diff options
author | William Joye <wjoye@cfa.harvard.edu> | 2016-10-19 18:28:08 (GMT) |
---|---|---|
committer | William Joye <wjoye@cfa.harvard.edu> | 2016-10-19 18:28:08 (GMT) |
commit | 7f402e480a04b974617c04142993d3919457d342 (patch) | |
tree | a39cf901d49b9767df3fe076ebf8a06c92277995 /libxml2/c14n.c | |
parent | 818fd731f62993dc8aa5d07b82e1d1b2c04aae7d (diff) | |
parent | 6c75f0dc676d73c4cbcdca5a9cf05c6797a2ac0b (diff) | |
download | blt-7f402e480a04b974617c04142993d3919457d342.zip blt-7f402e480a04b974617c04142993d3919457d342.tar.gz blt-7f402e480a04b974617c04142993d3919457d342.tar.bz2 |
Merge commit '6c75f0dc676d73c4cbcdca5a9cf05c6797a2ac0b' as 'libxml2'
Diffstat (limited to 'libxml2/c14n.c')
-rw-r--r-- | libxml2/c14n.c | 2238 |
1 files changed, 2238 insertions, 0 deletions
diff --git a/libxml2/c14n.c b/libxml2/c14n.c new file mode 100644 index 0000000..ca77f92 --- /dev/null +++ b/libxml2/c14n.c @@ -0,0 +1,2238 @@ +/* + * "Canonical XML" implementation + * http://www.w3.org/TR/xml-c14n + * + * "Exclusive XML Canonicalization" implementation + * http://www.w3.org/TR/xml-exc-c14n + * + * See Copyright for the status of this software. + * + * Author: Aleksey Sanin <aleksey@aleksey.com> + */ +#define IN_LIBXML +#include "libxml.h" +#ifdef LIBXML_C14N_ENABLED +#ifdef LIBXML_OUTPUT_ENABLED + +#ifdef HAVE_STDLIB_H +#include <stdlib.h> +#endif +#include <string.h> + +#include <libxml/tree.h> +#include <libxml/parser.h> +#include <libxml/uri.h> +#include <libxml/xmlerror.h> +#include <libxml/globals.h> +#include <libxml/xpathInternals.h> +#include <libxml/c14n.h> + +#include "buf.h" + +/************************************************************************ + * * + * Some declaration better left private ATM * + * * + ************************************************************************/ + +typedef enum { + XMLC14N_BEFORE_DOCUMENT_ELEMENT = 0, + XMLC14N_INSIDE_DOCUMENT_ELEMENT = 1, + XMLC14N_AFTER_DOCUMENT_ELEMENT = 2 +} xmlC14NPosition; + +typedef struct _xmlC14NVisibleNsStack { + int nsCurEnd; /* number of nodes in the set */ + int nsPrevStart; /* the begginning of the stack for previous visible node */ + int nsPrevEnd; /* the end of the stack for previous visible node */ + int nsMax; /* size of the array as allocated */ + xmlNsPtr *nsTab; /* array of ns in no particular order */ + xmlNodePtr *nodeTab; /* array of nodes in no particular order */ +} xmlC14NVisibleNsStack, *xmlC14NVisibleNsStackPtr; + +typedef struct _xmlC14NCtx { + /* input parameters */ + xmlDocPtr doc; + xmlC14NIsVisibleCallback is_visible_callback; + void* user_data; + int with_comments; + xmlOutputBufferPtr buf; + + /* position in the XML document */ + xmlC14NPosition pos; + int parent_is_doc; + xmlC14NVisibleNsStackPtr ns_rendered; + + /* C14N mode */ + xmlC14NMode mode; + + /* exclusive canonicalization */ + xmlChar **inclusive_ns_prefixes; + + /* error number */ + int error; +} xmlC14NCtx, *xmlC14NCtxPtr; + +static xmlC14NVisibleNsStackPtr xmlC14NVisibleNsStackCreate (void); +static void xmlC14NVisibleNsStackDestroy (xmlC14NVisibleNsStackPtr cur); +static void xmlC14NVisibleNsStackAdd (xmlC14NVisibleNsStackPtr cur, + xmlNsPtr ns, + xmlNodePtr node); +static void xmlC14NVisibleNsStackSave (xmlC14NVisibleNsStackPtr cur, + xmlC14NVisibleNsStackPtr state); +static void xmlC14NVisibleNsStackRestore (xmlC14NVisibleNsStackPtr cur, + xmlC14NVisibleNsStackPtr state); +static void xmlC14NVisibleNsStackShift (xmlC14NVisibleNsStackPtr cur); +static int xmlC14NVisibleNsStackFind (xmlC14NVisibleNsStackPtr cur, + xmlNsPtr ns); +static int xmlExcC14NVisibleNsStackFind (xmlC14NVisibleNsStackPtr cur, + xmlNsPtr ns, + xmlC14NCtxPtr ctx); + +static int xmlC14NIsNodeInNodeset (xmlNodeSetPtr nodes, + xmlNodePtr node, + xmlNodePtr parent); + + + +static int xmlC14NProcessNode(xmlC14NCtxPtr ctx, xmlNodePtr cur); +static int xmlC14NProcessNodeList(xmlC14NCtxPtr ctx, xmlNodePtr cur); +typedef enum { + XMLC14N_NORMALIZE_ATTR = 0, + XMLC14N_NORMALIZE_COMMENT = 1, + XMLC14N_NORMALIZE_PI = 2, + XMLC14N_NORMALIZE_TEXT = 3 +} xmlC14NNormalizationMode; + +static xmlChar *xmlC11NNormalizeString(const xmlChar * input, + xmlC14NNormalizationMode mode); + +#define xmlC11NNormalizeAttr( a ) \ + xmlC11NNormalizeString((a), XMLC14N_NORMALIZE_ATTR) +#define xmlC11NNormalizeComment( a ) \ + xmlC11NNormalizeString((a), XMLC14N_NORMALIZE_COMMENT) +#define xmlC11NNormalizePI( a ) \ + xmlC11NNormalizeString((a), XMLC14N_NORMALIZE_PI) +#define xmlC11NNormalizeText( a ) \ + xmlC11NNormalizeString((a), XMLC14N_NORMALIZE_TEXT) + +#define xmlC14NIsVisible( ctx, node, parent ) \ + (((ctx)->is_visible_callback != NULL) ? \ + (ctx)->is_visible_callback((ctx)->user_data, \ + (xmlNodePtr)(node), (xmlNodePtr)(parent)) : 1) + +#define xmlC14NIsExclusive( ctx ) \ + ( (ctx)->mode == XML_C14N_EXCLUSIVE_1_0 ) + +/************************************************************************ + * * + * Some factorized error routines * + * * + ************************************************************************/ + +/** + * xmlC14NErrMemory: + * @extra: extra informations + * + * Handle a redefinition of memory error + */ +static void +xmlC14NErrMemory(const char *extra) +{ + __xmlRaiseError(NULL, NULL, NULL, NULL, NULL, XML_FROM_C14N, + XML_ERR_NO_MEMORY, XML_ERR_ERROR, NULL, 0, extra, + NULL, NULL, 0, 0, + "Memory allocation failed : %s\n", extra); +} + +/** + * xmlC14NErrParam: + * @extra: extra informations + * + * Handle a redefinition of param error + */ +static void +xmlC14NErrParam(const char *extra) +{ + __xmlRaiseError(NULL, NULL, NULL, NULL, NULL, XML_FROM_C14N, + XML_ERR_INTERNAL_ERROR, XML_ERR_ERROR, NULL, 0, extra, + NULL, NULL, 0, 0, + "Invalid parameter : %s\n", extra); +} + +/** + * xmlC14NErrInternal: + * @extra: extra informations + * + * Handle a redefinition of internal error + */ +static void +xmlC14NErrInternal(const char *extra) +{ + __xmlRaiseError(NULL, NULL, NULL, NULL, NULL, XML_FROM_C14N, + XML_ERR_INTERNAL_ERROR, XML_ERR_ERROR, NULL, 0, extra, + NULL, NULL, 0, 0, + "Internal error : %s\n", extra); +} + +/** + * xmlC14NErrInvalidNode: + * @extra: extra informations + * + * Handle a redefinition of invalid node error + */ +static void +xmlC14NErrInvalidNode(const char *node_type, const char *extra) +{ + __xmlRaiseError(NULL, NULL, NULL, NULL, NULL, XML_FROM_C14N, + XML_C14N_INVALID_NODE, XML_ERR_ERROR, NULL, 0, extra, + NULL, NULL, 0, 0, + "Node %s is invalid here : %s\n", node_type, extra); +} + +/** + * xmlC14NErrUnknownNode: + * @extra: extra informations + * + * Handle a redefinition of unknown node error + */ +static void +xmlC14NErrUnknownNode(int node_type, const char *extra) +{ + __xmlRaiseError(NULL, NULL, NULL, NULL, NULL, XML_FROM_C14N, + XML_C14N_UNKNOW_NODE, XML_ERR_ERROR, NULL, 0, extra, + NULL, NULL, 0, 0, + "Unknown node type %d found : %s\n", node_type, extra); +} + +/** + * xmlC14NErrRelativeNamespace: + * @extra: extra informations + * + * Handle a redefinition of relative namespace error + */ +static void +xmlC14NErrRelativeNamespace(const char *ns_uri) +{ + __xmlRaiseError(NULL, NULL, NULL, NULL, NULL, XML_FROM_C14N, + XML_C14N_RELATIVE_NAMESPACE, XML_ERR_ERROR, NULL, 0, NULL, + NULL, NULL, 0, 0, + "Relative namespace UR is invalid here : %s\n", ns_uri); +} + + + +/** + * xmlC14NErr: + * @ctxt: a C14N evaluation context + * @node: the context node + * @error: the erorr code + * @msg: the message + * @extra: extra informations + * + * Handle a redefinition of attribute error + */ +static void +xmlC14NErr(xmlC14NCtxPtr ctxt, xmlNodePtr node, int error, + const char * msg) +{ + if (ctxt != NULL) + ctxt->error = error; + __xmlRaiseError(NULL, NULL, NULL, + ctxt, node, XML_FROM_C14N, error, + XML_ERR_ERROR, NULL, 0, + NULL, NULL, NULL, 0, 0, "%s", msg); +} + +/************************************************************************ + * * + * The implementation internals * + * * + ************************************************************************/ +#define XML_NAMESPACES_DEFAULT 16 + +static int +xmlC14NIsNodeInNodeset(xmlNodeSetPtr nodes, xmlNodePtr