Yes. But I was thinking there might be a section heading somewhere called Location Information that specifically identifies each of the location patterns and discusses the overhead of each. This is one area where I miss the C preprocessor as the location stuff could have been hidden in a macro so it could be passed along at compile time.
Sent from my iPad On Apr 27, 2013, at 2:52 AM, Remko Popma <rem...@yahoo.com> wrote: > Ok, then to summarize my understanding for LOG4J2-219 the solution would be > to: > * internally create a root logger if one isn't explicitly configured (like in > beta5) > * this default root logger would have level ERROR (like in beta5) > * this default root logger would not have any appenders configured > (different from beta5) > * named loggers in the config are picked up correctly and not ignored > (different from beta5) > > Does that match what you had in mind? > > About location, I was thinking that the majority of sample configs would be > without location and a few would keep location layouts. For those we'd put a > comment in the config XML as well as in the accompanying text that explains > the performance impact. Does that sound ok? > > > Sent from my iPhone > > On 2013/04/27, at 15:09, Ralph Goers <ralph.go...@dslextreme.com> wrote: > >> It occurs to me that given the way Log4j and Logback work they probably >> always have a root logger that starts out not configured with an appender. I >> guess it would make sense for us to do the same. I do think it makes sense >> to emit a warning when no root logger is specified. >> >> I agree with your points about the use of location in the examples. It would >> probably make sense though to have a section that does have examples that >> clearly calls out the overhead of using them. >> >> Ralph >> >> On Apr 26, 2013, at 11:04 PM, Remko Popma wrote: >> >>> On the log4j user mailing list I encountered a user who had made a few >>> wrong turns, config-wise, and had trouble getting back on track. >>> >>> One issue was that by default the File appender is buffered but does not >>> flush. >>> So his log file stayed empty and he had trouble figuring out why. >>> >>> Should we change the examples a little? >>> * use bufferedIO=false in all File appender examples? >>> * add a comment that the root logger is mandatory (LOG4J2-219) >>> >>> Also I would like to propose (and opinions may differ on this one) >>> to avoid using layouts that use location in the examples, >>> unless the example is used to explain how location works. >>> (E.g., the API > Flow Tracing page would need location layout examples.) >>> >>> Currently most examples use %C and %M and I would greatly prefer that the >>> examples use %c and other layouts that do not include location. >>> >>> I believe that many people will start using Log4j by copying an example and >>> only read the manual when they get stuck. >>> I would hate to give first-time users the impression that Log4j is slow and >>> have our own example configs to blame... :-) >>> >>> On the same topic, in the layouts page I would like to add remarks >>> describing performance impact to the attributes that include location. >>> >>> Thoughts? >>
