On 02/06/2010 20:35, Konstantin Kolinko wrote:
> If I search for the calls of FastHttpDateFormat.getCurrentDate()
> method I see that in trunk
> 
> a) Http11AprProcessor() and  Http11NioProcessor() constructors call
> FastHttpDateFormat.getCurrentDate() to preload the class
> 
> b) Http11Processor uses PrivilegedAction
> http://svn.apache.org/viewvc?view=revision&revision=299009
> 
> c) The proposed patch implements the third way to solve this issue:
> - preload the class inside SecurityClassLoad and
> - call a constructor, to trigger initialization of static fields.
> 
> The same could be achieved with calling any static method of the
> class, as said in ch.2.17.4 of JVM spec,
> http://java.sun.com/docs/books/jvms/second_edition/html/Concepts.doc.html#19075
> though I am fine with this implementation, as calling the constructor
> looks faster and independent of the API.
> 
> 
> I think that only one of those three approaches should be used,
> and that will be c), as it covers not only HTTP, but AJP connectors as well.

Fixed in trunk. It passes a simple test and I plan to do more extensive
testing later today.

> By the way,
> http://svn.apache.org/viewvc?rev=947717&view=rev
> contains changes to SecurityClassLoad.loadUtilPackage(). Those should
> not/cannot be backported as they are, because that method differs in
> tc6.0.x.

Those can be skipped.

Mark



---------------------------------------------------------------------
To unsubscribe, e-mail: dev-unsubscr...@tomcat.apache.org
For additional commands, e-mail: dev-h...@tomcat.apache.org

Reply via email to