node, xmlNodePtr parent) { + if((nodes != NULL) && (node != NULL)) { + if(node->type != XML_NAMESPACE_DECL) { + return(xmlXPathNodeSetContains(nodes, node)); + } else { + xmlNs ns; + + memcpy(&ns, node, sizeof(ns)); + + /* this is a libxml hack! check xpath.c for details */ + if((parent != NULL) && (parent->type == XML_ATTRIBUTE_NODE)) { + ns.next = (xmlNsPtr)parent->parent; + } else { + ns.next = (xmlNsPtr)parent; + } + + /* + * If the input is an XPath node-set, then the node-set must explicitly + * contain every node to be rendered to the canonical form. + */ + return(xmlXPathNodeSetContains(nodes, (xmlNodePtr)&ns)); + } + } + return(1); +} + +static xmlC14NVisibleNsStackPtr +xmlC14NVisibleNsStackCreate(void) { + xmlC14NVisibleNsStackPtr ret; + + ret = (xmlC14NVisibleNsStackPtr) xmlMalloc(sizeof(xmlC14NVisibleNsStack)); + if (ret == NULL) { + xmlC14NErrMemory("creating namespaces stack"); + return(NULL); + } + memset(ret, 0 , (size_t) sizeof(xmlC14NVisibleNsStack)); + return(ret); +} + +static void +xmlC14NVisibleNsStackDestroy(xmlC14NVisibleNsStackPtr cur) { + if(cur == NULL) { + xmlC14NErrParam("destroying namespaces stack"); + return; + } + if(cur->nsTab != NULL) { + memset(cur->nsTab, 0, cur->nsMax * sizeof(xmlNsPtr)); + xmlFree(cur->nsTab); + } + if(cur->nodeTab != NULL) { + memset(cur->nodeTab, 0, cur->nsMax * sizeof(xmlNodePtr)); + xmlFree(cur->nodeTab); + } + memset(cur, 0, sizeof(xmlC14NVisibleNsStack)); + xmlFree(cur); + +} + +static void +xmlC14NVisibleNsStackAdd(xmlC14NVisibleNsStackPtr cur, xmlNsPtr ns, xmlNodePtr node) { + if((cur == NULL) || + ((cur->nsTab == NULL) && (cur->nodeTab != NULL)) || + ((cur->nsTab != NULL) && (cur->nodeTab == NULL))) { + xmlC14NErrParam("adding namespace to stack"); + return; + } + + if ((cur->nsTab == NULL) && (cur->nodeTab == NULL)) { + cur->nsTab = (xmlNsPtr*) xmlMalloc(XML_NAMESPACES_DEFAULT * sizeof(xmlNsPtr)); + cur->nodeTab = (xmlNodePtr*) xmlMalloc(XML_NAMESPACES_DEFAULT * sizeof(xmlNodePtr)); + if ((cur->nsTab == NULL) || (cur->nodeTab == NULL)) { + xmlC14NErrMemory("adding node to stack"); + return; + } + memset(cur->nsTab, 0 , XML_NAMESPACES_DEFAULT * sizeof(xmlNsPtr)); + memset(cur->nodeTab, 0 , XML_NAMESPACES_DEFAULT * sizeof(xmlNodePtr)); + cur->nsMax = XML_NAMESPACES_DEFAULT; + } else if(cur->nsMax == cur->nsCurEnd) { + void *tmp; + int tmpSize; + + tmpSize = 2 * cur->nsMax; + tmp = xmlRealloc(cur->nsTab, tmpSize * sizeof(xmlNsPtr)); + if (tmp == NULL) { + xmlC14NErrMemory("adding node to stack"); + return; + } + cur->nsTab = (xmlNsPtr*)tmp; + + tmp = xmlRealloc(cur->nodeTab, tmpSize * sizeof(xmlNodePtr)); + if (tmp == NULL) { + xmlC14NErrMemory("adding node to stack"); + return; + } + cur->nodeTab = (xmlNodePtr*)tmp; + + cur->nsMax = tmpSize; + } + cur->nsTab[cur->nsCurEnd] = ns; + cur->nodeTab[cur->nsCurEnd] = node; + + ++cur->nsCurEnd; +} + +static void +xmlC14NVisibleNsStackSave(xmlC14NVisibleNsStackPtr cur, xmlC14NVisibleNsStackPtr state) { + if((cur == NULL) || (state == NULL)) { + xmlC14NErrParam("saving namespaces stack"); + return; + } + + state->nsCurEnd = cur->nsCurEnd; + state->nsPrevStart = cur->nsPrevStart; + state->nsPrevEnd = cur->nsPrevEnd; +} + +static void +xmlC14NVisibleNsStackRestore(xmlC14NVisibleNsStackPtr cur, xmlC14NVisibleNsStackPtr state) { + if((cur == NULL) || (state == NULL)) { + xmlC14NErrParam("restoring namespaces stack"); + return; + } + cur->nsCurEnd = state->nsCurEnd; + cur->nsPrevStart = state->nsPrevStart; + cur->nsPrevEnd = state->nsPrevEnd; +} + +static void +xmlC14NVisibleNsStackShift(xmlC14NVisibleNsStackPtr cur) { + if(cur == NULL) { + xmlC14NErrParam("shifting namespaces stack"); + return; + } + cur->nsPrevStart = cur->nsPrevEnd; + cur->nsPrevEnd = cur->nsCurEnd; +} + +static int +xmlC14NStrEqual(const xmlChar *str1, const xmlChar *str2) { + if (str1 == str2) return(1); + if (str1 == NULL) return((*str2) == '\0'); + if (str2 == NULL) return((*str1) == '\0'); + do { + if (*str1++ != *str2) return(0); + } while (*str2++); + return(1); +} + +/** + * xmlC14NVisibleNsStackFind: + * @ctx: the C14N context + * @ns: the namespace to check + * + * Checks whether the given namespace was already rendered or not + * + * Returns 1 if we already wrote this namespace or 0 otherwise + */ +static int +xmlC14NVisibleNsStackFind(xmlC14NVisibleNsStackPtr cur, xmlNsPtr ns) +{ + int i; + const xmlChar *prefix; + const xmlChar *href; + int has_empty_ns; + + if(cur == NULL) { + xmlC14NErrParam("searching namespaces stack (c14n)"); + return (0); + } + + /* + * if the default namespace xmlns="" is not defined yet then + * we do not want to print it out + */ + prefix = ((ns == NULL) || (ns->prefix == NULL)) ? BAD_CAST "" : ns->prefix; + href = ((ns == NULL) || (ns->href == NULL)) ? BAD_CAST "" : ns->href; + has_empty_ns = (xmlC14NStrEqual(prefix, NULL) && xmlC14NStrEqual(href, NULL)); + + if (cur->nsTab != NULL) { + int start = (has_empty_ns) ? 0 : cur->nsPrevStart; + for (i = cur->nsCurEnd - 1; i >= start; --i) { + xmlNsPtr ns1 = cur->nsTab[i]; + + if(xmlC14NStrEqual(prefix, (ns1 != NULL) ? ns1->prefix : NULL)) { + return(xmlC14NStrEqual(href, (ns1 != NULL) ? ns1->href : NULL)); + } + } + } + return(has_empty_ns); +} + +static int +xmlExcC14NVisibleNsStackFind(xmlC14NVisibleNsStackPtr cur, xmlNsPtr ns, xmlC14NCtxPtr ctx) { + int i; + const xmlChar *prefix; + const xmlChar *href; + int has_empty_ns; + + if(cur == NULL) { + xmlC14NErrParam("searching namespaces stack (exc c14n)"); + return (0); + } + + /* + * if the default namespace xmlns="" is not defined yet then + * we do not want to print it out + */ + prefix = ((ns == NULL) || (ns->prefix == NULL)) ? BAD_CAST "" : ns->prefix; + href = ((ns == NULL) || (ns->href == NULL)) ? BAD_CAST "" : ns->href; + has_empty_ns = (xmlC14NStrEqual(prefix, NULL) && xmlC14NStrEqual(href, NULL)); + + if (cur->nsTab != NULL) { + int start = 0; + for (i = cur->nsCurEnd - 1; i >= start; --i) { + xmlNsPtr ns1 = cur->nsTab[i]; + + if(xmlC14NStrEqual(prefix, (ns1 != NULL) ? ns1->prefix : NULL)) { + if(xmlC14NStrEqual(href, (ns1 != NULL) ? ns1->href : NULL)) { + return(xmlC14NIsVisible(ctx, ns1, cur->nodeTab[i])); + } else { + return(0); + } + } + } + } + return(has_empty_ns); +} + + + + +/** + * xmlC14NIsXmlNs: + * @ns: the namespace to check + * + * Checks whether the given namespace is a default "xml:" namespace + * with href="http://www.w3.org/XML/1998/namespace" + * + * Returns 1 if the node is default or 0 otherwise + */ + +/* todo: make it a define? */ +static int +xmlC14NIsXmlNs(xmlNsPtr ns) +{ + return ((ns != NULL) && + (xmlStrEqual(ns->prefix, BAD_CAST "xml")) && + (xmlStrEqual(ns->href, XML_XML_NAMESPACE))); +} + + +/** + * xmlC14NNsCompare: + * @ns1: the pointer to first namespace + * @ns2: the pointer to second namespace + * + * Compares the namespaces by names (prefixes). + * + * Returns -1 if ns1 < ns2, 0 if ns1 == ns2 or 1 if ns1 > ns2. + */ +static int +xmlC14NNsCompare(xmlNsPtr ns1, xmlNsPtr ns2) +{ + if (ns1 == ns2) + return (0); + if (ns1 == NULL) + return (-1); + if (ns2 == NULL) + return (1); + + return (xmlStrcmp(ns1->prefix, ns2->prefix)); +} + + +/** + * xmlC14NPrintNamespaces: + * @ns: the pointer to namespace + * @ctx: the C14N context + * + * Prints the given namespace to the output buffer from C14N context. + * + * Returns 1 on success or 0 on fail. + */ +static int +xmlC14NPrintNamespaces(const xmlNsPtr ns, xmlC14NCtxPtr ctx) +{ + + if ((ns == NULL) || (ctx == NULL)) { + xmlC14NErrParam("writing namespaces"); + return 0; + } + + if (ns->prefix != NULL) { + xmlOutputBufferWriteString(ctx->buf, " xmlns:"); + xmlOutputBufferWriteString(ctx->buf, (const char *) ns->prefix); + xmlOutputBufferWriteString(ctx->buf, "="); + } else { + xmlOutputBufferWriteString(ctx->buf, " xmlns="); + } + if(ns->href != NULL) { + xmlBufWriteQuotedString(ctx->buf->buffer, ns->href); + } else { + xmlOutputBufferWriteString(ctx->buf, "\"\""); + } + return (1); +} + +/** + * xmlC14NProcessNamespacesAxis: + * @ctx: the C14N context + * @node: the current node + * + * Prints out canonical