[XERCESJ-1749] XInclude parsing fails on legal construct

• Type: Bug
• Status: Open
• Priority: Critical
• Resolution: Unresolved
• Affects Version/s: 2.12.2
• Component/s: XInclude 1.0
• Labels: XInclude

Environment

jlpoole@ares /usr/local/src/xerces_2.12.2/xerces-2_12_2 $ java -version
openjdk version "1.8.0_332"
OpenJDK Runtime Environment (Temurin)(build 1.8.0_332-b09)
OpenJDK 64-Bit Server VM (Temurin)(build 25.332-b09, mixed mode)
jlpoole@ares /usr/local/src/xerces_2.12.2/xerces-2_12_2 $

I downloaded Xerces-J-src.2.12.2.tar.gz and Xerces-J-tools.2.12.2.tar.gz and built same, then tested with the example file above.

Attaching examples_log.txt, dated Wednesday Aug 17, 2022 11:35:34 AM, of runs for both files also showing xmllint output successfully resolving the XIncludes.

• Flags: Important

Language

java

• Assignee: Unassigned
• Reporter: John Poole
• Created: 17/Aug/22 18:38
• Updated: 05/Sep/22 09:01

Description

Suppose a3.xml:

<book xmlns="http://docbook.org/ns/docbook"
    xmlns:xlink="http://www.w3.org/1999/xlink"
    xmlns:xi="http://www.w3.org/2001/XInclude"
    xml:id="a-book"
    version="5.0">
    <title>A</title>
    <chapter xml:id="a-chapter">
        <title>a chapter</title>
        <BurstingPara xml:id="t1000">Warning: Read following paragraph carefully.</BurstingPara>
        <para>foo</para>
        <xi:include href="a3.xml" xpointer="t1000"/>
        <para>foo2</para>
        <xi:include href="a3.xml" xpointer="t1000"/>
        <para>foo3</para>
    </chapter>
   <xi:include href="a3.xml" xpointer="a-chapter"/>
</book>

Xerces 2.12.2 rejects the XML with:

edit: added CLASSPATH recital below to provide context for following session:

jlpoole@ares /usr/local/src/xerces_2.12.2/xerces-2_12_2 $ date;echo $CLASSPATH
Wed Aug 17 12:19:41 PDT 2022
.:build/xercesSamples.jar:build/xercesImpl.jar:build/serializer.jar:build/resolver.jar
jlpoole@ares /usr/local/src/xerces_2.12.2/xerces-2_12_2 $

jlpoole@ares /usr/local/src/xerces_2.12.2/xerces-2_12_2 $ java dom.Writer -xi a3.xml
[Fatal Error] a3.xml:1:1: Recursive include detected.  Document 'file:///usr/local/src/xerces_2.12.2/xerces-2_12_2/a3.xml' was already processed.
[Fatal Error] a3.xml:11:53: Error attempting to parse XML file (href='a3.xml').
jlpoole@ares /usr/local/src/xerces_2.12.2/xerces-2_12_2 $ {noformat}
The XML XIncludes above are a legal construct, yet Xerces throws an error when the parser should process the XIncludes.

This bug mimics the bug encountered in the Gnome XMLParser2 project:[ xmllint handling of recursion|[https://gitlab.gnome.org/GNOME/libxml2/-/issues/348]]

Here is another sample file illustrating the failure:
{code:java}
<book xmlns="http://docbook.org/ns/docbook"
    xmlns:xlink="http://www.w3.org/1999/xlink"
    xmlns:xi="http://www.w3.org/2001/XInclude"
    xml:id="b-book"
    version="5.0">
    <title>B</title>
    <chapter xml:id="b-chapter">
        <title>b chapter</title>
        <para>foo</para>
    </chapter>
    <xi:include href="a3.xml" xpointer="a-chapter"/>
</book>