[C2] and xalan2D06 and multiple CPUs ...

[Rick Tessner <ri...@dingle.myra.com>]

Hello All!

[ I think this is a Xalan or JDK implementation issue,
  but since xalan is included along with CVS Cocoon2, I thot I'd
  post this hear first as a heads-up kinda thing. ]

This is all with latest revision of Cocoon2 from CVS.

I've been banging my head today on some odd behaviour between
Cocoon2, xalan2_D06 and multiple CPU machines.

When trying out new CVS revisions of Cocoon2, I'm on
a single-CPU Solaris machine.  When I'm happy with
it, I move the identical cocoon.war file over to
a 3-CPU Solaris machine.  (Same OS, servlet engine [Resin],
CLASSPATHs, etc -- double checking that all day :)

Now, on the 3-CPU machine, I find myself frequently, but not
always, getting errors in the WEB-INF/logs/cocoon.log.  Those
messages found at bottom of this email.

At this point (after pulling out what now remains of my hair),
I start taking the CPUs offline one at a time.

 2-CPU system: same errors.
 1-CPU system: works like a hot-damn.  No problems.  Zero.  Zip. Nadda

Okay, maybe the other CPUs are flakey and just not generating errors.
Go through all combinations of single and dual CPU systems.

Single ALWAYS works.  Dual and 3-CPU consistenly generate errors.

Now, this used to work when Cocoon2 was using xalan2_D01.  So,
download versions of xalan from http://xml.apache.org/dist/xalan-j/
and start replacing the xalan JAR in WEB-INF/lib with these
other versions.

  xalan2_D01	No longer works due to extensive sitemap changes
  xalan2_D05	Works!!  No errors whatsoever with CPUs.
  xalan2_D06	Frequently generates errors (but sometimes not)
  xalan2_D07	Same as D06.

It's this that leads me to believe that it's a problem with
xalan.  (Of course, it could be that the newer xalan is
just trying to do something clever that my SUN JDK 1.3.0 just
hasn't implemented well ...)

My platform
-----------

  3-CPU SUN E450
  SUN JDK 1.3.0
  Apache 1.3.14
  Resin 1.2.1 Servlet Engine
  Cocoon2, latest from CVS (includes xalan jars)

My Resolution
-------------

  Temporary resolution, until I get time to look at the xalan 
  sources and differences between D05 and D06, is to use
  xalan2_D05.jar.

  Again, this may be a JDK issue.  I'll try it with JDK 1.2.2 
  and see how that goes.


Errors I'm seeing in WEB-INF/logs/cocoon.log
--------------------------------------------

  DEBUG   24956   [cocoon  ] (tcpConnection-6802-2): DefaultComponentManager \
   using poolable instance of org.apache.cocoon.components.parser.XercesParser 
\
   for role org.apache.cocoon.components.parser.Parser.
  DEBUG   24957   [cocoon  ] (tcpConnection-6802-2): ComponentPool retrieved
   org.apache.cocoon.components.parser.XercesParser \
   (org.apache.cocoon.components.parser.XercesParser@46572b)

  ERROR   24970   [cocoon  ] (tcpConnection-6802-2): FileGenerator.generate()
  java.lang.ArrayIndexOutOfBoundsException
        at org.apache.xerces.framework.XMLParser.parse(XMLParser.java:965)
        at org.apache.cocoon.components.parser.XercesParser.parse(XercesParser.
java:51)
        at org.apache.cocoon.generation.FileGenerator.generate(FileGenerator.ja
va:38)
        at org.apache.cocoon.sitemap.ResourcePipeline.process(ResourcePipeline.
java:201)
        at _usr._local._resin1_2_1._webapps._cocoon._sitemap_xmap.process(_site
map_xmap.java:1819)
        at org.apache.cocoon.sitemap.Handler.process(Handler.java:142)
        at org.apache.cocoon.sitemap.Manager.invoke(Manager.java:89)
        at org.apache.cocoon.Cocoon.process(Cocoon.java:248)
        at org.apache.cocoon.servlet.CocoonServlet.service(CocoonServlet.java:3
44)


-- 
Rick Tessner	rick@myra.com
MYRA Systems Corp. Fone: (250) 381 1335 x125  Phax: (250) 381 1304
                   Cell: (250) 885 9452

"Hmmm.  They have the Internet on computers now!"