logging-log4j-dev mailing list archives

Site index · List index
Message view « Date » · « Thread »
Top « Date » · « Thread »
From Anders Kristensen <akristen...@dynamicsoft.com>
Subject Re: Log4j improvement
Date Thu, 05 Jul 2001 15:00:10 GMT
One thing I don't understand is how the Java2 security model, whose
correct working depends on knowing things about stack layouts (see for
example SecurityManager.getClassContext()) deals with this problem.
Anyone know? It would seem that they have exactly the same problem and
presumably it works more than 85% of the time.

Anders


Jim Moore wrote:
> 
> FYI: At the Logging API session at JavaOne, the presenter mentioned exactly
> that issue -- the "unreliability" of the stack-trace information.  He said
> they (the JSR47 team at Sun) talked to the VM people to see if they could
> get it fixed and were told that it's not possible.  The information will be
> right "most" of the time, but not all of the time.  (I think he threw out a
> number like it being roughly 85% reliable in most environments.)  The only
> time it has a chance of being 98% reliable or better is when there are NO
> optimizations being done by either the compiler or the VM, which people only
> run under those conditions if they really like slow performance, and even
> then you can't be sure SOMETHING's not going on behind the scenes.
> 
> Just thought you'd like to know that even the people within Sun couldn't
> solve the problem -- even by trying to go in through a "back-door."
> 
> -Jim Moore
> 
> -----Original Message-----
> From: Ceki Gülcü [mailto:cgu@qos.ch]
> Sent: Tuesday, July 03, 2001 4:24 AM
> To: LOG4J Developers Mailing List
> Subject: RE: Log4j improvement
> 
> Hello Pavel,
> 
> At 12:07 03.07.2001 +0400, Pavel Muhataev wrote:
> >Hi
> >
> >> 1) It's already provided via the "locationInfo" in
> >> LoggingEvent.  If you want to see it, use your layout's means
> >> for doing so.  (For example, PatternLayout will show
> >> everything you mentioned and more.)  It's usually not used
> >> because it dramiticly slows things down, but if you've got
> >> the speed to spare and want that info, it's available for you.
> >
> >I didn't talk about discovering log caller's method. I talked about
> >automatically Category determining. In such case, user would not have to
> get
> >Category before logging.
> >Instead of Category.getInstance(_My_object_.class.getName()).debug(...)
> code
> >would be Log.debug(...).
> 
> The extraction of caller information is not reliable. We cannot build a
> system on an unreliable feature.
> 
> ---------------------------------------------------------------------
> To unsubscribe, e-mail: log4j-dev-unsubscribe@jakarta.apache.org
> For additional commands, e-mail: log4j-dev-help@jakarta.apache.org

---------------------------------------------------------------------
To unsubscribe, e-mail: log4j-dev-unsubscribe@jakarta.apache.org
For additional commands, e-mail: log4j-dev-help@jakarta.apache.org


Mime
View raw message