logging-log4j-dev mailing list archives

Site index · List index
Message view « Date » · « Thread »
Top « Date » · « Thread »
From Elias Ross <er...@m-qube.com>
Subject thank you on bug 24159
Date Sat, 11 Feb 2006 05:07:24 GMT

Thanks for checking in my concurrent library.

There's a bunch of little unrelated changes, you found some already I
noticed.

If you have any questions about the .diff for bug 24159 let me know.

Here are some reasons for some changes:

src/java/org/apache/log4j/FileAppender.java

Setting buffered IO improves the output speed significantly for dumping
large stack traces and IMHO should be default.  There's also no reason
to turn off flushing when using buffering.

src/java/org/apache/log4j/scheduler/Scheduler.java

Setting shutdown = true should have been done in a synchronized block.
This ensures thread visibility of this flag.

src/java/org/apache/log4j/helpers/OnlyOnceErrorHandler.java

It makes sense to create a singleton instance for this class, since it's
unmodifiable, rather than recreate it for every AppenderSkeleton
instance.

src/java/org/apache/log4j/PatternLayout.java

It makes no sense to reuse a StringBuffer object on modern JVM.  This is
false economy.  These classes could be fixed to not use this pattern.

./src/java/org/apache/log4j/helpers/SyslogWriter.java
./src/java/org/apache/log4j/net/SyslogAppender.java
./src/java/org/apache/log4j/TTCCLayout.java

I suggest spending some time with a real profiling tool and a multi-
processor box.



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


Mime
View raw message