be.
>>>
>>> Then you eliminate the need to add code to pd_initialize()
>>>
>>> bill
>>>
>>> On 5/20/2013 3:26 PM, JOSEPH PROVINO wrote:
>>>> Make it easy to disable NMT detail in platform specific code.
>>>>
>>>> A new variable NMT_detail_supported is defined and set to true.
>>>> It can be disabled in platform specific code as needed.
>>>>
>>>> Webrev is here: http://cr.openjdk.java.net/~jprovino/8011064/webrev/
>>>>
>>>> JDK-8011064 <https://jbs.oracle.com/bugs/browse/JDK-8011064>Some
>>>> tests have failed with SIGSEGV on arm-hflt on build b82
>>>> <https://jbs.oracle.com/bugs/browse/JDK-8011064>
>>>
>>>
>>
>