WFXMLScanner.cpp

                    -1
                    , attNameRawBuf
                    , XMLUni::fgZeroLenString
                    , fAttValueBuf.getRawBuffer()
                );
                curAtt->setSpecified(true);
                fAttrNameHashList->setElementAt(attNameHash, attCount);
            }
            attCount++;

            // And jump back to the top of the loop
            continue;
        }

        //  It was some special case character so do all of the checks and
        //  deal with it.
        if (!nextCh)
            ThrowXML(UnexpectedEOFException, XMLExcepts::Gen_UnexpectedEOF);

        if (nextCh == chForwardSlash)
        {
            fReaderMgr.getNextChar();
            isEmpty = true;
            if (!fReaderMgr.skippedChar(chCloseAngle))
                emitError(XMLErrs::UnterminatedStartTag, qnameRawBuf);
            break;
        }
        else if (nextCh == chCloseAngle)
        {
            fReaderMgr.getNextChar();
            break;
        }
        else if (nextCh == chOpenAngle)
        {
            //  Check for this one specially, since its going to be common
            //  and it is kind of auto-recovering since we've already hit the
            //  next open bracket, which is what we would have seeked to (and
            //  skipped this whole tag.)
            emitError(XMLErrs::UnterminatedStartTag, elemDecl->getFullName());
            break;
        }
        else if ((nextCh == chSingleQuote) || (nextCh == chDoubleQuote))
        {
            //  Check for this one specially, which is probably a missing
            //  attribute name, e.g. ="value". Just issue expected name
            //  error and eat the quoted string, then jump back to the
            //  top again.
            emitError(XMLErrs::ExpectedAttrName);
            fReaderMgr.getNextChar();
            fReaderMgr.skipQuotedString(nextCh);
            fReaderMgr.skipPastSpaces();
            continue;
        }
    }

    //  If empty, validate content right now if we are validating and then
    //  pop the element stack top. Else, we have to update the current stack
    //  top's namespace mapping elements.
    if (isEmpty)
    {
        // Pop the element stack back off since it'll never be used now
        fElemStack.popTop();

        // If the elem stack is empty, then it was an empty root
        if (isRoot)
            gotData = false;
    }

    //  If we have a document handler, then tell it about this start tag. We
    //  don't have any URI id to send along, so send fEmptyNamespaceId. We also do not send
    //  any prefix since its just one big name if we are not doing namespaces.
    if (fDocHandler)
    {
        fDocHandler->startElement
        (
            *elemDecl
            , fEmptyNamespaceId
            , 0
            , *fAttrList
            , attCount
            , isEmpty
            , isRoot
        );
    }

    return true;
}


//  This method is called to scan a start tag when we are processing
//  namespaces. There are two different versions of this method, one for
//  namespace aware processing an done for non-namespace aware processing.
//
//  This method is called after we've scanned the < of a start tag. So we
//  have to get the element name, then scan the attributes, after which
//  we are either going to see >, />, or attributes followed by one of those
//  sequences.
bool WFXMLScanner::scanStartTagNS(bool& gotData)
{
    //  Assume we will still have data until proven otherwise. It will only
    //  ever be false if this is the root and its empty.
    gotData = true;

    //  The current position is after the open bracket, so we need to read in
    //  in the element name.
    if (!fReaderMgr.getName(fQNameBuf))
    {
        emitError(XMLErrs::ExpectedElementName);
        fReaderMgr.skipToChar(chOpenAngle);
        return false;
    }

    // See if its the root element
    const bool isRoot = fElemStack.isEmpty();

	// Assume it won't be an empty tag
    bool isEmpty = false;

    // Skip any whitespace after the name
    fReaderMgr.skipPastSpaces();

    //  Lets try to look up the element
    const XMLCh* qnameRawBuf = fQNameBuf.getRawBuffer();
    XMLElementDecl* elemDecl = fElementLookup->get(qnameRawBuf);

    if (!elemDecl) {
        if (!XMLString::compareNString(qnameRawBuf, XMLUni::fgXMLNSColonString, 6))
            emitError(XMLErrs::NoXMLNSAsElementPrefix, qnameRawBuf);

        if (fElementIndex < fElements->size()) {
            elemDecl = fElements->elementAt(fElementIndex);
        }
        else {
            elemDecl = new DTDElementDecl();
            fElements->addElement(elemDecl);
        }

        elemDecl->setElementName(qnameRawBuf, fEmptyNamespaceId);
        fElementLookup->put((void*)elemDecl->getFullName(), elemDecl);
        fElementIndex++;
    }

