[xml] Patch suggestion for "fixing" 10 MB limit when using xmlNewTextWriterDoc

[Stian Hvatum <stian hvatum net>]

Hi,
I am one of those who have been bit by the 10 MB limit when building an 
XML using xmlNewTextWriterDoc as constructor the xmlTextWriter.

In order to work around the problem I have forked libxml2 and added two 
functions to xmlwriter.c that enables the user to get and set the parser 
options used in the underlying parser. My reason for doing so was that 
it seemed to me as the most non-destructive way of passing 
XML_PARSER_HUGE to the underlying parser context.

The functions that I have added are xmlTextWriterSetParserOptions and 
xmlTextWriterGetParserOption (to get options set other places if 
necessary. The code is available on GitHub on 
https://github.com/hvatum/libxml2 and I would be very happy to create a 
pull request if you think that this is a viable solution.

I struggled a bit with the double free that occurs when dictNames == 1 
as commented on xmlwriter.c:447. This is currently handled by resetting 
dictNames after setting the options, but I believe that the real problem 
is that the doc->dict and ctxt->dict points to the same structure when 
dictNames == 1, and later when you free the writer and the doc, the dict 
is free'd twice. See SAX2.c:1029.



/*
 * Example code for creating xmlTextWriter capable of creating larger 
than 10 MB files.
 *
 * It might be unnecessary to get the options first, as they seems to 
be 0 by default.
 *
 */

    xmlDocPtr doc;
    xmlTextWriterPtr writer = xmlNewTextWriterDoc(&doc, 0);
    if (writer == NULL) {
        puts("Error while creating the XML writer");
    } else {
        int options = xmlTextWriterGetParserOptions(writer);
        xmlTextWriterSetParserOptions(writer, XML_PARSE_HUGE | options);
    }



Best regards,
Stian Hvatum