Magic-Mirror/gecko-dev
1
0
Fork 0
mirror of https://github.com/mozilla/gecko-dev.git synced 2024-10-12 12:55:46 +00:00
Code Issues Actions Packages Projects Releases Wiki Activity

bug 299036: Fix crash in CNavDTD by moving newline stripping into the tokenizer. This reduces our reliance on the invariants that the tokenizer tries to provide. r+sr=jst a=chofmann

Browse Source
This commit is contained in:
mrbkap%gmail.com 2005-06-29 02:46:45 +00:00
parent 6b6f99f32b
commit 47de12372b
4 changed files with 46 additions and 60 deletions
Show Stats Download Patch File Download Diff File Expand all files Collapse all files

5
parser/htmlparser/public/nsHTMLTokens.h
View File

@ -291,15 +291,14 @@ public:
nsScannerIterator& aEnd);
virtual void Bind(const nsAString& aStr);
nsresult ConsumeCharacterData(PRUnichar aChar,
PRBool aConservativeConsume,
nsresult ConsumeCharacterData(PRBool aConservativeConsume,
PRBool aIgnoreComments,
nsScanner& aScanner,
const nsAString& aEndTagName,
PRInt32 aFlag,
PRBool& aFlushTokens);
nsresult ConsumeParsedCharacterData(PRUnichar aChar,
nsresult ConsumeParsedCharacterData(PRBool aDiscardFirstNewline,
PRBool aConservativeConsume,
nsScanner& aScanner,
const nsAString& aEndTagName,

43
parser/htmlparser/src/CNavDTD.cpp
View File

@ -1017,49 +1017,6 @@ nsresult CNavDTD::DidHandleStartTag(nsIParserNode& aNode,eHTMLTags aChildTag){
}//if
}
break;
case eHTMLTag_textarea:
{
// In HTML, we need to strip the first newline from the textarea's text.
CToken* theNextToken = mTokenizer->PeekToken();
if (theNextToken) {
#ifdef DEBUG
eHTMLTokenTypes theType = eHTMLTokenTypes(theNextToken->GetTokenType());
NS_ASSERTION(eToken_text == theType, "Textareas should always have at "
"least one text token as a child.");
#endif
CTextToken* text = NS_STATIC_CAST(CTextToken*, theNextToken);
const nsSubstring &content = text->GetStringValue();
PRBool chop = PR_FALSE;
if (!content.IsEmpty()) {
nsSubstring::const_iterator start, end;
content.BeginReading(start);
content.EndReading(end);
if (*start == nsCRT::CR) {
++start;
if (start != end && *start == nsCRT::LF) {
++start;
}
chop = PR_TRUE;
}
else if (*start == nsCRT::LF) {
++start;
chop = PR_TRUE;
}
if (chop) {
// XXX See bug 294599 for why the nsAutoString is necessary.
nsAutoString chopped(Substring(start, end));
text->Bind(chopped);
++mLineNumber;
}
}
}
break;
}
default:
break;
}//switch

16
parser/htmlparser/src/nsHTMLTokenizer.cpp
View File

@ -872,8 +872,7 @@ nsresult nsHTMLTokenizer::ConsumeStartTag(PRUnichar aChar,
if (isCDATA) {
// The only tags that consume conservatively are <script> and
// <style>, the rest all consume until the end of the document.
result = textToken->ConsumeCharacterData(0,
theTag==eHTMLTag_script ||
result = textToken->ConsumeCharacterData(theTag==eHTMLTag_script ||
theTag==eHTMLTag_style,
theTag!=eHTMLTag_script,
aScanner,
@ -888,12 +887,13 @@ nsresult nsHTMLTokenizer::ConsumeStartTag(PRUnichar aChar,
else if (isPCDATA) {
// Title is consumed conservatively in order to not regress
// bug 42945
result = textToken->ConsumeParsedCharacterData(0,
theTag==eHTMLTag_title,
aScanner,
endTagName,
mFlags,
done);
result = textToken->ConsumeParsedCharacterData(
theTag==eHTMLTag_textarea,
theTag==eHTMLTag_title,
aScanner,
endTagName,
mFlags,
done);
// Note: we *don't* set aFlushTokens here.
}

42
parser/htmlparser/src/nsHTMLTokens.cpp
View File

@ -643,8 +643,7 @@ nsresult CTextToken::Consume(PRUnichar aChar, nsScanner& aScanner,PRInt32 aFlag)
* @param aFlushTokens -- PR_TRUE if we found the terminal tag.
* @return error result
*/
nsresult CTextToken::ConsumeCharacterData(PRUnichar aChar,
PRBool aConservativeConsume,
nsresult CTextToken::ConsumeCharacterData(PRBool aConservativeConsume,
PRBool aIgnoreComments,
nsScanner& aScanner,
const nsAString& aEndTagName,
@ -798,7 +797,7 @@ nsresult CTextToken::ConsumeCharacterData(PRUnichar aChar,
* @param aFlushTokens -- PR_TRUE if we found the terminal tag.
* @return error result
*/
nsresult CTextToken::ConsumeParsedCharacterData(PRUnichar aChar,
nsresult CTextToken::ConsumeParsedCharacterData(PRBool aDiscardFirstNewline,
PRBool aConservativeConsume,
nsScanner& aScanner,
const nsAString& aEndTagName,
@ -845,6 +844,39 @@ nsresult CTextToken::ConsumeParsedCharacterData(PRUnichar aChar,
result = ConsumeUntil(theContent, mNewlineCount, aScanner,
theEndCondition, PR_TRUE, aFlag);
if (aDiscardFirstNewline &&
(NS_SUCCEEDED(result) || !aScanner.IsIncremental()) &&
!(aFlag & NS_IPARSER_FLAG_VIEW_SOURCE)) {
// Check if the very first character is a newline, and if so discard it.
// Note that we don't want to discard it in view source!
// Also note that this has to happen here (as opposed to before the
// ConsumeUntil) because we have to expand any entities.
// XXX It would be nice to be able to do this without calling
// writable()!
const nsSubstring &firstChunk = theContent.str();
if (!firstChunk.IsEmpty()) {
PRUint32 where = 0;
PRUnichar newline = firstChunk.First();
if (newline == kCR || newline == kNewLine) {
++where;
if (firstChunk.Length() > 1) {
if (newline == kCR && firstChunk.CharAt(1) == kNewLine) {
// Handle \r\n = 1 newline.
++where;
}
// Note: \n\r = 2 newlines.
}
}
if (where != 0) {
theContent.writable() = Substring(firstChunk, where);
}
}
}
aDiscardFirstNewline = PR_FALSE;
if (NS_FAILED(result)) {
if (kEOF == result && !aScanner.IsIncremental()) {
aFound = PR_TRUE; // this is as good as it gets.
@ -895,9 +927,7 @@ nsresult CTextToken::ConsumeParsedCharacterData(PRUnichar aChar,
}
}
}
// IE only consumes <!-- --> as comments in PCDATA. We'll accept a bit
// more in quirks mode, but lets ensure that this really is a comment
// start to maintain the illusion of compatability.
// IE only consumes <!-- --> as comments in PCDATA.
if (Distance(currPos, endPos) >= commentStartLen) {
nsScannerIterator start(currPos), end(currPos);
end.advance(commentStartLen);