Lars Huttar wrote:
Dear Geert,
Thanks for your reply.
Geert Josten wrote:
Hi Lars,
I have the impression that something is going wrong.
:-) Me too!
Have you checked the log files for errors?
Yes. There's nothing in the error.log corresponding to this problem.
The only hint there is that the count-rows.xsp is not compiled until I
call the count-rows URL
(http://localhost/mount/ethnologue/checks/count-rows?table=Development_Status)
directly... in other words, the cinclude transformer is not calling
the count-rows URL even though it is instructed to do so.
And what happens if you place the cincludes behind each other in
stead of inside each other?
In the input to the cinclude transformer (cocoon-view=raw), i.e. the
part that I can control, they are one-behind-the-other. It's only
after the cinclude transformer that they come out one-inside-the-other.
If you want us to look at your url, please supply a correct dns or ip
number and not just localhost.. ;-)
It would be nice if I could let everybody try it themselves, but our
project is on an intranet.
I gave the URL just to illustrate how the URL is formed.
Regards,
Lars
Cheers,
Geert
Lars Huttar wrote:
Dear Cocoonists,
We're getting strange results from cinclude. It's all the more
baffling because cinclude works fine in many cases.
FYI we've tried this with Tomcat 4.1, Tomcat 5.5, and Jetty, and
Cocoon 2.1.2, 2.1.5.1, and 2.1.7. All give the same results.
When I run the troubled URL,
"http://localhost:8888/mount/ethnologue/checks/list-tables";, the
browser gives an XML parsing error because the output from cinclude
is not well-formed XML: most elements have no close tags!
If I do a view-source on the output of the cinclud transform, it
looks like this:
<table xmlns:cinclude="http://apache.org/cocoon/include/1.0";>
<cinclude:include
src="cocoon:/checks/count-rows?table=Development_Status">
<cinclude:include
src="cocoon:/checks/count-rows?table=Development_Type">
...
<cinclude:include
src="cocoon:/checks/count-rows?table=Writing_System"/>
(That's really the end!)
Note that only the last element is closed.
The pipeline in the sitemap:
<map:match pattern="checks/list-tables">
<map:generate src="sources/ethnologue-schema.xml"/>
<!-- Generate cinclude statements for checks/count-rows from
<table> elements. -->
<map:transform src="transforms/list-tables-cinclude.xsl"
label="raw"/>
<!-- Process cinclude statements. -->
<map:transform type="cinclude" label="raw2"/>
<map:serialize type="xml"/>
</map:match>
The data in the pipeline before the cinclude transformer
(check/list-tables?cocoon-view=raw) looks like this:
<tables xmlns:cinclude="http://apache.org/cocoon/include/1.0";>
<cinclude:include
src="cocoon:/checks/count-rows?table=Development_Status"/>
<cinclude:include
src="cocoon:/checks/count-rows?table=Development_Type"/>
...
<cinclude:include
src="cocoon:/checks/count-rows?table=Writing_System"/>
</tables>
In other words, the output of the cinclude transformer is the same
as its input, except that each <cinclude:include> element becomes a
child of the previous one. Or to put it another way, all the
endElement events except one are dropped. There is no processing of
the src URLs.
Incidentally, if you run the src URLs on their own, e.g.
http://localhost/mount/ethnologue/checks/count-rows?table=Development_Status,
they run successfully and produce well-formed XML output:
<table [namespace declarations for xsp, xspdoc, esql, xsp-request]>
<name>Development Status</name>
<count>11875</count>
</table>
As I mentioned at the top, cinclude works fine for us in other
applications.
Any ideas?
One interesting thing is that if we follow the cinclude transformer
with an XSLT transformer that simply copies its input, that the XSLT
transformer's output is
<table xmlns:cinclude="http://apache.org/cocoon/include/1.0";>
<cinclude:include
src="cocoon:/checks/count-rows?table=Development_Status">
<cinclude:include
src="cocoon:/checks/count-rows?table=Development_Type">
...
<cinclude:include
src="cocoon:/checks/count-rows?table=Writing_System"/>
...
</cinclude:include>
</cinclude:include>
</table>
In other words, all the unclosed elements get closed. But still no
indication of why they became children of each other, nor why the
<cinclude:include> elements didn't get replaced with the output of
their respective src pipelines.
Any help would be appreciated...
Lars
More data:
The above weird behavior with cinclude seems to be tied to Java 1.5.
In the following environments, the weird behavior occurs:
======================================
- Tomcat 5.5.9 and Cocoon 2.1.7 on Win2K or WinXP with JDK 1.5 -
cinclude fails, weird behavior
- Tomcat 4.1 and Cocoon 2.1.2 on WinXP with JDK 1.5 - cinclude fails,
weird behavior
- Jetty and Cocoon 2.1.7 on WinXP with JDK 1.5 - cinclude fails, weird
behavior
- Tomcat 5.5.9 and Cocoon 2.1.7 on WinXP with JDK 1.4.2 and JRE 1.5 (JDK
1.5 removed; JVM shown in Tomcat Monitor is JRE 1.5) - cinclude fails,
weird behavior
- Jetty in cocoon-2.1.7 with JAVA_HOME=JDK 1.4.2 (JDK1.5 gone, JRE 1.5
is installed): - cinclude fails, weird behavior
In the following environment, cinclude works as expected (no weird
behavior):
=====================================================
- Tomcat 4.1 and Cocoon 2.1.7 on WinXP with JDK 1.4.2 and JRE 1.5
(JAVA_HOME is JDK 1.4.2; JDK 1.5 removed but JRE 1.5 still around) - had
to rebuild Cocoon for JDK 1.4.2, but then it works fine!
- Tomcat 4.1 and cocoon 2.1.2 on WinXP with JDK 1.4.2 and JRE 1.5
(JAVA_HOME is JDK 1.4.2; JDK 1.5 removed but JRE 1.5 still around) -
works fine.
- Tomcat 5.0.27 and Cocoon 2.1.2 on Win2K with JDK 1.4.2 - cinclude
works fine!
The exception to the above pattern is on Mac OSX, for which Java 1.5 is
not yet available:
=============================================================
- Tomcat 5.5.9 (with compatibility pack for JDK 1.4.2) and Cocoon 2.1.7
on Mac OS X, with JDK 1.4.2, JAVA_HOME is pointing to JRE - cinclude
fails with weird behavior
So... anybody know why cinclude would misbehave under Java 1.5?
If I were to run Cocoon in a Java debugger (say, in Eclipse or JDB), any
hints for finding the problem?
Again, very similar cincludes are working fine in the same environments
where the weird behavior shows up for the cincludes mentioned in my OP.
But the behavior is very reproducible.
So barring any better ideas, I'll probably try writing a working minimal
cinclude and then gradually transform it into what I want and see at
what point it starts to malfunction.
Thanks for any help.
Lars
---------------------------------------------------------------------
To unsubscribe, e-mail: [EMAIL PROTECTED]
For additional commands, e-mail: [EMAIL PROTECTED]
