Please Review (S): CR 6782663: Data produced by PrintGCApplicationConcurrentTime and PrintGCApplicationStoppedTime is not accurate

[David Holmes - Sun Microsystems]

John,

John Coomes said the following on 07/22/09 10:57:
> Your original webrev (webrev.00) removed the print statements in
> vmThread.cpp.  Based upon David Holmes' comments (I believe) this
> version restores them, which keeps the existing (buggy) behavior of
> missing time in jdk6 when the old names are used.  I think that's
> taking compatibility too far; we should get rid of the buggy behavior
> completely.

The intent was to give JDK6 users a transition path to the new flags 
which perform a different (albeit better) measurement than the current 
flags, without simply changing the meaning of the existing flags in 
mid-stream via a JDK6 update release.

If a customer uses these flags to track performance changes across 
releases then changing the meaning of them under-the-covers is not a 
user-friendly thing to do. This might seem overly conservative but it is 
the safer option.

If we were only interested in fixing the bug with the current flags then 
we'd be tracking only GC safepoint ops, and we'd keep the old flag names.

Cheers,
David