AbstractLogger should be doing more level checks

[GoogleCodeExporter]

What version of the product are you using? On what operating system?
Running the 2.3.1 client on RHEL. 

Tell me more...
We use the SunLogger, and have the logging level set at warning. The cases
in which debug and info don't check the logging levels before calling into
the abstract log(), end up spending time on getting a stack trace that
doesn't get used. We've seen noticeable performance improvements when the
log levels get checked first. 
Patch is attached. 

Thanks,
Paul.

Original issue reported on code.google.com by Paul.Bur...@gmail.com on 14 Apr 2010 at 3:28

Attachments:

• AbstractLogger.java.patch

[GoogleCodeExporter]

Original comment by mikewie...@gmail.com on 24 Aug 2011 at 10:43

• Changed state: Accepted
• Added labels: Milestone-Release2.7.2

[GoogleCodeExporter]

Paul,

I would like to apply you changes to Spymemcached. Can you please go to 
review.couchbase.org and sign a contributor license agreement. In order to do 
this you just need to register as a contributor to Spymemcached. From there you 
can either put you code up for review or I can do it for you and give you 
credit for the change.

Thanks,
- Mike

Original comment by mikewie...@gmail.com on 3 Sep 2011 at 12:30

[GoogleCodeExporter]

From Matt,

"on further thought, I am pretty sure the right answer here is don't do this at 
this level. We should change it in spy's default logger, but we want logger 
implementations to be somewhat in control over what gets logged, not us. We 
want to just send it up with the level we think makes sense. If they do 
something dynamic and we're doing something dynamic, there's too much 
dynamicism."

I'm going to try to change the loggin system to slf4j this weekend so that will 
probably solve your issue anyways.

Original comment by mikewie...@gmail.com on 7 Sep 2011 at 7:52

• Changed state: WontFix
• Removed labels: Milestone-Release2.7.2