hi all,
i'm seeing a weird behavior that probably stems from using
libxml/libxslt in a non-standard way. i am nonetheless perplexed.
the issue arises from attempting to stream top-level xmlNodes as if
the xmlParserCtxt is a pull parser.
i would like to avoid reading in a whole xml file, so i have an
xmlNode* getNextNode() function that calls xmlParseChunk() when it
needs to.
by "when it needs to", i mean it hands buffers to xmlParseChunk until
there is another top-level xmlNode *and* that node isn't the last
top-level node. said node is then xmlUnlinkNode'd, used, and possibly
operated upon and xmlFreeNode'd.
my thinking here is that if the parser hasn't finished building a node
yet, there may be some dangling references that would cause trouble.
by waiting until there is at least one more xmlNode, i hope to avoid
messing with things in progress.
this works fine when i use the node and free it.
if i use the node in a transformation, i often see xmlFree errors when
freeing the node.
here's a distilled code snippet of what i do with a node:
void item(xmlNode* node) {
if (xslt_ != NULL) {
xmlDoc* doc = xmlNewDoc((xmlChar*) "1.0");
xmlDocSetRootElement(doc, node);
xmlDoc* res = xsltApplyStylesheet(xslt_, doc, NULL);
xmlFreeDoc(doc); // free the made-up doc and input node <<---
causes breakage
xmlOut(xmlDocGetRootElement(res));
xmlFreeDoc(res); // free the result
} else {
xmlOut(node);
xmlFreeNode(node);
}
}
if i don't transform the node, all is happy. if i don't attempt to
free the node after using it in a transformation, all is happy (but my
process size grows). if i read in the whole xml file and walk through
the top-level nodes, all is happy.
ideas and suggestions welcome.
-bart
_______________________________________________
xslt mailing list, project page http://xmlsoft.org/XSLT/
[email protected]
http://mail.gnome.org/mailman/listinfo/xslt
