Re: opjitconv runs indefinitely

On 05/07/2015 05:34 PM, William Cohen wrote:
> On 05/07/2015 12:08 PM, Andrew Haley wrote:
>> On 05/07/2015 04:04 PM, Daniel Hansel wrote:
>>> in that case these zero sized samples must not be invalidated without any further investigation.
>>>
>>> Could you describe such a case where zero-sized samples will be generated?
>>> Maybe I could figure out how such samples could be used to produce a correct profiling report.
>>
>> Surely Oprofile generates the samples.  I have no idea how samples
>> could be generated against a zero-length region.  Zero-length
>> regions are perfectly legitimate.
> 
> I think the question was in reference to earlier email's:
> 
>   This is a particular PPC thing which generates zero-sized stubs in some cases.
> 
> There is a thread discussing this problem on the ppc-aix-port-dev mailing list:
> 
> http://mail.openjdk.java.net/pipermail/ppc-aix-port-dev/2014-July/002045.html
> 
> There seems to be some question whether jvmti should be generating that:
> 
> http://mail.openjdk.java.net/pipermail/ppc-aix-port-dev/2014-July/002048.html
> https://bugs.openjdk.java.net/browse/JDK-8049441
> 
> It looks like there is a patch to remove the zero-length regions.
> 
> However, oprofile should be more graceful in how it handles this.

Of course it should.  Zero is not a special number: it's a very ordinary
number.  Some code regions are zero in size.  This is not a Java thing; it
could happen anywhere.  I can't see any point oprofile resisting this bug
fix, if only for robustness.

Andrew.



------------------------------------------------------------------------------
One dashboard for servers and applications across Physical-Virtual-Cloud 
Widest out-of-the-box monitoring support with 50+ applications
Performance metrics, stats and reports that give you Actionable Insights
Deep dive visibility with transaction tracing using APM Insight.
http://ad.doubleclick.net/ddm/clk/290420510;117567292;y