    // Expand the element stack and add the new element
    fElemStack.addLevel(elemDecl, fReaderMgr.getCurrentReaderNum());

    // reset NS attribute list
    fAttrNSList->removeAllElements();

    // We loop until we either see a /> or >, handling attribute/value
    // pairs until we get there.
    unsigned int attCount = 0;
    unsigned int curAttListSize = fAttrList->size();
    while (true)
    {
        // And get the next non-space character
        XMLCh nextCh = fReaderMgr.peekNextChar();

        //  If the next character is not a slash or closed angle bracket,
        //  then it must be whitespace, since whitespace is required
        //  between the end of the last attribute and the name of the next
        //  one.
        if (attCount)
        {
            if ((nextCh != chForwardSlash) && (nextCh != chCloseAngle))
            {
                if (fReaderMgr.getCurrentReader()->isWhitespace(nextCh))
                {
                    // Ok, skip by them and peek another char
                    fReaderMgr.skipPastSpaces();
                    nextCh = fReaderMgr.peekNextChar();
                }
                else
                {
                    // Emit the error but keep on going
                    emitError(XMLErrs::ExpectedWhitespace);
                }
            }
        }

        //  Ok, here we first check for any of the special case characters.
        //  If its not one, then we do the normal case processing, which
        //  assumes that we've hit an attribute value, Otherwise, we do all
        //  the special case checks.
        if (!fReaderMgr.getCurrentReader()->isSpecialStartTagChar(nextCh))
        {
            //  Assume its going to be an attribute, so get a name from
            //  the input.
            if (!fReaderMgr.getName(fAttNameBuf))
            {
                emitError(XMLErrs::ExpectedAttrName);
                fReaderMgr.skipPastChar(chCloseAngle);
                return false;
            }

            // And next must be an equal sign
            if (!scanEq())
            {
                static const XMLCh tmpList[] =
                {
                    chSingleQuote, chDoubleQuote, chCloseAngle
                    , chOpenAngle, chForwardSlash, chNull
                };

                emitError(XMLErrs::ExpectedEqSign);

                //  Try to sync back up by skipping forward until we either
                //  hit something meaningful.
                const XMLCh chFound = fReaderMgr.skipUntilInOrWS(tmpList);

                if ((chFound == chCloseAngle) || (chFound == chForwardSlash))
                {
                    // Jump back to top for normal processing of these
                    continue;
                }
                else if ((chFound == chSingleQuote)
                      ||  (chFound == chDoubleQuote)
                      ||  fReaderMgr.getCurrentReader()->isWhitespace(chFound))
                {
                    // Just fall through assuming that the value is to follow
                }
                else if (chFound == chOpenAngle)
                {
                    // Assume a malformed tag and that new one is starting
                    emitError(XMLErrs::UnterminatedStartTag, qnameRawBuf);
                    return false;
                }
                else
                {
                    // Something went really wrong
                    return false;
                }
            }

            //  See if this attribute is declared more than one for this element.
            const XMLCh* attNameRawBuf = fAttNameBuf.getRawBuffer();
            unsigned int attNameHash = XMLString::hash(attNameRawBuf, 109);
            if (attCount) {

                for (unsigned int k=0; k < attCount; k++) {

                    if (fAttrNameHashList->elementAt(k) == attNameHash) {
                        if (XMLString::equals(
                                fAttrList->elementAt(k)->getQName()
                                , attNameRawBuf))
                        {
                            emitError
                            (
                                XMLErrs::AttrAlreadyUsedInSTag
                                , attNameRawBuf
                                , qnameRawBuf
                            );
                            break;
                        }
                    }
                }
            }

            //  Skip any whitespace before the value and then scan the att
            //  value. This will come back normalized with entity refs and
            //  char refs expanded.
            fReaderMgr.skipPastSpaces();
            if (!scanAttValue(attNameRawBuf, fAttValueBuf))
            {
                static const XMLCh tmpList[] =
                {
                    chCloseAngle, chOpenAngle, chForwardSlash, chNull
                };

                emitError(XMLErrs::ExpectedAttrValue);

