bug report

[Dmitri Etkine]

To whom it may concern (presumably everybody on the list),   lazy evaluation mode of DOMParser seems to have a nasty bug. Here's the story:   ======== my code ======= ..... DOMParser parser = new DOMParser("file_name"); parser.setErrorHandler(myHandler); // just to see if I'm missing something parser.parse("file.xml"); Document doc = parser.getDocument(); parser.reset();            // I tried with and without - just bear with me ..... Element elem = doc.getElementById("myid"); .....   ========= my file.xml ============= <?xml version="1.0" standalone="yes" ?> <!DOCTYPE mydoc [      <!ELEMENT mydoc (myelem1*)>      <!ELEMENT myelem1 (myelem2*)>      <!ATTLIST myelem1           id    ID      #REQUIRED>      <!ELEMENT myelem2 EMPTY>      <!ATTLIST account           id    ID      #REQUIRED           name   CDATA     #REQUIRED> ]> <mydoc>      <myelem1 id="dmitri">       <myelem2 id="id1" name="name1"/>       <myelem2 id="id2" name="name2"/>       <myelem3 id="id3" name="name3"/>     </myelem1>    </mydoc> ========= what happens ============ I get an exception with following stack trace:   java.lang.NullPointerException  at java.util.Hashtable.put(Hashtable.java, Compiled Code)  at org.apache.xerces.dom.DeferredDocumentImpl.putIdentifier0(DeferredDocumentImpl.java:1576)  at org.apache.xerces.dom.DeferredDocumentImpl.getNodeObject(DeferredDocumentImpl.java, Compiled Code)  at org.apache.xerces.dom.DeferredElementImpl.synchronizeChildren(DeferredElementImpl.java, Compiled Code)  at org.apache.xerces.dom.NodeContainer.getFirstChild(NodeContainer.java:273)  at org.apache.xerces.dom.DeferredDocumentImpl.synchronizeData(DeferredDocumentImpl.java, Compiled Code)  at org.apache.xerces.dom.DocumentImpl.getIdentifier(DocumentImpl.java:853)  at org.apache.xerces.dom.DocumentImpl.getElementById(DocumentImpl.java:813)  at om.poc.OV$Entitler.entitled(OV.java, Compiled Code)  at om.poc.OV$Entitler.canWrite(OV.java:158)  at om.poc.OV.main(OV.java, Compiled Code)   ========= what I do next ============ I comment out this "parser.reset()" - it's suspicious   ========= what happens after that ========= I get this:   java.lang.ClassCastException  at org.apache.xerces.dom.DeferredDocumentImpl.synchronizeData(DeferredDocumentImpl.java, Compiled Code)  at org.apache.xerces.dom.DocumentImpl.getIdentifier(DocumentImpl.java:853)  at org.apache.xerces.dom.DocumentImpl.getElementById(DocumentImpl.java:813)  at om.poc.OV$Entitler.entitled(OV.java, Compiled Code)  at om.poc.OV$Entitler.canWrite(OV.java:158)  at om.poc.OV.main(OV.java, Compiled Code)   ========= what I do next ========= I put in following line:     parser.setFeature("http://apache.org/xml/features/dom/defer-node-expansion", false);   ======== guessed what happens? ======== It works without any exceptions or parsing errors.   ======= my environment ============== It's Win32 (NT 4.0) Xerces 1.0.3 (downloaded March 14), Sun JDK 1.2.2_001, native threads Tried it with both javac and the latest jikes compiler - no difference in behaviour, except for compilation speed   I peeked into the sources: it looks like (null) String reference is being returned from fStringPool - StringPool does indeed return nulls, without anybody checking for this possibility in DeferredDocumentImpl code.   Contact me if you need more details.   Regards, Dmitri Etkine ___________________________________________             IntegraSolv Corp. Information and Communications Engineering ___________________________________________                                1715 Hway 35, suite 201                                Middletown NJ 07748 e-mail: [EMAIL PROTECTED]         voice:  (732) 615-0054 ext.30 fax:    (732) 615-9188