Request for review: JDK-8011569 ARM: avoid native stack walking

[Zhengyu Gu]

Good to me.

-Zhengyu

On Jul 16, 2013, at 10:00 AM, Joseph Provino wrote:

> The latest webrev is here: http://cr.openjdk.java.net/~jprovino/8011569/webrev.00/ <http://cr.openjdk.java.net/%7Ejprovino/8011569/webrev.00/>
> 
> David Holmes already approved these changes when the review request was for 8017473.
> 
> thanks.
> 
> joe
> 
> 
> On 7/3/2013 4:52 PM, Joseph Provino wrote:
>> It's even more confusing now!  ;-)
>> 
>> https://jbs.oracle.com/bugs/browse/JDK-8017473 turns out to be a duplicate of
>> 
>> https://jbs.oracle.com/bugs/browse/JDK-8011569
>> 
>> I closed 8017473 as a duplicate and will make a webrev for 8011569
>> then backport to 7u.
>> 
>> More comments below as to what to do.
>> 
>> On 7/3/2013 4:15 PM, Vladimir Kozlov wrote:
>>> Thank you for explaining situation, it was confusing.
>>> 
>>> Since you are going to backport it into 7u40 the fix should be simple and targeted. For me the suggestion to use new PLATFORM_NATIVE_STACK_WALKING_SUPPORTED to fix this 8017473.
>>> 
>>> As you said NMT is nothing to do with this. So using PLATFORM_NATIVE_STACK_WALKING_SUPPORTED instead of PLATFORM_NMT_DETAIL_SUPPORTED is also questionable.
>> 
>> I'm not sure what you would like.  It is okay to change PLATFORM_NMT_DETAIL_SUPPORTED to
>> PLATFORM_NATIVE_STACK_WALKING_SUPPORTED for this bug fix?  I think this would be the easiest
>> way to clean it up.  If you have another way you'd rather see it done, let me know.
>> 
>>> Why you use "PLATFORM_" prefix in names?
>> 
>> That's a good question.  I thought there were other names like that but I don't see any now.
>> NATIVE_STACK_WALKING_SUPPORTED seems better but perhaps because it's platform specific
>> maybe it's okay as is?  I don't feel strongly either way...
>> 
>> thanks.
>> 
>> joe
>> 
>>> 
>>> Thanks,
>>> Vladimir
>>> 
>>> On 7/3/13 12:53 PM, Joseph Provino wrote:
>>>> On 07/03/2013 03:27 PM, Vladimir Kozlov wrote:
>>>>> I don't like to have renaming done together with the fix. Is renaming
>>>>> required?
>>>>> 
>>>>> Thanks,
>>>>> Vladimir
>>>> 
>>>> Renaming isn't required but if I keep PLATFORM_NMT_DETAIL_SUPPORTED
>>>> I would need to add the new flag PLATFORM_NATIVE_STACK_WALKING_SUPPORTED.
>>>> 
>>>> I could use PLATFORM_NMT_DETAIL_SUPPORTED but NMT doesn't have anything
>>>> to do
>>>> with this bug 8017473.
>>>> 
>>>> What happened is that 8011064 got reported and fixed first. The fix was
>>>> to disallow
>>>> NMT_detail in some cases so PLATFORM_NMT_DETAIL_SUPPORTED made sense.
>>>> 
>>>> Bug 8017473 makes it clear that there are times when native stack
>>>> walking can't be done.
>>>> PLATFORM_NATIVE_STACK_WALKING_SUPPORTED is more general and makes sense for
>>>> 8011064 and 8017473.
>>>> 
>>>> Do you think it would be better to use a new name and then file another
>>>> bug to change
>>>> PLATFORM_NMT_DETAIL_SUPPORTED to PLATFORM_NATIVE_STACK_WALKING_SUPPORTED?
>>>> 
>>>> joe
>>>> 
>>>>> 
>>>>> On 7/3/13 12:10 PM, Joseph Provino wrote:
>>>>>> Bug report: https://jbs.oracle.com/bugs/browse/JDK-8017473
>>>>>> 
>>>>>> This is for SE_8 but will be backported to 7u.
>>>>>> 
>>>>>> Webrev: http://cr.openjdk.java.net/~jprovino/8017473/webrev.00/
>>>>>> <http://cr.openjdk.java.net/%7Ejprovino/8017473/webrev.00/>
>>>>>> 
>>>>>> I changed PLATFORM_NMT_DETAIL_SUPPORTED to
>>>>>> PLATFORM_NATIVE_STACK_WALKING_SUPPORTED
>>>>>> to make the name more general.
>>>>>> 
>>>>>> Added -DPLATFORM_NMT_DETAIL_SUPPORTED=1 to linux/makefiles/debug.make
>>>>>> because
>>>>>> with the low optimization for debug builds, -fno-omit-frame-pointer is
>>>>>> set and stack walking
>>>>>> is always permissible.
>>>>>> 
>>>>>> Changed vm.make to optionally include an architecture specific makefile
>>>>>> in case some files
>>>>>> need to be compiled with special options such as
>>>>>> -fno-omit-frame-pointer.
>>>>>> 
>>>>>> Thanks.
>>>>>> 
>>>>>> joe
>>>> 
>> 
>