                //  It failed, so lets try to get synced back up. We skip
                //  forward until we find some whitespace or one of the
                //  chars in our list.
                const XMLCh chFound = fReaderMgr.skipUntilInOrWS(tmpList);

                if ((chFound == chCloseAngle)
                ||  (chFound == chForwardSlash)
                ||  fReaderMgr.getCurrentReader()->isWhitespace(chFound))
                {
                    //  Just fall through and process this attribute, though
                    //  the value will be "".
                }
                else if (chFound == chOpenAngle)
                {
                    // Assume a malformed tag and that new one is starting
                    emitError(XMLErrs::UnterminatedStartTag, qnameRawBuf);
                    return false;
                }
                else
                {
                    // Something went really wrong
                    return false;
                }
            }

            //  Add this attribute to the attribute list that we use to
            //  pass them to the handler. We reuse its existing elements
            //  but expand it as required.
            const XMLCh* attValueRawBuf = fAttValueBuf.getRawBuffer();
            XMLAttr* curAtt = 0;
            if (attCount >= curAttListSize)
            {
                curAtt = new XMLAttr
                (
                    fEmptyNamespaceId
                    , attNameRawBuf
                    , attValueRawBuf
                );
                fAttrList->addElement(curAtt);
                fAttrNameHashList->addElement(attNameHash);
            }
            else
            {
                curAtt = fAttrList->elementAt(attCount);
                curAtt->set
                (
                    fEmptyNamespaceId
                    , attNameRawBuf
                    , attValueRawBuf
                );
                curAtt->setSpecified(true);
                fAttrNameHashList->setElementAt(attNameHash, attCount);
            }

            // Make sure that the name is basically well formed for namespace
            //  enabled rules. It either has no colons, or it has one which
            //  is neither the first or last char.
            const int colonFirst = XMLString::indexOf(attNameRawBuf, chColon);
            if (colonFirst != -1)
            {
                const int colonLast = XMLString::lastIndexOf(attNameRawBuf, chColon);

                if (colonFirst != colonLast)
                {
                    emitError(XMLErrs::TooManyColonsInName);
                    continue;
                }
                else if ((colonFirst == 0)
                      ||  (colonLast == (int)fAttNameBuf.getLen() - 1))
                {
                    emitError(XMLErrs::InvalidColonPos);
                    continue;
                }
            }

            // Map prefix to namespace
            const XMLCh* attPrefix = curAtt->getPrefix();
            const XMLCh* attLocalName = curAtt->getName();
            const XMLCh* namespaceURI = fAttValueBuf.getRawBuffer();

            if (attPrefix && *attPrefix) {
                if (XMLString::equals(attPrefix, XMLUni::fgXMLString)) {
                    curAtt->setURIId(fXMLNamespaceId);
                }
                else if (XMLString::equals(attPrefix, XMLUni::fgXMLNSString)) {

                    if (XMLString::equals(attLocalName, XMLUni::fgXMLNSString))
                        emitError(XMLErrs::NoUseOfxmlnsAsPrefix);
                    else if (XMLString::equals(attLocalName, XMLUni::fgXMLString)) {
                        if (!XMLString::equals(namespaceURI, XMLUni::fgXMLURIName))
                            emitError(XMLErrs::PrefixXMLNotMatchXMLURI);
                    }

                    if (!namespaceURI)
                        emitError(XMLErrs::NoEmptyStrNamespace, attNameRawBuf);
                    else if(!*namespaceURI && fXMLVersion == XMLReader::XMLV1_0)
                        emitError(XMLErrs::NoEmptyStrNamespace, attNameRawBuf);

                    fElemStack.addPrefix
                    (
                        attLocalName
                        , fURIStringPool->addOrFind(namespaceURI)
                    );
                    curAtt->setURIId(fXMLNSNamespaceId);
                }
                else {
                    fAttrNSList->addElement(curAtt);
                }
            }
            else {
                if (XMLString::equals(XMLUni::fgXMLNSString, attLocalName)) {

                    if (XMLString::equals(namespaceURI, XMLUni::fgXMLNSURIName))
                        emitError(XMLErrs::NoUseOfxmlnsURI);
                    else if (XMLString::equals(namespaceURI, XMLUni::fgXMLURIName))
                        emitError(XMLErrs::XMLURINotMatchXMLPrefix);

                    fElemStack.addPrefix
                    (
                        XMLUni::fgZeroLenString
                        , fURIStringPool->addOrFind(namespaceURI)
                    );
                }
            }

