RFR (S): 7129724: Fix location of core file in dump message on macosx

[Dmitry Samersoff]

David,

OK. Thank you for clarification.

-Dmitry

On 2012-06-26 16:00, David Holmes wrote:
> Hi Dmitry,
> 
> On 26/06/2012 6:30 PM, Dmitry Samersoff wrote:
>> Mikael,
>>
>> Stepping aside MAC OS related changes, this code doesn't
>> respect OS settings of coredump pattern and location.
> 
> This CR was simply trying to address an inaccurate error message
> regarding the "default location" for core files. It seems on OSX they
> are always in /cores/.
> 
>> So is it better to create os::coredump_pattern() function
>> and change it to handle /proc/sys/kernel/core_pattern
>> under Linux as well as MAC OS specific things?
> 
> This has been raised in the past. The VM may not have a way of readily
> determining the actual core location if it has been set via coreadm or
> other equivalent functionality in the OS. You can always file a RFE to
> try and do this, but it is out of scope for this CR.
> 
> David
> -----
> 
>> -Dmitry
>>
>>
>> On 2012-06-26 02:22, Mikael Vidstedt wrote:
>>>
>>> If HotSpot crashes the dump message prints the default location of the
>>> generated core file. On other "posix" like OSes the path is normally
>>> <cwd>/core or core.<pid>, but on macosx the cores always end up in
>>> /cores/core.<pid>. This change aims to reflect that. Verified manually
>>> on Linux and macosx. Passes JPRT.
>>>
>>> Webrev: http://cr.openjdk.java.net/~mikael/7129724/webrev.00
>>>
>>> Cheers,
>>> Mikael
>>>
>>
>>


-- 
Dmitry Samersoff
Java Hotspot development team, SPB04
* There will come soft rains ...