Request for review (S): 7198130: G1: PrintReferenceGC output comes out of order

[Bengt Rutisson]

Hi again,

Updated webrev based on merging a similar patch from John Cuthbertson:
http://cr.openjdk.java.net/~brutisso/7198130/webrev.02/

Bengt


On 2012-09-13 10:07, Bengt Rutisson wrote:
>
> Hi all,
>
> Can I have some reviews for this change?
> http://cr.openjdk.java.net/~brutisso/7198130/webrev.01/
>
> This fixes two issues that were introduced by a previous change I made 
> to the G1 logging. Most importantly the time stamp for the GC log 
> entry is now the start of the GC. Secondly, any interleaved logging 
> that is done by other modules will be seen within the GC log entry. 
> This is the way it used to be.
>
> As an example using these switches on the command line "-XX:+UseG1GC 
> -XX:+PrintGCDetails -XX:+PrintReferenceGC" currently produce output like:
>
> 3.648: [SoftReference, 0 refs, 0.0000110 secs]3.648: [WeakReference, 
> 207 refs, 0.0000440 secs]3.648: [FinalReference, 14473 refs, 0.0499260 
> secs]3.698: [PhantomReference, 0 refs, 0.0000050 secs]3.698: [JNI Weak 
> Reference, 0.0000120 secs]3.700: [GC pause (G1 Evacuation Pause) 
> (young), 0.0853910 secs]
>    [Parallel Time: 31.9 ms, GC Workers: 8]
>
> But with my suggested fix it will be:
>
> 5.814: [GC pause (young)6.030: [SoftReference, 0 refs, 0.0000924 
> secs]6.030: [WeakReference, 3 refs, 0.0001098 secs]6.031: 
> [FinalReference, 6 refs, 0.0000924 secs]6.032: [PhantomReference, 0 
> refs, 0.0000893 secs]6.032: [JNI Weak Reference, 0 refs, 0.0001699 
> secs], 0.2200814 secs]
>    [Parallel Time: 214.3 ms, GC Workers: 4]
>
>
> This is consistent with the way it looked before my previous change to 
> the G1 logging:
>
> 3.746: [GC pause (young)3.802: [SoftReference, 0 refs, 0.0000090 
> secs]3.802: [WeakReference, 148 refs, 0.0000400 secs]3.802: 
> [FinalReference, 13020 refs, 0.0462290 secs]3.848: [PhantomReference, 
> 7 refs, 0.0000110 secs]3.848: [JNI Weak Reference, 0.0000150 secs], 
> 0.10515900 secs]
>    [Parallel Time:  52.2 ms]
>
>
> Thanks,
> Bengt