            // increment attribute count
            attCount++;

            // And jump back to the top of the loop
            continue;
        }

        //  It was some special case character so do all of the checks and
        //  deal with it.
        if (!nextCh)
            ThrowXML(UnexpectedEOFException, XMLExcepts::Gen_UnexpectedEOF);

        if (nextCh == chForwardSlash)
        {
            fReaderMgr.getNextChar();
            isEmpty = true;
            if (!fReaderMgr.skippedChar(chCloseAngle))
                emitError(XMLErrs::UnterminatedStartTag, qnameRawBuf);
            break;
        }
        else if (nextCh == chCloseAngle)
        {
            fReaderMgr.getNextChar();
            break;
        }
        else if (nextCh == chOpenAngle)
        {
            //  Check for this one specially, since its going to be common
            //  and it is kind of auto-recovering since we've already hit the
            //  next open bracket, which is what we would have seeked to (and
            //  skipped this whole tag.)
            emitError(XMLErrs::UnterminatedStartTag, qnameRawBuf);
            break;
        }
        else if ((nextCh == chSingleQuote) || (nextCh == chDoubleQuote))
        {
            //  Check for this one specially, which is probably a missing
            //  attribute name, e.g. ="value". Just issue expected name
            //  error and eat the quoted string, then jump back to the
            //  top again.
            emitError(XMLErrs::ExpectedAttrName);
            fReaderMgr.getNextChar();
            fReaderMgr.skipQuotedString(nextCh);
            fReaderMgr.skipPastSpaces();
            continue;
        }
    }

    // Handle provided attributes that we did not map their prefixes
    for (unsigned int i=0; i < fAttrNSList->size(); i++) {

        XMLAttr* providedAttr = fAttrNSList->elementAt(i);

        providedAttr->setURIId
        (
	        resolvePrefix
            (
                providedAttr->getPrefix(),
                ElemStack::Mode_Attribute
            )
        );
    }

    // Resolve the qualified name to a URI.
    unsigned int uriId = resolvePrefix
    (
        elemDecl->getElementName()->getPrefix()
        , ElemStack::Mode_Element
    );

    // Now we can update the element stack
    fElemStack.setCurrentURI(uriId);

    // Tell the document handler about this start tag
    if (fDocHandler)
    {
        fDocHandler->startElement
        (
            *elemDecl
            , uriId
            , elemDecl->getElementName()->getPrefix()
            , *fAttrList
            , attCount
            , false
            , isRoot
        );
    }

    //  If empty, validate content right now if we are validating and then
    //  pop the element stack top. Else, we have to update the current stack
    //  top's namespace mapping elements.
    if (isEmpty)
    {
        // Pop the element stack back off since it'll never be used now
        fElemStack.popTop();

        // If we have a doc handler, tell it about the end tag
        if (fDocHandler)
        {
            fDocHandler->endElement
            (
                *elemDecl
                , uriId
                , isRoot
                , elemDecl->getElementName()->getPrefix()
            );
        }

        // If the elem stack is empty, then it was an empty root
        if (isRoot)
            gotData = false;
    }

    return true;
}

unsigned int
WFXMLScanner::resolveQName(const   XMLCh* const qName
                           ,       XMLBuffer&   prefixBuf
                           , const short        mode
                           ,       int&         prefixColonPos)
{
    //  Lets split out the qName into a URI and name buffer first. The URI
    //  can be empty.
    prefixColonPos = XMLString::indexOf(qName, chColon);
    if (prefixColonPos == -1)
    {
        //  Its all name with no prefix, so put the whole thing into the name
        //  buffer. Then map the empty string to a URI, since the empty string
        //  represents the default namespace. This will either return some
        //  explicit URI which the default namespace is mapped to, or the
        //  the default global namespace.
        bool unknown = false;

        prefixBuf.reset();
        return fElemStack.mapPrefixToURI(XMLUni::fgZeroLenString, (ElemStack::MapModes) mode, unknown);
    }
    else
    {
        //  Copy the chars up to but not including the colon into the prefix
        //  buffer.
        prefixBuf.set(qName, prefixColonPos);