namespace axis of the current node to the + * buffer from C14N context as follows + * + * Canonical XML v 1.0 (http://www.w3.org/TR/xml-c14n) + * + * Namespace Axis + * Consider a list L containing only namespace nodes in the + * axis and in the node-set in lexicographic order (ascending). To begin + * processing L, if the first node is not the default namespace node (a node + * with no namespace URI and no local name), then generate a space followed + * by xmlns="" if and only if the following conditions are met: + * - the element E that owns the axis is in the node-set + * - The nearest ancestor element of E in the node-set has a default + * namespace node in the node-set (default namespace nodes always + * have non-empty values in XPath) + * The latter condition eliminates unnecessary occurrences of xmlns="" in + * the canonical form since an element only receives an xmlns="" if its + * default namespace is empty and if it has an immediate parent in the + * canonical form that has a non-empty default namespace. To finish + * processing L, simply process every namespace node in L, except omit + * namespace node with local name xml, which defines the xml prefix, + * if its string value is http://www.w3.org/XML/1998/namespace. + * + * Exclusive XML Canonicalization v 1.0 (http://www.w3.org/TR/xml-exc-c14n) + * Canonical XML applied to a document subset requires the search of the + * ancestor nodes of each orphan element node for attributes in the xml + * namespace, such as xml:lang and xml:space. These are copied into the + * element node except if a declaration of the same attribute is already + * in the attribute axis of the element (whether or not it is included in + * the document subset). This search and copying are omitted from the + * Exclusive XML Canonicalization method. + * + * Returns 0 on success or -1 on fail. + */ +static int +xmlC14NProcessNamespacesAxis(xmlC14NCtxPtr ctx, xmlNodePtr cur, int visible) +{ + xmlNodePtr n; + xmlNsPtr ns, tmp; + xmlListPtr list; + int already_rendered; + int has_empty_ns = 0; + + if ((ctx == NULL) || (cur == NULL) || (cur->type != XML_ELEMENT_NODE)) { + xmlC14NErrParam("processing namespaces axis (c14n)"); + return (-1); + } + + /* + * Create a sorted list to store element namespaces + */ + list = xmlListCreate(NULL, (xmlListDataCompare) xmlC14NNsCompare); + if (list == NULL) { + xmlC14NErrInternal("creating namespaces list (c14n)"); + return (-1); + } + + /* check all namespaces */ + for(n = cur; n != NULL; n = n->parent) { + for(ns = n->nsDef; ns != NULL; ns = ns->next) { + tmp = xmlSearchNs(cur->doc, cur, ns->prefix); + + if((tmp == ns) && !xmlC14NIsXmlNs(ns) && xmlC14NIsVisible(ctx, ns, cur)) { + already_rendered = xmlC14NVisibleNsStackFind(ctx->ns_rendered, ns); + if(visible) { + xmlC14NVisibleNsStackAdd(ctx->ns_rendered, ns, cur); + } + if(!already_rendered) { + xmlListInsert(list, ns); + } + if(xmlStrlen(ns->prefix) == 0) { + has_empty_ns = 1; + } + } + } + } + + /** + * if the first node is not the default namespace node (a node with no + * namespace URI and no local name), then generate a space followed by + * xmlns="" if and only if the following conditions are met: + * - the element E that owns the axis is in the node-set + * - the nearest ancestor element of E in the node-set has a default + * namespace node in the node-set (default namespace nodes always + * have non-empty values in XPath) + */ + if(visible && !has_empty_ns) { + static xmlNs ns_default; + + memset(&ns_default, 0, sizeof(ns_default)); + if(!xmlC14NVisibleNsStackFind(ctx->ns_rendered, &ns_default)) { + xmlC14NPrintNamespaces(&ns_default, ctx); + } + } + + + /* + * print out all elements from list + */ + xmlListWalk(list, (xmlListWalker) xmlC14NPrintNamespaces, (const void *) ctx); + + /* + * Cleanup + */ + xmlListDelete(list); + return (0); +} + + +/** + * xmlExcC14NProcessNamespacesAxis: + * @ctx: the C14N context + * @node: the current node + * + * Prints out exclusive canonical namespace axis of the current node to the + * buffer from C14N context as follows + * + * Exclusive XML Canonicalization + * http://www.w3.org/TR/xml-exc-c14n + * + * If the element node is in the XPath subset then output the node in + * accordance with Canonical XML except for namespace nodes which are + * rendered as follows: + * + * 1. Render each namespace node iff: + * * it is visibly utilized by the immediate parent element or one of + * its attributes, or is present in InclusiveNamespaces PrefixList, and + * * its prefix and value do not appear in ns_rendered. ns_rendered is + * obtained by popping the state stack in order to obtain a list of + * prefixes and their values which have already been rendered by + * an output ancestor of the namespace node's parent element. + * 2. Append the rendered namespace node to the list ns_rendered of namespace + * nodes rendered by output ancestors. Push ns_rendered on state stack and + * recurse. + * 3. After the recursion returns, pop thestate stack. + * + * + * Returns 0 on success or -1 on fail. + */ +static int +xmlExcC14NProcessNamespacesAxis(xmlC14NCtxPtr ctx, xmlNodePtr cur, int visible) +{ + xmlNsPtr ns; + xmlListPtr list; + xmlAttrPtr attr; + int already_rendered; + int has_empty_ns = 0; + int has_visibly_utilized_empty_ns = 0; + int has_empty_ns_in_inclusive_list = 0; + + if ((ctx == NULL) || (cur == NULL) || (cur->type != XML_ELEMENT_NODE)) { + xmlC14NErrParam("processing namespaces axis (exc c14n)"); + return (-1); + } + + if(!xmlC14NIsExclusive(ctx)) { + xmlC14NErrParam("processing namespaces axis (exc c14n)"); + return (-1); + + } + + /* + * Create a sorted list to store element namespaces + */ + list = xmlListCreate(NULL, (xmlListDataCompare) xmlC14NNsCompare); + if (list == NULL) { + xmlC14NErrInternal("creating namespaces list (exc c14n)"); + return (-1); + } + + /* + * process inclusive namespaces: + * All namespace nodes appearing on inclusive ns list are + * handled as provided in Canonical XML + */ + if(ctx->inclusive_ns_prefixes != NULL) { + xmlChar *prefix; + int i; + + for (i = 0; ctx->inclusive_ns_prefixes[i] != NULL; ++i) { + prefix = ctx->inclusive_ns_prefixes[i]; + /* + * Special values for namespace with empty prefix + */ + if (xmlStrEqual(prefix, BAD_CAST "#default") + || xmlStrEqual(prefix, BAD_CAST "")) { + prefix = NULL; + has_empty_ns_in_inclusive_list = 1; + } + + ns = xmlSearchNs(cur->doc, cur, prefix); + if((ns != NULL) && !xmlC14NIsXmlNs(ns) && xmlC14NIsVisible(ctx, ns, cur)) { + already_rendered = xmlC14NVisibleNsStackFind(ctx->ns_rendered, ns); + if(visible) { + xmlC14NVisibleNsStackAdd(ctx->ns_rendered, ns, cur); + } + if(!already_rendered) { + xmlListInsert(list, ns); + } + if(xmlStrlen(ns->prefix) == 0) { + has_empty_ns = 1; + } + } + } + } + + /* add node namespace */ + if(cur->ns != NULL) { + ns = cur->ns; + } else { + ns = xmlSearchNs(cur->doc, cur, NULL); + has_visibly_utilized_empty_ns = 1; + } + if((ns != NULL) && !xmlC14NIsXmlNs(ns)) { + if(visible && xmlC14NIsVisible(ctx, ns, cur)) { + if(!xmlExcC14NVisibleNsStackFind(ctx->ns_rendered, ns, ctx)) { + xmlListInsert(list, ns); + } + } + if(visible) { + xmlC14NVisibleNsStackAdd(ctx->ns_rendered, ns, cur); + } + if(xmlStrlen(ns->prefix) == 0) { + has_empty_ns = 1; + } + } + + + /* add attributes */ + for(attr = cur->properties; attr != NULL; attr = attr->next) { + /* + * we need to check that attribute is visible and has non + * default namespace (XML Namespaces: "default namespaces + * do not apply directly to attributes") + */ + if((attr->ns != NULL) && !xmlC14NIsXmlNs(attr->ns) && xmlC14NIsVisible(ctx, attr, cur)) { + already_rendered = xmlExcC14NVisibleNsStackFind(ctx->ns_rendered, attr->ns, ctx); + xmlC14NVisibleNsStackAdd(ctx->ns_rendered, attr->ns, cur); + if(!already_rendered && visible) { + xmlListInsert(list, attr->ns); + } + if(xmlStrlen(attr->ns->prefix) == 0) { + has_empty_ns = 1; + } + } else if((attr->ns != NULL) && (xmlStrlen(attr->ns->prefix) == 0) && (xmlStrlen(attr->ns->href) == 0)) { + has_visibly_utilized_empty_ns = 1; + } + } + + /* + * Process