RFR(xxs): 8144702: Using tid decorator in Unified Logging may crash VM
Thomas Stüfe
thomas.stuefe at gmail.com
Sun Dec 6 05:45:18 UTC 2015
Hi David,
On Sat, Dec 5, 2015 at 10:23 PM, David Holmes <david.holmes at oracle.com>
wrote:
> On 4/12/2015 10:05 PM, Thomas Stüfe wrote:
>
>> Hi all,
>>
>> please review this tiny fix:
>>
>> webrev:
>>
>> http://cr.openjdk.java.net/~stuefe/webrevs/8144702-Using-tid-decorator-in-Unified-Logging-may-crash-VM/webrev.00/webrev/
>> bug: https://bugs.openjdk.java.net/browse/JDK-8144702
>>
>> Unified Logging uses Thread::current() when printing out the "tid"
>> decorator, which will crash the VM if logging is done after thread start
>> and before thread TLS slot initialization or after thread TLS slot cleanup
>> and before thread finish.
>>
>
> Thanks for fixing that. This indicates a hole in our UL testing if we did
> not encounter an assertion failure with the use of decorators in the
> earliest stages of logging ... or did you find this when you added your own
> logging that occurs very early during initialization.
>
>
The latter. I used logging to trace thread creation to find out why AIX
stacks are not aligned.
..Thomas
> David
> -----
>
> Kind regards, Thomas
>>
>>
More information about the hotspot-runtime-dev
mailing list