Daniel-Constantin Mierla &lt;miconda(a)gmail.com&gt; hat am 7. Mai 2014 um 15:45 geschrieben: Hello, can you send the patch to be reviewed? Cheers, Daniel On 07/05/14 13:15, Klaus Feichtinger wrote: > > Hi Daniel, > > I think we´ve found the reason, why this problem occurs! > > The problem is caused in the "agregate_xmls" function in file > "notify_body.c" of the "presence_dialoginfo" module: > > /* loop over all bodies and create the aggregated body */ > for(i=0; i<j; i++) > { > /* LM_DBG("[n]=%d, [i]=%d, [j]=%d xml_array[i]=%p\n", n, i, j, > xml_array[j] ); */ > p_root= xmlDocGetRootElement(xml_array[i]); > if(p_root ==NULL) { > LM_ERR("while geting the xml_tree root element\n"); > goto error; > } > if (p_root->children) { > for (node = p_root->children; node; node = node->next) { > if (node->type == XML_ELEMENT_NODE) { > LM_DBG("node type: Element, name: %s\n", > node->name); > /* we do not copy the node, but unlink it and then > add it ot the new node > * this destroys the original document but we do > not need it anyway. > * using "copy" instead of "unlink" would also copy > the namespace which > * would then be declared redundant (libxml > unfortunately can not remove > * namespaces) > */ > if (!force_single_dialog || (j==1)) { > xmlUnlinkNode(node); > if(xmlAddChild(root_node, node)== NULL) { > LM_ERR("while adding child\n"); > goto error; > } > > It seems to be not the best idea to "unlink" the XML node (= > "xmlUnlinkNode(node);"), as then no "node->next" is available any more. > Therefore, this loop will _always_ stop after one dialog entry and ignore > any additional one! > > But we "solved" the problem in this way that the loop will not directly > use "node->next", but a variable, which is set within the loop. It looks > like this: > xmlNodePtr next_node = NULL; > [...] > if (p_root->children) { > for (node = p_root->children; node; node = next_node) { > next_node = node->next; > if (node->type == XML_ELEMENT_NODE) { > [...] > > This solution has been tested with 2 and 3 dialog entries in a single > PUBLISH request and it is working fine. We should discuss if this is a > problem that should be solved generally or if it is a "private" problem in > our use case. > > What do you mean? > > regards, > Klaus > > > >