xmlns="" + */ + if(visible && has_visibly_utilized_empty_ns && + !has_empty_ns && !has_empty_ns_in_inclusive_list) { + static xmlNs ns_default; + + memset(&ns_default, 0, sizeof(ns_default)); + + already_rendered = xmlExcC14NVisibleNsStackFind(ctx->ns_rendered, &ns_default, ctx); + if(!already_rendered) { + xmlC14NPrintNamespaces(&ns_default, ctx); + } + } else if(visible && !has_empty_ns && has_empty_ns_in_inclusive_list) { + static xmlNs ns_default; + + memset(&ns_default, 0, sizeof(ns_default)); + if(!xmlC14NVisibleNsStackFind(ctx->ns_rendered, &ns_default)) { + xmlC14NPrintNamespaces(&ns_default, ctx); + } + } + + + + /* + * print out all elements from list + */ + xmlListWalk(list, (xmlListWalker) xmlC14NPrintNamespaces, (const void *) ctx); + + /* + * Cleanup + */ + xmlListDelete(list); + return (0); +} + + +/** + * xmlC14NIsXmlAttr: + * @attr: the attr to check + * + * Checks whether the given attribute is a default "xml:" namespace + * with href="http://www.w3.org/XML/1998/namespace" + * + * Returns 1 if the node is default or 0 otherwise + */ + +/* todo: make it a define? */ +static int +xmlC14NIsXmlAttr(xmlAttrPtr attr) +{ + return ((attr->ns != NULL) && + (xmlC14NIsXmlNs(attr->ns) != 0)); +} + + +/** + * xmlC14NAttrsCompare: + * @attr1: the pointer tls o first attr + * @attr2: the pointer to second attr + * + * Prints the given attribute to the output buffer from C14N context. + * + * Returns -1 if attr1 < attr2, 0 if attr1 == attr2 or 1 if attr1 > attr2. + */ +static int +xmlC14NAttrsCompare(xmlAttrPtr attr1, xmlAttrPtr attr2) +{ + int ret = 0; + + /* + * Simple cases + */ + if (attr1 == attr2) + return (0); + if (attr1 == NULL) + return (-1); + if (attr2 == NULL) + return (1); + if (attr1->ns == attr2->ns) { + return (xmlStrcmp(attr1->name, attr2->name)); + } + + /* + * Attributes in the default namespace are first + * because the default namespace is not applied to + * unqualified attributes + */ + if (attr1->ns == NULL) + return (-1); + if (attr2->ns == NULL) + return (1); + if (attr1->ns->prefix == NULL) + return (-1); + if (attr2->ns->prefix == NULL) + return (1); + + ret = xmlStrcmp(attr1->ns->href, attr2->ns->href); + if (ret == 0) { + ret = xmlStrcmp(attr1->name, attr2->name); + } + return (ret); +} + + +/** + * xmlC14NPrintAttrs: + * @attr: the pointer to attr + * @ctx: the C14N context + * + * Prints out canonical attribute urrent node to the + * buffer from C14N context as follows + * + * Canonical XML v 1.0 (http://www.w3.org/TR/xml-c14n) + * + * Returns 1 on success or 0 on fail. + */ +static int +xmlC14NPrintAttrs(const xmlAttrPtr attr, xmlC14NCtxPtr ctx) +{ + xmlChar *value; + xmlChar *buffer; + + if ((attr == NULL) || (ctx == NULL)) { + xmlC14NErrParam("writing attributes"); + return (0); + } + + xmlOutputBufferWriteString(ctx->buf, " "); + if (attr->ns != NULL && xmlStrlen(attr->ns->prefix) > 0) { + xmlOutputBufferWriteString(ctx->buf, + (const char *) attr->ns->prefix); + xmlOutputBufferWriteString(ctx->buf, ":"); + } + xmlOutputBufferWriteString(ctx->buf, (const char *) attr->name); + xmlOutputBufferWriteString(ctx->buf, "=\""); + + value = xmlNodeListGetString(ctx->doc, attr->children, 1); + /* todo: should we log an error if value==NULL ? */ + if (value != NULL) { + buffer = xmlC11NNormalizeAttr(value); + xmlFree(value); + if (buffer != NULL) { + xmlOutputBufferWriteString(ctx->buf, (const char *) buffer); + xmlFree(buffer); + } else { + xmlC14NErrInternal("normalizing attributes axis"); + return (0); + } + } + xmlOutputBufferWriteString(ctx->buf, "\""); + return (1); +} + +/** + * xmlC14NFindHiddenParentAttr: + * + * Finds an attribute in a hidden parent node. + * + * Returns a pointer to the attribute node (if found) or NULL otherwise. + */ +static xmlAttrPtr +xmlC14NFindHiddenParentAttr(xmlC14NCtxPtr ctx, xmlNodePtr cur, const xmlChar * name, const xmlChar * ns) +{ + xmlAttrPtr res; + while((cur != NULL) && (!xmlC14NIsVisible(ctx, cur, cur->parent))) { + res = xmlHasNsProp(cur, name, ns); + if(res != NULL) { + return res; + } + + cur = cur->parent; + } + + return NULL; +} + +/** + * xmlC14NFixupBaseAttr: + * + * Fixes up the xml:base attribute + * + * Returns the newly created attribute or NULL + */ +static xmlAttrPtr +xmlC14NFixupBaseAttr(xmlC14NCtxPtr ctx, xmlAttrPtr xml_base_attr) +{ + xmlChar * res = NULL; + xmlNodePtr cur; + xmlAttrPtr attr; + xmlChar * tmp_str; + xmlChar * tmp_str2; + int tmp_str_len; + + if ((ctx == NULL) || (xml_base_attr == NULL) || (xml_base_attr->parent == NULL)) { + xmlC14NErrParam("processing xml:base attribute"); + return (NULL); + } + + /* start from current value */ + res = xmlNodeListGetString(ctx->doc, xml_base_attr->children, 1); + if(res == NULL) { + xmlC14NErrInternal("processing xml:base attribute - can't get attr value"); + return (NULL); + } + + /* go up the stack until we find a node that we rendered already */ + cur = xml_base_attr->parent->parent; + while((cur != NULL) && (!xmlC14NIsVisible(ctx, cur, cur->parent))) { + attr = xmlHasNsProp(cur, BAD_CAST "base", XML_XML_NAMESPACE); + if(attr != NULL) { + /* get attr value */ + tmp_str = xmlNodeListGetString(ctx->doc, attr->children, 1); + if(tmp_str == NULL) { + xmlFree(res); + + xmlC14NErrInternal("processing xml:base attribute - can't get attr value"); + return (NULL); + } + + /* we need to add '/' if our current base uri ends with '..' or '.' + to ensure that we are forced to go "up" all the time */ + tmp_str_len = xmlStrlen(tmp_str); + if(tmp_str_len > 1 && tmp_str[tmp_str_len - 2] == '.') { + tmp_str2 = xmlStrcat(tmp_str, BAD_CAST "/"); + if(tmp_str2 == NULL) { + xmlFree(tmp_str); + xmlFree(res); + + xmlC14NErrInternal("processing xml:base attribute - can't modify uri"); + return (NULL); + } + + tmp_str = tmp_str2; + } + + /* build uri */ + tmp_str2 = xmlBuildURI(res, tmp_str); + if(tmp_str2 == NULL) { + xmlFree(tmp_str); + xmlFree(res); + + xmlC14NErrInternal("processing xml:base attribute - can't construct uri"); + return (NULL); + } + + /* cleanup and set the new res */ + xmlFree(tmp_str); + xmlFree(res); + res = tmp_str2; + } + + /* next */ + cur = cur->parent; + } + + /* check if result uri is empty or not */ + if((res == NULL) || xmlStrEqual(res, BAD_CAST "")) { + xmlFree(res); + return (NULL); + } + + /* create and return the new attribute node */ + attr = xmlNewNsProp(NULL, xml_base_attr->ns, BAD_CAST "base", res); + if(attr == NULL) { + xmlFree(res); + + xmlC14NErrInternal("processing xml:base attribute - can't construct attribute"); + return (NULL); + } + + /* done */ + xmlFree(res); + return (attr); +} + +/** + * xmlC14NProcessAttrsAxis: + * @ctx: the C14N context + * @cur: the current node + * @parent_visible: the visibility of parent node + * @all_parents_visible: the visibility of all parent nodes + * + * Prints out canonical attribute axis of the current node to the + * buffer from C14N context as follows + * + * Canonical XML v 1.0 (http://www.w3.org/TR/xml-c14n) + * + * Attribute Axis + * In lexicographic order (ascending), process each node that + * is in the element's attribute axis and in the node-set. + * + * The processing of an element node E MUST be modified slightly + * when an XPath node-set is given as input and the element's + * parent is omitted from the node-set. + * + * + * Exclusive XML Canonicalization v 1.0 (http://www.w3.org/TR/xml-exc-c14n) + * + * Canonical XML applied to a document subset requires the search of the + * ancestor nodes of each orphan element node for attributes in the xml + * namespace, such as xml:lang and xml:space. These are copied into the + * element node except if a declaration of the same attribute is already + * in the attribute axis of the element (whether or not it is included in + * the document subset). This search and copying are omitted from the + * Exclusive XML Canonicalization method. + * + * Returns 0 on success or -1 on fail. + */ +static int +xmlC14NProcessAttrsAxis(xmlC14NCtxPtr ctx, xmlNodePtr cur, int parent_visible) +{ + xmlAttrPtr attr; + xmlListPtr list; + xmlAttrPtr attrs_to_delete = NULL; + + /* special processing for 1.1 spec */ + xmlAttrPtr xml_base_attr = NULL; + xmlAttrPtr xml_lang_attr = NULL; + xmlAttrPtr xml_space_attr = NULL; + + if ((ctx == NULL) || (cur == NULL) || (cur->type != XML_ELEMENT_NODE)) { + xmlC14NErrParam("processing attributes axis"); + return (-1); + } + + /* + * Create a sorted list to store element attributes + */ + list = xmlListCreate(NULL, (xmlListDataCompare) xmlC14NAttrsCompare); + if (list == NULL) { + xmlC14NErrInternal("creating attributes list"); + return (-1); + } + + switch(ctx->mode) { + case XML_C14N_1_0: + /* The processing of an element node E MUST be modified slightly when an XPath node-set is + * given as input and the element's parent is omitted from the node-set. The method for processing + * the attribute axis of an element E in the node-set is enhanced. All element nodes along E's + * ancestor axis are examined for nearest occurrences of attributes in the xml namespace, such + * as xml:lang and xml:space (whether or not they are in the node-set). From this list of attributes, + * remove any that are in E's attribute axis (whether or not they are in the node-set). Then, + * lexicographically merge this attribute list with the nodes of E's attribute axis that are in + * the node-set. The result of visiting the attribute axis is computed by processing the attribute + * nodes in this merged attribute list. + */ + + /* + * Add all visible attributes from current node. + */ + attr = cur->properties; + while (attr != NULL) { + /* check that attribute is visible */ + if (xmlC14NIsVisible(ctx, attr, cur)) { + xmlListInsert(list, attr); + } + attr = attr->next; + } + + /* + * Handle xml attributes + */ + if (parent_visible && (cur->parent != NULL) && + (!xmlC14NIsVisible(ctx, cur->parent, cur->parent->parent))) + { + xmlNodePtr tmp; + + /* + * If XPath node-set is not specified then the parent is always + * visible! + */ + tmp = cur->parent; + while (tmp != NULL) { + attr = tmp->properties; + while (attr != NULL) { + if (xmlC14NIsXmlAttr(attr) != 0) { + if (xmlListSearch(list, attr) == NULL) { + xmlListInsert(list, attr); + } + } + attr = attr->next; + } + tmp = tmp->parent; + } + } + + /* done */ + break; + case XML_C14N_EXCLUSIVE_1_0: + /* attributes in the XML namespace, such as xml:lang and xml:space + * are not imported into orphan nodes of the document subset + */ + + /* + * Add all visible attributes from current node. + */ + attr = cur->properties; + while (attr != NULL) { + /* check that attribute is visible */ + if (xmlC14NIsVisible(ctx, attr, cur)) { + xmlListInsert(list, attr); + } + attr = attr->next; + } + + /* do nothing special for xml attributes */ + break; + case XML_C14N_1_1: + /* The processing of an element node E MUST be modified slightly when an XPath node-set is + * given as input and some of the element's ancestors are omitted from the node-set. + * + * Simple inheritable attributes are attributes that have a value that requires at most a simple + * redeclaration. This redeclaration is done by supplying a new value in the child axis. The + * redeclaration of a simple inheritable attribute A contained in one of E's ancestors is done + * by supplying a value to an attribute Ae inside E with the same name. Simple inheritable attributes + * are xml:lang and xml:space. + * + * The method for processing the attribute axis of an element E in the node-set is hence enhanced. + * All element nodes along E's ancestor axis are examined for the nearest occurrences of simple + * inheritable attributes in the xml namespace, such as xml:lang and xml:space (whether or not they + * are in the node-set). From this list of attributes, any simple inheritable attributes that are + * already in E's attribute axis (whether or not they are in the node-set) are removed. Then, + * lexicographically merge this attribute list with the nodes of E's attribute axis that are in + * the node-set. The result of visiting the attribute axis is computed by processing the attribute + * nodes in this merged attribute list. + * + * The xml:id attribute is not a simple inheritable attribute and no processing of these attributes is + * performed. + * + * The xml:base attribute is not a simple inheritable attribute and requires special processing beyond + * a simple redeclaration. + * + * Attributes in the XML namespace other than xml:base, xml:id, xml:lang, and xml:space MUST be processed + * as ordinary attributes. + */ + + /* + * Add all visible attributes from current node. + */ + attr = cur->properties; + while (attr != NULL) { + /* special processing for XML attribute kiks in only when we have invisible parents */ + if ((!parent_visible) || (xmlC14NIsXmlAttr(attr) == 0)) { + /* check that attribute is visible */ + if (xmlC14NIsVisible(ctx, attr, cur)) { + xmlListInsert(list, attr); + } + } else { + int matched = 0; + + /* check for simple inheritance attributes */ + if((!matched) && (xml_lang_attr == NULL) && xmlStrEqual(attr->name, BAD_CAST "lang")) { + xml_lang_attr = attr; + matched = 1; + } + if((!matched) && (xml_space_attr == NULL) && xmlStrEqual(attr->name, BAD_CAST "space")) { + xml_space_attr = attr; + matched = 1; + } + + /* check for base attr */ + if((!matched) && (xml_base_attr == NULL) && xmlStrEqual(attr->name, BAD_CAST "base")) { + xml_base_attr = attr; + matched = 1; + } + + /* otherwise, it is a normal attribute, so just check if it is visible */ + if((!matched) && xmlC14NIsVisible(ctx, attr, cur)) { + xmlListInsert(list, attr); + } + } + + /* move to the next one */ + attr = attr->next; + } + + /* special processing for XML attribute kiks in only when we have invisible parents */ + if ((parent_visible)) { + + /* simple inheritance attributes - copy */ + if(xml_lang_attr == NULL) { + xml_lang_attr = xmlC14NFindHiddenParentAttr(ctx, cur->parent, BAD_CAST "lang", XML_XML_NAMESPACE); + } + if(xml_lang_attr != NULL) { + xmlListInsert(list, xml_lang_attr); + } + if(xml_space_attr == NULL) { + xml_space_attr = xmlC14NFindHiddenParentAttr(ctx, cur->parent, BAD_CAST "space", XML_XML_NAMESPACE); + } + if(xml_space_attr != NULL) { + xmlListInsert(list, xml_space_attr); + } + + /* base uri attribute - fix up */ + if(xml_base_attr == NULL) { + /* if we don't have base uri attribute, check if we have a "hidden" one above */ + xml_base_attr = xmlC14NFindHiddenParentAttr(ctx, cur->parent, BAD_CAST "base", XML_XML_NAMESPACE); + } + if(xml_base_attr != NULL) { + xml_base_attr = xmlC14NFixupBaseAttr(ctx, xml_base_attr); + if(xml_base_attr != NULL) { + xmlListInsert(list, xml_base_attr); + + /* note that we MUST delete returned attr node ourselves! */ + xml_base_attr->next = attrs_to_delete; + attrs_to_delete = xml_base_attr; + } + } + } + + /* done */ + break; + } + + /* + * print out all elements from list + */ + xmlListWalk(list, (xmlListWalker) xmlC14NPrintAttrs, (const void *) ctx); + + /* + * Cleanup + */ + xmlFreePropList(attrs_to_delete); + xmlListDelete(list); + return (0); +} + +/** + * xmlC14NCheckForRelativeNamespaces: + * @ctx: the C14N context + * @cur: the current element node + * + * Checks that current element node has no relative namespaces defined + * + * Returns 0 if the node has no relative namespaces or -1 otherwise. + */ +static int +xmlC14NCheckForRelativeNamespaces(xmlC14NCtxPtr ctx, xmlNodePtr cur) +{ + xmlNsPtr ns; + + if ((ctx == NULL) || (cur == NULL) || (cur->type != XML_ELEMENT_NODE)) { + xmlC14NErrParam("checking for relative namespaces"); + return (-1); + } + + ns = cur->nsDef; + while (ns != NULL) { + if (xmlStrlen(ns->href) > 0) { + xmlURIPtr uri; + + uri = xmlParseURI((const char *) ns->href); + if (uri == NULL) { + xmlC14NErrInternal("parsing namespace uri"); + return (-1); + } + if (xmlStrlen((const xmlChar *) uri->scheme) == 0) { + xmlC14NErrRelativeNamespace(uri->scheme); + xmlFreeURI(uri); + return (-1); + } + if ((xmlStrcasecmp((const xmlChar *) uri->scheme, BAD_CAST "urn") != 0) + && (xmlStrcasecmp((const xmlChar *) uri->scheme, BAD_CAST "dav") !