svn commit: r179710 - /jakarta/commons/proper/digester/trunk/src/java/org/apache/commons/digester/

classic Classic list List threaded Threaded
1 message Options
Reply | Threaded
Open this post in threaded view

svn commit: r179710 - /jakarta/commons/proper/digester/trunk/src/java/org/apache/commons/digester/

Simon Kitching
Author: skitching
Date: Thu Jun  2 20:19:51 2005
New Revision: 179710

Added javadoc comment re flaw in CallMethodRule design.


Modified: jakarta/commons/proper/digester/trunk/src/java/org/apache/commons/digester/
--- jakarta/commons/proper/digester/trunk/src/java/org/apache/commons/digester/ (original)
+++ jakarta/commons/proper/digester/trunk/src/java/org/apache/commons/digester/ Thu Jun  2 20:19:51 2005
@@ -76,6 +76,19 @@
  * target function takes something other than a string as a parameter) then
  * the conversion will fail if the converter class does not accept an empty
  * string as valid input.</p>
+ *
+ * <p>CallMethodRule has a design flaw which can cause it to fail under
+ * certain rule configurations. All CallMethodRule instances share a single
+ * parameter stack, and all CallParamRule instances simply store their data
+ * into the parameter-info structure that is on the top of the stack. This
+ * means that two CallMethodRule instances cannot be associated with the
+ * same pattern without getting scrambled parameter data. This same issue
+ * also applies when a CallMethodRule matches some element X, a different
+ * CallMethodRule matches a child element Y and some of the CallParamRules
+ * associated with the first CallMethodRule match element Y or one of its
+ * child elements. This issue has been present since the very first release
+ * of Digester. Note, however, that this configuration of CallMethodRule
+ * instances is not commonly required.</p>
 public class CallMethodRule extends Rule {

To unsubscribe, e-mail: [hidden email]
For additional commands, e-mail: [hidden email]