RFR: 8241001: Improve logging in the ConcurrentGCBreakpoint mechanism

[Stefan Johansson]

> 19 mars 2020 kl. 08:53 skrev Per Liden <per.liden at oracle.com>:
> 
> On 3/19/20 5:41 AM, Kim Barrett wrote:
>> Please review this change to the logging output produced by the
>> ConcurrentGCBreakpoint facility.
>> Most of its log messages provide low-level information for someone
>> debugging the breakpoint mechanism or a test using it, with relatively
>> cryptic messages unless that is what one is doing.  As such, putting
>> these messages at the debug level generates too much uninformative
>> clutter for the fairly common case of gc*=debug logging; those
>> messages have been demoted to the trace level.
>> There were logging messages for the concurrent cycle start
>> (idle->active) and end (active->idle) transitions, which were
>> duplicative of other logging messages from the normal concurrent cycle
>> operation.  Those logging messages have been removed.  However, the
>> message for the active->idle transition when there is an active run-to
>> request has been retained, but the message text has been improved.
>> gc+breakpoint logging has also been configured to produce the normal
>> "GC(n)" prefix, for those log messages produced by the collector.
>> User requests such as run-to don't (and can't) have that prefix, as
>> they don't occur within the context of the collection but instead come
>> from Java threads outside the collector.
>> CR:
>> https://bugs.openjdk.java.net/browse/JDK-8241001
>> Webrev:
>> https://cr.openjdk.java.net/~kbarrett/8241001/open.00/
> 
> Looks good!
Looks good to me too,
Stefan

> 
> /Per
> 
>> Testing:
>> Locally (linux-x64) ran TestConcurrentGCBreakpoints.java with various
>> logging options and verified the results were as described.