=0) + && (xmlStrlen((const xmlChar *) uri->server) == 0)) { + xmlC14NErrRelativeNamespace(uri->scheme); + xmlFreeURI(uri); + return (-1); + } + xmlFreeURI(uri); + } + ns = ns->next; + } + return (0); +} + +/** + * xmlC14NProcessElementNode: + * @ctx: the pointer to C14N context object + * @cur: the node to process + * @visible: this node is visible + * @all_parents_visible: whether all the parents of this node are visible + * + * Canonical XML v 1.0 (http://www.w3.org/TR/xml-c14n) + * + * Element Nodes + * If the element is not in the node-set, then the result is obtained + * by processing the namespace axis, then the attribute axis, then + * processing the child nodes of the element that are in the node-set + * (in document order). If the element is in the node-set, then the result + * is an open angle bracket (<), the element QName, the result of + * processing the namespace axis, the result of processing the attribute + * axis, a close angle bracket (>), the result of processing the child + * nodes of the element that are in the node-set (in document order), an + * open angle bracket, a forward slash (/), the element QName, and a close + * angle bracket. + * + * Returns non-negative value on success or negative value on fail + */ +static int +xmlC14NProcessElementNode(xmlC14NCtxPtr ctx, xmlNodePtr cur, int visible) +{ + int ret; + xmlC14NVisibleNsStack state; + int parent_is_doc = 0; + + if ((ctx == NULL) || (cur == NULL) || (cur->type != XML_ELEMENT_NODE)) { + xmlC14NErrParam("processing element node"); + return (-1); + } + + /* + * Check relative relative namespaces: + * implementations of XML canonicalization MUST report an operation + * failure on documents containing relative namespace URIs. + */ + if (xmlC14NCheckForRelativeNamespaces(ctx, cur) < 0) { + xmlC14NErrInternal("checking for relative namespaces"); + return (-1); + } + + + /* + * Save ns_rendered stack position + */ + memset(&state, 0, sizeof(state)); + xmlC14NVisibleNsStackSave(ctx->ns_rendered, &state); + + if (visible) { + if (ctx->parent_is_doc) { + /* save this flag into the stack */ + parent_is_doc = ctx->parent_is_doc; + ctx->parent_is_doc = 0; + ctx->pos = XMLC14N_INSIDE_DOCUMENT_ELEMENT; + } + xmlOutputBufferWriteString(ctx->buf, "<"); + + if ((cur->ns != NULL) && (xmlStrlen(cur->ns->prefix) > 0)) { + xmlOutputBufferWriteString(ctx->buf, + (const char *) cur->ns->prefix); + xmlOutputBufferWriteString(ctx->buf, ":"); + } + xmlOutputBufferWriteString(ctx->buf, (const char *) cur->name); + } + + if (!xmlC14NIsExclusive(ctx)) { + ret = xmlC14NProcessNamespacesAxis(ctx, cur, visible); + } else { + ret = xmlExcC14NProcessNamespacesAxis(ctx, cur, visible); + } + if (ret < 0) { + xmlC14NErrInternal("processing namespaces axis"); + return (-1); + } + /* todo: shouldn't this go to "visible only"? */ + if(visible) { + xmlC14NVisibleNsStackShift(ctx->ns_rendered); + } + + ret = xmlC14NProcessAttrsAxis(ctx, cur, visible); + if (ret < 0) { + xmlC14NErrInternal("processing attributes axis"); + return (-1); + } + + if (visible) { + xmlOutputBufferWriteString(ctx->buf, ">"); + } + if (cur->children != NULL) { + ret = xmlC14NProcessNodeList(ctx, cur->children); + if (ret < 0) { + xmlC14NErrInternal("processing childrens list"); + return (-1); + } + } + if (visible) { + xmlOutputBufferWriteString(ctx->buf, "</"); + if ((cur->ns != NULL) && (xmlStrlen(cur->ns->prefix) > 0)) { + xmlOutputBufferWriteString(ctx->buf, + (const char *) cur->ns->prefix); + xmlOutputBufferWriteString(ctx->buf, ":"); + } + xmlOutputBufferWriteString(ctx->buf, (const char *) cur->name); + xmlOutputBufferWriteString(ctx->buf, ">"); + if (parent_is_doc) { + /* restore this flag from the stack for next node */ + ctx->parent_is_doc = parent_is_doc; + ctx->pos = XMLC14N_AFTER_DOCUMENT_ELEMENT; + } + } + + /* + * Restore ns_rendered stack position + */ + xmlC14NVisibleNsStackRestore(ctx->ns_rendered, &state); + return (0); +} + +/** + * xmlC14NProcessNode: + * @ctx: the pointer to C14N context object + * @cur: the node to process + * + * Processes the given node + * + * Returns non-negative value on success or negative value on fail + */ +static int +xmlC14NProcessNode(xmlC14NCtxPtr ctx, xmlNodePtr cur) +{ + int ret = 0; + int visible; + + if ((ctx == NULL) || (cur == NULL)) { + xmlC14NErrParam("processing node"); + return (-1); + } + + visible = xmlC14NIsVisible(ctx, cur, cur->parent); + switch (cur->type) { + case XML_ELEMENT_NODE: + ret = xmlC14NProcessElementNode(ctx, cur, visible); + break; + case XML_CDATA_SECTION_NODE: + case XML_TEXT_NODE: + /* + * Text Nodes + * the string value, except all ampersands are replaced + * by &, all open angle brackets (<) are replaced by <, all closing + * angle brackets (>) are replaced by >, and all #xD characters are + * replaced by 
. + */ + /* cdata sections are processed as text nodes */ + /* todo: verify that cdata sections are included in XPath nodes set */ + if ((visible) && (cur->content != NULL)) { + xmlChar *buffer; + + buffer = xmlC11NNormalizeText(cur->content); + if (buffer != NULL) { + xmlOutputBufferWriteString(ctx->buf, + (const char *) buffer); + xmlFree(buffer); + } else { + xmlC14NErrInternal("normalizing text node"); + return (-1); + } + } + break; + case XML_PI_NODE: + /* + * Processing Instruction (PI) Nodes- + * The opening PI symbol (<?), the PI target name of the node, + * a leading space and the string value if it is not empty, and + * the closing PI symbol (?>). If the string value is empty, + * then the leading space is not added. Also, a trailing #xA is + * rendered after the closing PI symbol for PI children of the + * root node with a lesser document order than the document + * element, and a leading #xA is rendered before the opening PI + * symbol of PI children of the root node with a greater document + * order than the document element. + */ + if (visible) { + if (ctx->pos == XMLC14N_AFTER_DOCUMENT_ELEMENT) { + xmlOutputBufferWriteString(ctx->buf, "\x0A<?"); + } else { + xmlOutputBufferWriteString(ctx->buf, "<?"); + } + + xmlOutputBufferWriteString(ctx->buf, + (const char *) cur->name); + if ((cur->content != NULL) && (*(cur->content) != '\0')) { + xmlChar *buffer; + + xmlOutputBufferWriteString(ctx->buf, " "); + + /* todo: do we need to normalize pi? */ + buffer = xmlC11NNormalizePI(cur->content); + if (buffer != NULL) { + xmlOutputBufferWriteString(ctx->buf, + (const char *) buffer); + xmlFree(buffer); + } else { + xmlC14NErrInternal("normalizing pi node"); + return (-1); + } + } + + if (ctx->pos == XMLC14N_BEFORE_DOCUMENT_ELEMENT) { + xmlOutputBufferWriteString(ctx->buf, "?>\x0A"); + } else { + xmlOutputBufferWriteString(ctx->buf, "?>"); + } + } + break; + case XML_COMMENT_NODE: + /* + * Comment Nodes + * Nothing if generating canonical XML without comments. For + * canonical XML with comments, generate the opening comment + * symbol (<!--), the string value of the node, and the + * closing comment symbol (-->). Also, a trailing #xA is rendered + * after the closing comment symbol for comment children of the + * root node with a lesser document order than the document + * element, and a leading #xA is rendered before the opening + * comment symbol of comment children of the root node with a + * greater document order than the document element. (Comment + * children of the root node represent comments outside of the + * top-level document element and outside of the document type + * declaration). + */ + if (visible && ctx->with_comments) { + if (ctx->pos == XMLC14N_AFTER_DOCUMENT_ELEMENT) { + xmlOutputBufferWriteString(ctx->buf, "\x0A<!--"); + } else { + xmlOutputBufferWriteString(ctx->buf, "<!