        //  Watch for the special namespace prefixes. We always map these to
        //  special URIs. 'xml' gets mapped to the official URI that its defined
        //  to map to by the NS spec. xmlns gets mapped to a special place holder
        //  URI that we define (so that it maps to something checkable.)
        const XMLCh* prefixRawBuf = prefixBuf.getRawBuffer();
        if (XMLString::equals(prefixRawBuf, XMLUni::fgXMLNSString)) {

            // if this is an element, it is an error to have xmlns as prefix
            if (mode == ElemStack::Mode_Element)
                emitError(XMLErrs::NoXMLNSAsElementPrefix, qName);

            return fXMLNSNamespaceId;
        }
        else if (XMLString::equals(prefixRawBuf, XMLUni::fgXMLString)) {
            return  fXMLNamespaceId;
        }
        else
        {
            bool unknown = false;
            unsigned int uriId = fElemStack.mapPrefixToURI(prefixRawBuf, (ElemStack::MapModes) mode, unknown);

            if (unknown)
                emitError(XMLErrs::UnknownPrefix, prefixRawBuf);

            return uriId;
        }
    }
}

// ---------------------------------------------------------------------------
//  XMLScanner: Private parsing methods
// ---------------------------------------------------------------------------
bool WFXMLScanner::scanAttValue(const XMLCh* const attrName
                              ,     XMLBuffer&   toFill)
{
    // Reset the target buffer
    toFill.reset();

    // Get the next char which must be a single or double quote
    XMLCh quoteCh;
    if (!fReaderMgr.skipIfQuote(quoteCh))
        return false;

    //  We have to get the current reader because we have to ignore closing
    //  quotes until we hit the same reader again.
    const unsigned int curReader = fReaderMgr.getCurrentReaderNum();

    //  Loop until we get the attribute value. Note that we use a double
    //  loop here to avoid the setup/teardown overhead of the exception
    //  handler on every round.
    XMLCh   nextCh;
    XMLCh   secondCh = 0;
    bool    firstNonWS = false;
    bool    gotLeadingSurrogate = false;
    bool    escaped;
    while (true)
    {
    try
    {
        while(true)
        {
            nextCh = fReaderMgr.getNextChar();

            if (!nextCh)
                ThrowXML(UnexpectedEOFException, XMLExcepts::Gen_UnexpectedEOF);

            // Check for our ending quote in the same entity
            if (nextCh == quoteCh)
            {
                if (curReader == fReaderMgr.getCurrentReaderNum())
                    return true;

                // Watch for spillover into a previous entity
                if (curReader > fReaderMgr.getCurrentReaderNum())
                {
                    emitError(XMLErrs::PartialMarkupInEntity);
                    return false;
                }
            }

            //  Check for an entity ref now, before we let it affect our
            //  whitespace normalization logic below. We ignore the empty flag
            //  in this one.
            escaped = false;
            if (nextCh == chAmpersand)
            {
                if (scanEntityRef(true, nextCh, secondCh, escaped) != EntityExp_Returned)
                {
                    gotLeadingSurrogate = false;
                    continue;
                }
            }
            else if ((nextCh >= 0xD800) && (nextCh <= 0xDBFF))
            {
                // Deal with surrogate pairs
                //  Its a leading surrogate. If we already got one, then
                //  issue an error, else set leading flag to make sure that
                //  we look for a trailing next time.
                if (gotLeadingSurrogate)
                {
                    emitError(XMLErrs::Expected2ndSurrogateChar);
                }
                else
                    gotLeadingSurrogate = true;
            }
            else
            {
                //  If its a trailing surrogate, make sure that we are
                //  prepared for that. Else, its just a regular char so make
                //  sure that we were not expected a trailing surrogate.
                if ((nextCh >= 0xDC00) && (nextCh <= 0xDFFF))
                {
                    // Its trailing, so make sure we were expecting it
                    if (!gotLeadingSurrogate)
                        emitError(XMLErrs::Unexpected2ndSurrogateChar);
                }
                else
                {
                    //  Its just a char, so make sure we were not expecting a
                    //  trailing surrogate.
                    if (gotLeadingSurrogate) {
                        emitError(XMLErrs::Expected2ndSurrogateChar);
                    }
                    // Its got to at least be a valid XML character
                    else if (!fReaderMgr.getCurrentReader()->isXMLChar(nextCh))
                    {
                        XMLCh tmpBuf[9];
                        XMLString::binToText
                        (
                            nextCh
                            , tmpBuf
                            , 8
                            , 16
                        );
                        emitError(XMLErrs::InvalidCharacterInAttrValue, attrName, tmpBuf);
                    }
                }
                gotLeadingSurrogate = false;
            }

