parser: Make undeclared entities in XML content fatal

When parsing XML content with functions like xmlParseBalancedChunk or
xmlParseInNodeContext, make undeclared entities always a fatal error to
match 2.13 behavior.

This was deliberately changed in 4f329dc, probably to make the tests
pass.

Should fix #895.

Author: nwellnhof

parser.c

@@ -12368,13 +12368,6 @@ xmlCtxtParseContent(xmlParserCtxtPtr ctxt, xmlParserInputPtr input,
     {
         xmlCtxtInitializeLate(ctxt);
 
-        /*
-         * This hack lowers the error level of undeclared entities
-         * from XML_ERR_FATAL (well-formedness error) to XML_ERR_ERROR
-         * or XML_ERR_WARNING.
-         */
-        ctxt->hasExternalSubset = 1;
-
         /*
          * initialize the SAX2 namespaces stack
          */

runtest.c

@@ -2355,6 +2355,14 @@ testParseContent(xmlParserCtxtPtr ctxt, xmlDocPtr doc, const char *filename) {
     char *content, *roundTrip;
     int ret = 0;
 
+    /*
+     * Undeclared entities aren't a fatal error if there's an
+     * external DTD. When parsing content, we assume there's no
+     * DTD, so the undeclared entity test would fail.
+     */
+    if (strcmp(filename, "./test/undeclared-entity.xml") == 0)
+        return 0;
+
     if (ctxt->html) {
         xmlNodePtr cur;
 

testparser.c

@@ -17,6 +17,13 @@
 
 #include <string.h>
 
+#ifdef LIBXML_SAX1_ENABLED
+static void
+ignoreError(void *ctxt ATTRIBUTE_UNUSED,
+            const xmlError *error ATTRIBUTE_UNUSED) {
+}
+#endif
+
 static int
 testNewDocNode(void) {
     xmlNodePtr node;
@@ -128,6 +135,53 @@ testCFileIO(void) {
     return err;
 }
 
+/*
+ * The exact rules when undeclared entities are a fatal error
+ * depend on some conditions that aren't recovered from the
+ * context document when parsing XML content. This test case
+ * demonstrates such an asymmetry.
+ */
+static int
+testUndeclEntInContent(void) {
+    const char xml[] = "<!DOCTYPE doc SYSTEM 'my.dtd'><doc>&undecl;</doc>";
+    const char content[] = "<doc>&undecl;</doc>";
+    xmlDocPtr doc;
+    xmlNodePtr root, list;
+    int options = XML_PARSE_NOENT | XML_PARSE_NOERROR;
+    int err = 0;
+    int res;
+
+    /* Parsing the document succeeds because of the external DTD. */
+    doc = xmlReadDoc(BAD_CAST xml, NULL, NULL, options);
+    root = xmlDocGetRootElement(doc);
+
+    /* Parsing content fails. */
+
+    res = xmlParseInNodeContext(root, content, sizeof(content) - 1, options,
+                                &list);
+    if (res != XML_ERR_UNDECLARED_ENTITY || list != NULL) {
+        fprintf(stderr, "Wrong result from xmlParseInNodeContext\n");
+        err = 1;
+    }
+    xmlFreeNodeList(list);
+
+#ifdef LIBXML_SAX1_ENABLED
+    xmlSetStructuredErrorFunc(NULL, ignoreError);
+    res = xmlParseBalancedChunkMemory(doc, NULL, NULL, 0, BAD_CAST content,
+                                      &list);
+    if (res != XML_ERR_UNDECLARED_ENTITY || list != NULL) {
+        fprintf(stderr, "Wrong result from xmlParseBalancedChunkMemory\n");
+        err = 1;
+    }
+    xmlFreeNodeList(list);
+    xmlSetStructuredErrorFunc(NULL, NULL);
+#endif /* LIBXML_SAX1_ENABLED */
+
+    xmlFreeDoc(doc);
+
+    return err;
+}
+
 #ifdef LIBXML_VALID_ENABLED
 static void
 testSwitchDtdExtSubset(void *vctxt, const xmlChar *name ATTRIBUTE_UNUSED,
@@ -1195,6 +1249,7 @@ main(void) {
     err |= testUnsupportedEncoding();
     err |= testNodeGetContent();
     err |= testCFileIO();
+    err |= testUndeclEntInContent();
 #ifdef LIBXML_VALID_ENABLED
     err |= testSwitchDtd();
 #endif