--"); + } + + if (cur->content != NULL) { + xmlChar *buffer; + + /* todo: do we need to normalize comment? */ + buffer = xmlC11NNormalizeComment(cur->content); + if (buffer != NULL) { + xmlOutputBufferWriteString(ctx->buf, + (const char *) buffer); + xmlFree(buffer); + } else { + xmlC14NErrInternal("normalizing comment node"); + return (-1); + } + } + + if (ctx->pos == XMLC14N_BEFORE_DOCUMENT_ELEMENT) { + xmlOutputBufferWriteString(ctx->buf, "-->\x0A"); + } else { + xmlOutputBufferWriteString(ctx->buf, "-->"); + } + } + break; + case XML_DOCUMENT_NODE: + case XML_DOCUMENT_FRAG_NODE: /* should be processed as document? */ +#ifdef LIBXML_DOCB_ENABLED + case XML_DOCB_DOCUMENT_NODE: /* should be processed as document? */ +#endif +#ifdef LIBXML_HTML_ENABLED + case XML_HTML_DOCUMENT_NODE: /* should be processed as document? */ +#endif + if (cur->children != NULL) { + ctx->pos = XMLC14N_BEFORE_DOCUMENT_ELEMENT; + ctx->parent_is_doc = 1; + ret = xmlC14NProcessNodeList(ctx, cur->children); + } + break; + + case XML_ATTRIBUTE_NODE: + xmlC14NErrInvalidNode("XML_ATTRIBUTE_NODE", "processing node"); + return (-1); + case XML_NAMESPACE_DECL: + xmlC14NErrInvalidNode("XML_NAMESPACE_DECL", "processing node"); + return (-1); + case XML_ENTITY_REF_NODE: + xmlC14NErrInvalidNode("XML_ENTITY_REF_NODE", "processing node"); + return (-1); + case XML_ENTITY_NODE: + xmlC14NErrInvalidNode("XML_ENTITY_NODE", "processing node"); + return (-1); + + case XML_DOCUMENT_TYPE_NODE: + case XML_NOTATION_NODE: + case XML_DTD_NODE: + case XML_ELEMENT_DECL: + case XML_ATTRIBUTE_DECL: + case XML_ENTITY_DECL: +#ifdef LIBXML_XINCLUDE_ENABLED + case XML_XINCLUDE_START: + case XML_XINCLUDE_END: +#endif + /* + * should be ignored according to "W3C Canonical XML" + */ + break; + default: + xmlC14NErrUnknownNode(cur->type, "processing node"); + return (-1); + } + + return (ret); +} + +/** + * xmlC14NProcessNodeList: + * @ctx: the pointer to C14N context object + * @cur: the node to start from + * + * Processes all nodes in the row starting from cur. + * + * Returns non-negative value on success or negative value on fail + */ +static int +xmlC14NProcessNodeList(xmlC14NCtxPtr ctx, xmlNodePtr cur) +{ + int ret; + + if (ctx == NULL) { + xmlC14NErrParam("processing node list"); + return (-1); + } + + for (ret = 0; cur != NULL && ret >= 0; cur = cur->next) { + ret = xmlC14NProcessNode(ctx, cur); + } + return (ret); +} + + +/** + * xmlC14NFreeCtx: + * @ctx: the pointer to C14N context object + * + * Cleanups the C14N context object. + */ + +static void +xmlC14NFreeCtx(xmlC14NCtxPtr ctx) +{ + if (ctx == NULL) { + xmlC14NErrParam("freeing context"); + return; + } + + if (ctx->ns_rendered != NULL) { + xmlC14NVisibleNsStackDestroy(ctx->ns_rendered); + } + xmlFree(ctx); +} + +/** + * xmlC14NNewCtx: + * @doc: the XML document for canonization + * @is_visible_callback:the function to use to determine is node visible + * or not + * @user_data: the first parameter for @is_visible_callback function + * (in most cases, it is nodes set) + * @mode: the c14n mode (see @xmlC14NMode) + * @inclusive_ns_prefixe the list of inclusive namespace prefixes + * ended with a NULL or NULL if there is no + * inclusive namespaces (only for ` + * canonicalization) + * @with_comments: include comments in the result (!=0) or not (==0) + * @buf: the output buffer to store canonical XML; this + * buffer MUST have encoder==NULL because C14N requires + * UTF-8 output + * + * Creates new C14N context object to store C14N parameters. + * + * Returns pointer to newly created object (success) or NULL (fail) + */ +static xmlC14NCtxPtr +xmlC14NNewCtx(xmlDocPtr doc, + xmlC14NIsVisibleCallback is_visible_callback, void* user_data, + xmlC14NMode mode, xmlChar ** inclusive_ns_prefixes, + int with_comments, xmlOutputBufferPtr buf) +{ + xmlC14NCtxPtr ctx = NULL; + + if ((doc == NULL) || (buf == NULL)) { + xmlC14NErrParam("creating new context"); + return (NULL); + } + + /* + * Validate the encoding output buffer encoding + */ + if (buf->encoder != NULL) { + xmlC14NErr(ctx, (xmlNodePtr) doc, XML_C14N_REQUIRES_UTF8, +"xmlC14NNewCtx: output buffer encoder != NULL but C14N requires UTF8 output\n"); + return (NULL); + } + + /* + * Validate the XML document encoding value, if provided. + */ + if (doc->charset != XML_CHAR_ENCODING_UTF8) { + xmlC14NErr(ctx, (xmlNodePtr) doc, XML_C14N_REQUIRES_UTF8, + "xmlC14NNewCtx: source document not in UTF8\n"); + return (NULL); + } + + /* + * Allocate a new xmlC14NCtxPtr and fill the fields. + */ + ctx = (xmlC14NCtxPtr) xmlMalloc(sizeof(xmlC14NCtx)); + if (ctx == NULL) { + xmlC14NErrMemory("creating context"); + return (NULL); + } + memset(ctx, 0, sizeof(xmlC14NCtx)); + + /* + * initialize C14N context + */ + ctx->doc = doc; + ctx->with_comments = with_comments; + ctx->is_visible_callback = is_visible_callback; + ctx->user_data = user_data; + ctx->buf = buf; + ctx->parent_is_doc = 1; + ctx->pos = XMLC14N_BEFORE_DOCUMENT_ELEMENT; + ctx->ns_rendered = xmlC14NVisibleNsStackCreate(); + + if(ctx->ns_rendered == NULL) { + xmlC14NErr(ctx, (xmlNodePtr) doc, XML_C14N_CREATE_STACK, + "xmlC14NNewCtx: xmlC14NVisibleNsStackCreate failed\n"); + xmlC14NFreeCtx(ctx); + return (NULL); + } + + /* + * Set "mode" flag and remember list of incluseve prefixes + * for exclusive c14n + */ + ctx->mode = mode; + if(xmlC14NIsExclusive(ctx)) { + ctx->inclusive_ns_prefixes = inclusive_ns_prefixes; + } + return (ctx); +} + +/** + * xmlC14NExecute: + * @doc: the XML document for canonization + * @is_visible_callback:the function to use to determine is node visible + * or not + * @user_data: the first parameter for @is_visible_callback function + * (in most cases, it is nodes set) + * @mode: the c14n mode (see @xmlC14NMode) + * @inclusive_ns_prefixes: the list of inclusive namespace prefixes + * ended with a NULL or NULL if there is no + * inclusive namespaces (only for exclusive + * canonicalization, ignored otherwise) + * @with_comments: include comments in the result (!=0) or not (==0) + * @buf: the output buffer to store canonical XML; this + * buffer MUST have encoder==NULL because C14N requires + * UTF-8 output + * + * Dumps the canonized image of given XML document into the provided buffer. + * For details see "Canonical XML" (http://www.w3.org/TR/xml-c14n) or + * "Exclusive XML Canonicalization" (http://www.w3.org/TR/xml-exc-c14n) + * + * Returns non-negative value on success or a negative value on fail + */ +int +xmlC14NExecute(xmlDocPtr doc, xmlC14NIsVisibleCallback is_visible_callback, + void* user_data, int mode, xmlChar **inclusive_ns_prefixes, + int with_comments, xmlOutputBufferPtr buf) { + + xmlC14NCtxPtr ctx; + xmlC14NMode c14n_mode = XML_C14N_1_0; + int ret; + + if ((buf == NULL) || (doc == NULL)) { + xmlC14NErrParam("executing c14n"); + return (-1); + } + + /* for backward compatibility, we have to have "mode" as "int" + and here we check that user gives valid value */ + switch(mode) { + case XML_C14N_1_0: + case XML_C14N_EXCLUSIVE_1_0: + case XML_C14N_1_1: + c14n_mode = (xmlC14NMode)mode; + break; + default: + xmlC14NErrParam("invalid mode for executing c14n"); + return (-1); + } + + /* + * Validate the encoding output buffer encoding + */ + if (buf->encoder != NULL) { + xmlC14NErr(NULL, (xmlNodePtr) doc, XML_C14N_REQUIRES_UTF8, +"xmlC14NExecute: output buffer encoder != NULL but C14N requires UTF8 output\n"); + return (-1); + } + + ctx = xmlC14NNewCtx(doc, is_visible_callback, user_data, + c14n_mode, inclusive_ns_prefixes, + with_comments, buf); + if (ctx == NULL) { + xmlC14NErr(NULL, (xmlNodePtr) doc, XML_C14N_CREATE_CTXT, + "xmlC14NExecute: unable to create C14N context\n"); + return (-1); + } + + + + /* + * Root Node + * The root node is the parent of the top-level document element. The + * result of processing each of its child nodes that is in the node-set + * in document order. The root node does not generate a byte order mark, + * XML declaration, nor anything from within the document type + * declaration. + */ + if (doc->children != NULL) { + ret = xmlC14NProcessNodeList(ctx, doc->children); + if (ret < 0) { + xmlC14NErrInternal("processing docs children list"); + xmlC14NFreeCtx(ctx); + return (-1); + } + } + + /* + * Flush buffer to get number of bytes written + */ + ret = xmlOutputBufferFlush(buf); + if (ret < 0) { + xmlC14NErrInternal("flushing output buffer"); + xmlC14NFreeCtx(ctx); + return (-1); + } + + /* + * Cleanup + */ + xmlC14NFreeCtx(ctx); + return (ret); +} + +/** + * xmlC14NDocSaveTo: + * @doc: the XML document for canonization + * @nodes: the nodes set to be included in the canonized image + * or NULL if all document nodes should be included + * @mode: the c14n mode (see @xmlC14NMode) + * @inclusive_ns_prefixes: the list of inclusive namespace prefixes + * ended with a NULL or NULL if there is no + * inclusive namespaces (only for exclusive + * canonicalization, ignored otherwise) + * @with_comments: include comments in the result (!