            //  If its not escaped, then make sure its not a < character, which
            //  is not allowed in attribute values.
            if (!escaped) {
                if (nextCh == chOpenAngle)
                    emitError(XMLErrs::BracketInAttrValue, attrName);
                else if (fReaderMgr.getCurrentReader()->isWhitespace(nextCh))
                    nextCh = chSpace;
            }

            // Else add it to the buffer
            toFill.append(nextCh);

            if (secondCh)
               toFill.append(secondCh);
        }
    }
    catch(const EndOfEntityException&)
    {
        // Just eat it and continue.
        gotLeadingSurrogate = false;
        escaped = false;
    }
    }
    return true;
}


//  This method scans a CDATA section. It collects the character into one
//  of the temp buffers and calls the document handler, if any, with the
//  characters. It assumes that the <![CDATA string has been scanned before
//  this call.
void WFXMLScanner::scanCDSection()
{
    //  This is the CDATA section opening sequence, minus the '<' character.
    //  We use this to watch for nested CDATA sections, which are illegal.
    static const XMLCh CDataPrefix[] =
    {
            chBang, chOpenSquare, chLatin_C, chLatin_D, chLatin_A
        ,   chLatin_T, chLatin_A, chOpenSquare, chNull
    };

    static const XMLCh CDataClose[] =
    {
            chCloseSquare, chCloseAngle, chNull
    };

    //  The next character should be the opening square bracket. If not
    //  issue an error, but then try to recover by skipping any whitespace
    //  and checking again.
    if (!fReaderMgr.skippedChar(chOpenSquare))
    {
        emitError(XMLErrs::ExpectedOpenSquareBracket);
        fReaderMgr.skipPastSpaces();

        // If we still don't find it, then give up, else keep going
        if (!fReaderMgr.skippedChar(chOpenSquare))
            return;
    }

    // Get a buffer for this
    XMLBufBid bbCData(&fBufMgr);

    //  We just scan forward until we hit the end of CDATA section sequence.
    //  CDATA is effectively a big escape mechanism so we don't treat markup
    //  characters specially here.
    bool            emittedError = false;
    bool    gotLeadingSurrogate = false;
    while (true)
    {
        const XMLCh nextCh = fReaderMgr.getNextChar();

        // Watch for unexpected end of file
        if (!nextCh)
        {
            emitError(XMLErrs::UnterminatedCDATASection);
            ThrowXML(UnexpectedEOFException, XMLExcepts::Gen_UnexpectedEOF);
        }

        //  If this is a close square bracket it could be our closing
        //  sequence.
        if (nextCh == chCloseSquare && fReaderMgr.skippedString(CDataClose))
        {
            //  make sure we were not expecting a trailing surrogate.
            if (gotLeadingSurrogate)
                emitError(XMLErrs::Expected2ndSurrogateChar);

            // If we have a doc handler, call it
            if (fDocHandler)
            {
                fDocHandler->docCharacters
                (
                    bbCData.getRawBuffer()
                    , bbCData.getLen()
                    , true
                );
            }

            // And we are done
            break;
        }

        //  Make sure its a valid character. But if we've emitted an error
        //  already, don't bother with the overhead since we've already told
        //  them about it.
        if (!emittedError)
        {
            // Deal with surrogate pairs
            if ((nextCh >= 0xD800) && (nextCh <= 0xDBFF))
            {
                //  Its a leading surrogate. If we already got one, then
                //  issue an error, else set leading flag to make sure that
                //  we look for a trailing next time.
                if (gotLeadingSurrogate)
                    emitError(XMLErrs::Expected2ndSurrogateChar);
                else
                    gotLeadingSurrogate = true;
            }
            else
            {
                //  If its a trailing surrogate, make sure that we are
                //  prepared for that. Else, its just a regular char so make
                //  sure that we were not expected a trailing surrogate.
                if ((nextCh >= 0xDC00) && (nextCh <= 0xDFFF))
                {
                    // Its trailing, so make sure we were expecting it
                    if (!gotLeadingSurrogate)
                        emitError(XMLErrs::Unexpected2ndSurrogateChar);
                }
                else
                {
                    //  Its just a char, so make sure we were not expecting a
                    //  trailing surrogate.
                    if (gotLeadingSurrogate)
                        emitError(XMLErrs::Expected2ndSurrogateChar);

