Andreas Hartmann wrote:
Andreas Hartmann wrote:
OK, I narrowed down the problem. The XSPGenerator object is not re-created when the page is requested the second time. So, the XSPModuleHelper isn't re-initialized. Because its input modules were released at the end of the previous generate() execution, the inputModules map is null and the exception is thrown.
BTW, as a workaround you can add
<xsp:logic> _xsp_module_helper = null; </xsp:logic>
at the bottom of your logicsheet. This way, the module helper is initialized the second time generate() is called.
I tracked the problem down to a refactoring that took place apparently before 2.1.3, so this was broken in that release IIUC.
Basically, the XSPModuleHelper was changed so that it assumed it needed to be setup() again before each use. Probably the right assumption, but does not appear to have been assumed before and input.xsl was not changed accordingly. Because there is no sample using input.xsl, this was only noticed by people who use the input module logicsheet and attempted to use 2.1.3.
The following simple patch fixes the problem:
Index: input.xsl
===================================================================
RCS file: /home/cvs//cocoon-2.1/src/java/org/apache/cocoon/components/language/markup/xsp/java/input.xsl,v
retrieving revision 1.1
diff -u -r1.1 input.xsl
--- input.xsl 9 Mar 2003 00:08:56 -0000 1.1
+++ input.xsl 7 Feb 2004 13:29:12 -0000
@@ -79,8 +79,8 @@
// create module cache
if (this._xsp_module_helper == null) {
this._xsp_module_helper = new XSPModuleHelper();
- this._xsp_module_helper.setup(manager);
}
+ this._xsp_module_helper.setup(manager);
</xsp:init-page>
<xsp:exit-page>
I can commit this change for 2.1.4 release (thursday) but would like Carsten to confirm this is what should happen since he's the one that refactored the module helper.
Geoff
I explored another workaround, by substituting: <input:get-attribute ../> with: <util:get-sitemap-parameter .../>
Of course I have to pass parameters via sitemap to the XSP... which has taken me a while to do, but the result was cleaner (I know which parameters are used by an XSP just by looking at the sitemap); hence, I'm moderately happy with the outcome
Anyway, this is a bug... who's going to add it to bugzilla, me or you ?
I don't have time right now, but it would be good if this was officially entered in bugzilla so the record is there, and so it doesn't get lost.
Geoff