=0) or not (==0) + * @buf: the output buffer to store canonical XML; this + * buffer MUST have encoder==NULL because C14N requires + * UTF-8 output + * + * Dumps the canonized image of given XML document into the provided buffer. + * For details see "Canonical XML" (http://www.w3.org/TR/xml-c14n) or + * "Exclusive XML Canonicalization" (http://www.w3.org/TR/xml-exc-c14n) + * + * Returns non-negative value on success or a negative value on fail + */ +int +xmlC14NDocSaveTo(xmlDocPtr doc, xmlNodeSetPtr nodes, + int mode, xmlChar ** inclusive_ns_prefixes, + int with_comments, xmlOutputBufferPtr buf) { + return(xmlC14NExecute(doc, + (xmlC14NIsVisibleCallback)xmlC14NIsNodeInNodeset, + nodes, + mode, + inclusive_ns_prefixes, + with_comments, + buf)); +} + + +/** + * xmlC14NDocDumpMemory: + * @doc: the XML document for canonization + * @nodes: the nodes set to be included in the canonized image + * or NULL if all document nodes should be included + * @mode: the c14n mode (see @xmlC14NMode) + * @inclusive_ns_prefixes: the list of inclusive namespace prefixes + * ended with a NULL or NULL if there is no + * inclusive namespaces (only for exclusive + * canonicalization, ignored otherwise) + * @with_comments: include comments in the result (!=0) or not (==0) + * @doc_txt_ptr: the memory pointer for allocated canonical XML text; + * the caller of this functions is responsible for calling + * xmlFree() to free allocated memory + * + * Dumps the canonized image of given XML document into memory. + * For details see "Canonical XML" (http://www.w3.org/TR/xml-c14n) or + * "Exclusive XML Canonicalization" (http://www.w3.org/TR/xml-exc-c14n) + * + * Returns the number of bytes written on success or a negative value on fail + */ +int +xmlC14NDocDumpMemory(xmlDocPtr doc, xmlNodeSetPtr nodes, + int mode, xmlChar ** inclusive_ns_prefixes, + int with_comments, xmlChar ** doc_txt_ptr) +{ + int ret; + xmlOutputBufferPtr buf; + + if (doc_txt_ptr == NULL) { + xmlC14NErrParam("dumping doc to memory"); + return (-1); + } + + *doc_txt_ptr = NULL; + + /* + * create memory buffer with UTF8 (default) encoding + */ + buf = xmlAllocOutputBuffer(NULL); + if (buf == NULL) { + xmlC14NErrMemory("creating output buffer"); + return (-1); + } + + /* + * canonize document and write to buffer + */ + ret = xmlC14NDocSaveTo(doc, nodes, mode, inclusive_ns_prefixes, + with_comments, buf); + if (ret < 0) { + xmlC14NErrInternal("saving doc to output buffer"); + (void) xmlOutputBufferClose(buf); + return (-1); + } + + ret = xmlBufUse(buf->buffer); + if (ret > 0) { + *doc_txt_ptr = xmlStrndup(xmlBufContent(buf->buffer), ret); + } + (void) xmlOutputBufferClose(buf); + + if ((*doc_txt_ptr == NULL) && (ret > 0)) { + xmlC14NErrMemory("coping canonicanized document"); + return (-1); + } + return (ret); +} + +/** + * xmlC14NDocSave: + * @doc: the XML document for canonization + * @nodes: the nodes set to be included in the canonized image + * or NULL if all document nodes should be included + * @mode: the c14n mode (see @xmlC14NMode) + * @inclusive_ns_prefixes: the list of inclusive namespace prefixes + * ended with a NULL or NULL if there is no + * inclusive namespaces (only for exclusive + * canonicalization, ignored otherwise) + * @with_comments: include comments in the result (!=0) or not (==0) + * @filename: the filename to store canonical XML image + * @compression: the compression level (zlib requred): + * -1 - libxml default, + * 0 - uncompressed, + * >0 - compression level + * + * Dumps the canonized image of given XML document into the file. + * For details see "Canonical XML" (http://www.w3.org/TR/xml-c14n) or + * "Exclusive XML Canonicalization" (http://www.w3.org/TR/xml-exc-c14n) + * + * Returns the number of bytes written success or a negative value on fail + */ +int +xmlC14NDocSave(xmlDocPtr doc, xmlNodeSetPtr nodes, + int mode, xmlChar ** inclusive_ns_prefixes, + int with_comments, const char *filename, int compression) +{ + xmlOutputBufferPtr buf; + int ret; + + if (filename == NULL) { + xmlC14NErrParam("saving doc"); + return (-1); + } +#ifdef HAVE_ZLIB_H + if (compression < 0) + compression = xmlGetCompressMode(); +#endif + + /* + * save the content to a temp buffer, use default UTF8 encoding. + */ + buf = xmlOutputBufferCreateFilename(filename, NULL, compression); + if (buf == NULL) { + xmlC14NErrInternal("creating temporary filename"); + return (-1); + } + + /* + * canonize document and write to buffer + */ + ret = xmlC14NDocSaveTo(doc, nodes, mode, inclusive_ns_prefixes, + with_comments, buf); + if (ret < 0) { + xmlC14NErrInternal("cannicanize document to buffer"); + (void) xmlOutputBufferClose(buf); + return (-1); + } + + /* + * get the numbers of bytes written + */ + ret = xmlOutputBufferClose(buf); + return (ret); +} + + + +/* + * Macro used to grow the current buffer. + */ +#define growBufferReentrant() { \ + buffer_size *= 2; \ + buffer = (xmlChar *) \ + xmlRealloc(buffer, buffer_size * sizeof(xmlChar)); \ + if (buffer == NULL) { \ + xmlC14NErrMemory("growing buffer"); \ + return(NULL); \ + } \ +} + +/** + * xmlC11NNormalizeString: + * @input: the input string + * @mode: the normalization mode (attribute, comment, PI or text) + * + * Converts a string to a canonical (normalized) format. The code is stolen + * from xmlEncodeEntitiesReentrant(). Added normalization of \x09, \x0a, \x0A + * and the @mode parameter + * + * Returns a normalized string (caller is responsible for calling xmlFree()) + * or NULL if an error occurs + */ +static xmlChar * +xmlC11NNormalizeString(const xmlChar * input, + xmlC14NNormalizationMode mode) +{ + const xmlChar *cur = input; + xmlChar *buffer = NULL; + xmlChar *out = NULL; + int buffer_size = 0; + + if (input == NULL) + return (NULL); + + /* + * allocate an translation buffer. + */ + buffer_size = 1000; + buffer = (xmlChar *) xmlMallocAtomic(buffer_size * sizeof(xmlChar)); + if (buffer == NULL) { + xmlC14NErrMemory("allocating buffer"); + return (NULL); + } + out = buffer; + + while (*cur != '\0') { + if ((out - buffer) > (buffer_size - 10)) { + int indx = out - buffer; + + growBufferReentrant(); + out = &buffer[indx]; + } + + if ((*cur == '<') && ((mode == XMLC14N_NORMALIZE_ATTR) || + (mode == XMLC14N_NORMALIZE_TEXT))) { + *out++ = '&'; + *out++ = 'l'; + *out++ = 't'; + *out++ = ';'; + } else if ((*cur == '>') && (mode == XMLC14N_NORMALIZE_TEXT)) { + *out++ = '&'; + *out++ = 'g'; + *out++ = 't'; + *out++ = ';'; + } else if ((*cur == '&') && ((mode == XMLC14N_NORMALIZE_ATTR) || + (mode == XMLC14N_NORMALIZE_TEXT))) { + *out++ = '&'; + *out++ = 'a'; + *out++ = 'm'; + *out++ = 'p'; + *out++ = ';'; + } else if ((*cur == '"') && (mode == XMLC14N_NORMALIZE_ATTR)) { + *out++ = '&'; + *out++ = 'q'; + *out++ = 'u'; + *out++ = 'o'; + *out++ = 't'; + *out++ = ';'; + } else if ((*cur == '\x09') && (mode == XMLC14N_NORMALIZE_ATTR)) { + *out++ = '&'; + *out++ = '#'; + *out++ = 'x'; + *out++ = '9'; + *out++ = ';'; + } else if ((*cur == '\x0A') && (mode == XMLC14N_NORMALIZE_ATTR)) { + *out++ = '&'; + *out++ = '#'; + *out++ = 'x'; + *out++ = 'A'; + *out++ = ';'; + } else if ((*cur == '\x0D') && ((mode == XMLC14N_NORMALIZE_ATTR) || + (mode == XMLC14N_NORMALIZE_TEXT) || + (mode == XMLC14N_NORMALIZE_COMMENT) || + (mode == XMLC14N_NORMALIZE_PI))) { + *out++ = '&'; + *out++ = '#'; + *out++ = 'x'; + *out++ = 'D'; + *out++ = ';'; + } else { + /* + * Works because on UTF-8, all extended sequences cannot + * result in bytes in the ASCII range. + */ + *out++ = *cur; + } + cur++; + } + *out = 0; + return (buffer); +} +#endif /* LIBXML_OUTPUT_ENABLED */ +#define bottom_c14n +#include "elfgcchack.h" +#endif /* LIBXML_C14N_ENABLED */ |