RFR (S): 7176220 'Full GC' events miss date stamp information occasionally

[Bengt Rutisson]

Andreas,

On 5/14/13 11:21 AM, Andreas Eriksson wrote:
> Hi,
>
> I changed the date and time stamping calls to use the guarded 
> convenience variants.
> Both of these calls suffixes a ": " automatically. See 
> share/vm/utilities/ostream.hpp line 99 and 105.

Right. Sorry  I missed that!

The change looks good.

Bengt

>
> It currently looks like this:
> % ./hsx24_hotspot-build/bin/java -Xmx2m -verbose:gc 
> -XX:+PrintGCTimeStamps -XX:+PrintGCDateStamps HumongousAlloc
> 2013-05-14T11:17:28.340+0200: 0.050: [GC 1896K->421K(3136K), 0.0011640 
> secs]
> 2013-05-14T11:17:28.341+0200: 0.051: [GC 2329K->437K(3136K), 0.0029750 
> secs]
>
> This is how it looked before as well:
> % ~/builds/linux-amd64/7u25/bin/java -Xmx2m -verbose:gc 
> -XX:+PrintGCTimeStamps -XX:+PrintGCDateStamps HumongousAlloc
> 2013-05-14T11:17:31.603+0200: 0.059: [GC 2025K->417K(3136K), 0.0008100 
> secs]
> 2013-05-14T11:17:31.605+0200: 0.060: [GC 2325K->417K(3136K), 0.0015140 
> secs]
>
> Thanks for looking at this.
>
> /Andreas
>
> On 2013-05-14 08:23, Bengt Rutisson wrote:
>>
>> Hi Andreas,
>>
>> Looks good.
>>
>> One question regarding the change in GCTraceTime:
>>
>> http://cr.openjdk.java.net/~aeriksso/7176220/webrev.00/src/share/vm/gc_implementation/shared/gcTraceTime.cpp.udiff.html
>>
>> It looks to me like we are missing a ": " in the new version. Maybe 
>> it is ok to loose this character in the log format, but a safer 
>> change may be:
>>
>> if (PrintGCTimeStamps || PrintGCDateStamps) {
>>   gclog_or_tty->date_stamp(PrintGCDateStamps);
>>   gclog_or_tty->stamp(PrintGCTimeStamps);
>> gclog_or_tty->print(": ");
>> }
>>
>> This actually still changes the ": " behavior since since you didn't 
>> use to get a ": " if you only had PrintGCDateStamps enabled. But I 
>> think this is a more expected behavior.
>>
>> Thanks,
>> Bengt
>>
>> On 5/13/13 6:01 PM, Andreas Eriksson wrote:
>>> Hi all,
>>>
>>> Please take a look at my changes for bug 7176220.
>>> The problem is that CMS sometimes misses datestamps for time trace 
>>> printouts.
>>>
>>> The proposed fix is to move the datestamp print logic into the class 
>>> that handles time trace printouts, GCTraceTime.
>>> This would make it consistent with timestamps, which are already 
>>> printed from there.
>>>
>>> Bugs.sun:
>>> http://bugs.sun.com/view_bug.do?bug_id=7176220
>>>
>>> JIRA:
>>> https://jbs.oracle.com/bugs/browse/JDK-7176220
>>>
>>> Webrev:
>>> http://cr.openjdk.java.net/~aeriksso/7176220/webrev.00/
>>>
>>> Testing:
>>> jprt, jtreg
>>>
>>> Thanks,
>>> Andreas
>>>
>>
>

-------------- next part --------------
An HTML attachment was scrubbed...
URL: <https://mail.openjdk.org/pipermail/hotspot-gc-dev/attachments/20130514/c04c7699/attachment.htm>