Failing miserably at nailing a suspected bug in Excalibur xmlutils or Cocoon TraxTransformer

["Florian G. Haas" <f....@gmx.net>]

-----BEGIN PGP SIGNED MESSAGE-----
Hash: SHA1

Hello,

since I recently introduced myself in my first post to the users list, which I 
suppose many of you are reading at least occasionally, here's the brief 
version: my name is Florian, I am currently using Cocoon in order to build 
web sites from XML topic maps, among other things. 

For about three weeks now, I'm trying to dig up the reason for the behavior 
described on August 30 in my post "2.1: Neither LinkSerializer nor 
LinkGatherer producing a complete link list" to the users list.[1] Upayavira 
provided lots of help and I have a hunch that the namespace-related issues 
described in the thread really have nothing to do with the ExtendedXLinkPipe 
as both he and I originally supposed, but that its due to a bug buried 
somewhere in the Excalibur JAXP parser wrapper, the Cocoon TraxTransformer, 
or a combination of both.

To illustrate the issue, I'll take a DocBook example. I'm quite certain that 
most of you are familiar with DocBook XML 4.2, and with Norm Walsh's 
docbook-xsl stylesheets. So for brevity's sake, I'll only post a source and 
output code snippet.

Source DocBook XML (this is an excerpt of a document I've written to put on my 
personal web site):

<para>This entire process is automated using the <ulink
  url="http://cocoon.apache.org/2.1/index.html#What+is+Cocoon%3F">Cocoon XML
    publishing framework</ulink> brought to you courtesy of the <ulink
  url="http://cocoon.apache.org/">Apache Cocoon
      project</ulink>. Currently, I run Cocoon off-line (using the
  <ulink url="http://wiki.cocoondev.org/Wiki.jsp?page=CommandLine">Cocoon
    command line interface</ulink>), and upload the generated pages onto a web
    server serving static content.</para>

Here's the output when running Xalan 2.5.1 from the command line with the 
unaltered XHTML style sheet from Norm's docbook-xsl package, version 1.62:

<p>This entire process is automated using the <a 
href="http://cocoon.apache.org/2.1/index.html#What+is+Cocoon%3F" 
target="_top">Cocoon XML
    publishing framework</a> brought to you courtesy of the <a 
href="http://cocoon.apache.org/" target="_top">Apache Cocoon
      project</a>. Currently, I run Cocoon off-line (using the
  <a href="http://wiki.cocoondev.org/Wiki.jsp?page=CommandLine" 
target="_top">Cocoon
    command line interface</a>), and upload the generated pages onto a web
    server serving static content.</p>

This is just what's expected. Now, here's the output when using the same 
stylesheet in a simple Cocoon pipeline (file generator, xslt transformer 
using Xalan 2.5.1, XHTML serializer)[2]:

<p>This entire process is automated using the <a ="" target="_top">Cocoon XML
    publishing framework</a> brought to you courtesy of the <a ="" 
target="_top">Apache Cocoon
      project</a>. Currently, I run Cocoon off-line (using the
  <a ="" target="_top">Cocoon
    command line interface</a>), and upload the generated pages onto a web
    server serving static content.</p>

Something is really not quite right here. What swallows the href attributes? 
Not only their values, but also their names are empty -- strange IMHO, 
particularly because this type of behavior seems to be limited to <a> 
elements. I've run into a couple of other issues as well, e.g. xmlns 
attributes on some elements where they aren't strictly necessary, but nothing 
else as bad as this.

I've already spent hours debugging the transformation and serialization 
process, but I've been unable to nail this apparent bug. Perhaps someone with 
more TraxTransformer or Excalibur experience could look into it.

I haven't yet filed a bug on Bugzilla about this as I'm currently only 
observing symptoms and can't even confirm whether it's a real issue. If I 
should suspect it to be, I'd be grateful for a shove in the right direction 
where to look closer. Currently I must confess I'm stuck.

Best regards,
Florian


[1] thread archived at: 
http://www.mail-archive.com/users@cocoon.apache.org/msg02100.html

[2] My setup is a current Cocoon CVS checkout, J2 SDK 1.4.2_01, and Tomcat 
4.1.24 for JDK 1.4 with the required XML-related jars in the endorsed 
directory. The same files are also in jre/lib/endorsed in my JAVA_HOME.

- -- 
Florian G. Haas <f....@gmx.net>

GnuPG key ID: 0x46D00BE3
Key fingerprint: 18B4 3E7B 191E F534 254A  1F7C 816D 950B 46D0 0BE3

My GnuPG key is available from the public PGP key server at
pgp.mit.edu (and various other key servers).

-----BEGIN PGP SIGNATURE-----
Version: GnuPG v1.0.7 (GNU/Linux)

iD8DBQE/aOSBgW2VC0bQC+MRAkOCAKCu1yr7JtW7niD6W6AfzxYnE3kEXgCfW0Cu
3OUP25HMTSznmoX34idWioc=
=fVzd
-----END PGP SIGNATURE-----