RE: [xml] Differences in DOM parsing on Windows and UNIX?

["Stéphane Bidoul" <stephane bidoul softwareag com>]

> Well, there is not that much to it.  When the code was as follows:
>
>          xmlInitParser();
>          xmlKeepBlanksDefault(0);
>          xmlLineNumbersDefault(1);  // set line numbers
>
>          pXmlDoc = xmlParseFile (fileSpec);
>
>          if (0 == pXmlDoc) {
>             printf ("Error parsing XML file\n");
>             return 1;
>          }
>
>          xmlCleanupParser();
>
> I got the blank nodes.  When I changed it to:
>
>          xmlInitParser();
> #ifdef LIBXML_THREAD_ENABLED
>          xmlThrDefKeepBlanksDefaultValue(0);
> #else
>          xmlKeepBlanksDefault(0);
> #endif
>          xmlLineNumbersDefault(1);  // set line numbers
>
>          pXmlDoc = xmlParseFile (fileSpec);
>
>          if (0 == pXmlDoc) {
>             printf ("Error parsing XML file\n");
>             return 1;
>          }
>
>          xmlCleanupParser();
>
> They went away.

I really don't see why it would be different for xmlKeepBlanksDefault
and xmlLineNumbersDefault: all globals are treated equally
in libxml (I'm pretty sure, because that code is actually generated). 
You should check again, IMHO.

> It really should not be necessary for me to have to deal with 
> a #define like
> LIBXML_THREAD_ENABLED in my mainline code.

If your program has only one thread, you don't need to. 
Just have a look at xmllint.c: it only
does xmlKeepBlanksDefault(0) and that does not
depend on LIBXML_THREAD_ENABLED.

Assume test.xml:

<!DOCTYPE a [
<!ELEMENT a (b)>
<!ELEMENT b EMPTY>
]>
<a> 
    <b/>    
</a>

~>xmllint --noblanks test.xml
<?xml version="1.0"?>
<!DOCTYPE a [
<!ELEMENT a (b)>
<!ELEMENT b EMPTY>
]>
<a><b/></a>

If your program has several threads, either
do xmlThrDefKeepBlanksDefaultValue() before creating
your threads or do xmlKeepBlanksDefault() in each thread.

-sbi