Daniel Veillard wrote:
A few months ago, I came across a "bug" where whitespace nodes as a direct child of the <body> tag would be removed. The problem is similar in that pure whitespace nodes are forbidden by the strict DTD, but allowed by the transitional DTD. In this case, the applied patch checked the DTD in use with code like dtd = xmlGetIntSubset(ctxt->myDoc); if (dtd != NULL && dtd->ExternalID != NULL) { if (!xmlStrcasecmp(dtd->ExternalID, BAD_CAST "-//W3C//DTD HTML 4.01//EN") || !xmlStrcasecmp(dtd->ExternalID, BAD_CAST "-//W3C//DTD HTML 4//EN")) { (line 2060, HTMLparser.c). This code assumes that HTML 4 and HTML 4.01 are the only strict non-XML DTDs in existence. Something similar might be useful for this issue - the <p> tags are not needed for a Transitional DTD. I'll have a look to see if there's an easy fix at the weekend, if nobody's supplied a patch before that :-)Yes, thanks ! That sounds the right approach to me, I would just turn merge that with a new htmlParserOption HTML_PARSE_STRICT, which could be either passed by the user to maintain the current behaviour or activated by default when the DOCTYPE is read if it happen to be a Strict HTML one. make sense ?
Actually, I got around to testing this today, and it looks it was fixed by the above checkin - version 1.196 of HTMLparser.c. iSteve, can you check the behaviour with the latest release, or from CVS? Daniel, this behaviour was not the same after the patch - htmlNoContentElements[] was altered, so the <p> tag doesn't get added for either the strict or non-strict DTD. Attached is a patch which does restores the original behaviour for strict DTDs. I'll leave it up to your judgement as to whether you want it back. There is no support for HTML_PARSE_STRICT, do you still think it might be needed? I don't like the cast to xmlChar** in the patch - but I couldn't find a pointer definition for 'elements' which wouldn't give a type mismatch warning... Gary.
Index: HTMLparser.c
===================================================================
RCS file: /cvs/gnome/libxml2/HTMLparser.c,v
retrieving revision 1.199
diff -c -r1.199 HTMLparser.c
*** HTMLparser.c 10 Dec 2005 11:11:11 -0000 1.199
--- HTMLparser.c 12 Jan 2006 13:36:54 -0000
***************
*** 968,973 ****
--- 968,985 ----
};
/*
+ * The list of HTML elements which are supposed not to have
+ * CDATA content in a Strict DTD and where a p element will
+ * be implied
+ */
+ static const char *htmlNoContentElementsStrict[] = {
+ "html",
+ "head",
+ "body",
+ NULL
+ };
+
+ /*
* The list of HTML attributes which are of content %Script;
* NOTE: when adding ones, check htmlIsScriptAttribute() since
* it assumes the name starts with 'on'
***************
*** 1132,1137 ****
--- 1144,1171 ----
}
/**
+ * htmlHasStrictDtd:
+ * @ctxt: a HTML parser context
+ *
+ * Does this document have a strict DTD?
+ *
+ * Returns 1 if the document has a strict HTML DTD, 0 otherwise.
+ */
+
+ static int htmlHasStrictDtd(htmlParserCtxtPtr ctxt) {
+ xmlDtdPtr dtd;
+
+ dtd = xmlGetIntSubset(ctxt->myDoc);
+ if (dtd != NULL && dtd->ExternalID != NULL) {
+ if (!xmlStrcasecmp(dtd->ExternalID, BAD_CAST "-//W3C//DTD HTML 4.01//EN") ||
+ !xmlStrcasecmp(dtd->ExternalID, BAD_CAST "-//W3C//DTD HTML 4//EN"))
+ return(1);
+ }
+
+ return(0);
+ }
+
+ /**
* htmlAutoCloseOnClose:
* @ctxt: an HTML parser context
* @newtag: The new tag name
***************
*** 1353,1358 ****
--- 1387,1393 ----
htmlCheckParagraph(htmlParserCtxtPtr ctxt) {
const xmlChar *tag;
int i;
+ xmlChar **elements;
if (ctxt == NULL)
return(-1);
***************
*** 1367,1374 ****
}
if (!htmlOmittedDefaultValue)
return(0);
! for (i = 0; htmlNoContentElements[i] != NULL; i++) {
! if (xmlStrEqual(tag, BAD_CAST htmlNoContentElements[i])) {
htmlAutoClose(ctxt, BAD_CAST"p");
htmlCheckImplied(ctxt, BAD_CAST"p");
htmlnamePush(ctxt, BAD_CAST"p");
--- 1402,1413 ----
}
if (!htmlOmittedDefaultValue)
return(0);
! if (htmlHasStrictDtd(ctxt))
! elements = (xmlChar **)htmlNoContentElementsStrict;
! else
! elements = (xmlChar **)htmlNoContentElements;
! for (i = 0; elements[i] != NULL; i++) {
! if (xmlStrEqual(tag, BAD_CAST elements[i])) {
htmlAutoClose(ctxt, BAD_CAST"p");
htmlCheckImplied(ctxt, BAD_CAST"p");
htmlnamePush(ctxt, BAD_CAST"p");
***************
*** 2041,2047 ****
unsigned int i;
int j;
xmlNodePtr lastChild;
- xmlDtdPtr dtd;
for (j = 0;j < len;j++)
if (!(IS_BLANK_CH(str[j]))) return(0);
--- 2080,2085 ----
***************
*** 2057,2070 ****
/* Only strip CDATA children of the body tag for strict HTML DTDs */
if (xmlStrEqual(ctxt->name, BAD_CAST "body") && ctxt->myDoc != NULL) {
! dtd = xmlGetIntSubset(ctxt->myDoc);
! if (dtd != NULL && dtd->ExternalID != NULL) {
! if (!xmlStrcasecmp(dtd->ExternalID, BAD_CAST "-//W3C//DTD HTML 4.01//EN") ||
! !xmlStrcasecmp(dtd->ExternalID, BAD_CAST "-//W3C//DTD HTML 4//EN"))
! return(1);
! }
}
-
if (ctxt->node == NULL) return(0);
lastChild = xmlGetLastChild(ctxt->node);
while ((lastChild) && (lastChild->type == XML_COMMENT_NODE))
--- 2095,2103 ----
/* Only strip CDATA children of the body tag for strict HTML DTDs */
if (xmlStrEqual(ctxt->name, BAD_CAST "body") && ctxt->myDoc != NULL) {
! if (htmlHasStrictDtd(ctxt))
! return(1);
}
if (ctxt->node == NULL) return(0);
lastChild = xmlGetLastChild(ctxt->node);
while ((lastChild) && (lastChild->type == XML_COMMENT_NODE))
Attachment:
signature.asc
Description: OpenPGP digital signature