                    // Its got to at least be a valid XML character
                    else if (!fReaderMgr.getCurrentReader()->isXMLChar(nextCh))
                    {
                        XMLCh tmpBuf[9];
                        XMLString::binToText
                        (
                            nextCh
                            , tmpBuf
                            , 8
                            , 16
                        );
                        emitError(XMLErrs::InvalidCharacter, tmpBuf);
                        emittedError = true;
                    }
                }
                gotLeadingSurrogate = false;
            }
        }

        // Add it to the buffer
        bbCData.append(nextCh);
    }
}


void WFXMLScanner::scanCharData(XMLBuffer& toUse)
{
    //  We have to watch for the stupid ]]> sequence, which is illegal in
    //  character data. So this is a little state machine that handles that.
    enum States
    {
        State_Waiting
        , State_GotOne
        , State_GotTwo
    };

    // Reset the buffer before we start
    toUse.reset();

    // Turn on the 'throw at end' flag of the reader manager
    ThrowEOEJanitor jan(&fReaderMgr, true);

    //  In order to be more efficient we have to use kind of a deeply nested
    //  set of blocks here. The outer block puts on a try and catches end of
    //  entity exceptions. The inner loop is the per-character loop. If we
    //  put the try inside the inner loop, it would work but would require
    //  the exception handling code setup/teardown code to be invoked for
    //  each character.
    XMLCh   nextCh;
    XMLCh   secondCh = 0;
    States  curState = State_Waiting;
    bool    escaped = false;
    bool    gotLeadingSurrogate = false;
    bool    notDone = true;
    while (notDone)
    {
        try
        {
            while (true)
            {
                //  Eat through as many plain content characters as possible without
                //  needing special handling.  Moving most content characters here,
                //  in this one call, rather than running the overall loop once
                //  per content character, is a speed optimization.
                if (curState == State_Waiting  &&  !gotLeadingSurrogate)
                {
                     fReaderMgr.movePlainContentChars(toUse);
                }

                // Try to get another char from the source
                //   The code from here on down covers all contengencies,
                if (!fReaderMgr.getNextCharIfNot(chOpenAngle, nextCh))
                {
                    // If we were waiting for a trailing surrogate, its an error
                    if (gotLeadingSurrogate)
                        emitError(XMLErrs::Expected2ndSurrogateChar);

                    notDone = false;
                    break;
                }

                //  Watch for a reference. Note that the escapement mechanism
                //  is ignored in this content.
                escaped = false;
                if (nextCh == chAmpersand)
                {
                    sendCharData(toUse);

                    // Turn off the throwing at the end of entity during this
                    ThrowEOEJanitor jan(&fReaderMgr, false);

                    if (scanEntityRef(false, nextCh, secondCh, escaped) != EntityExp_Returned)
                    {
                        gotLeadingSurrogate = false;
                        continue;
                    }
                }
                else if ((nextCh >= 0xD800) && (nextCh <= 0xDBFF))
                {
                    // Deal with surrogate pairs
                    //  Its a leading surrogate. If we already got one, then
                    //  issue an error, else set leading flag to make sure that
                    //  we look for a trailing next time.
                    if (gotLeadingSurrogate)
                    {
                        emitError(XMLErrs::Expected2ndSurrogateChar);
                    }
                    else
                        gotLeadingSurrogate = true;
                }
                else
                {
                    //  If its a trailing surrogate, make sure that we are
                    //  prepared for that. Else, its just a regular char so make
                    //  sure that we were not expected a trailing surrogate.
                    if ((nextCh >= 0xDC00) && (nextCh <= 0xDFFF))
                    {
                        // Its trailing, so make sure we were expecting it
                        if (!gotLeadingSurrogate)
                            emitError(XMLErrs::Unexpected2ndSurrogateChar);
                    }
                    else
                    {
                        //  Its just a char, so make sure we were not expecting a
                        //  trailing surrogate.
                        if (gotLeadingSurrogate) {
                            emitError(XMLErrs::Expected2ndSurrogateChar);
                        }
                        // Its got to at least be a valid XML character
                        else if (!fReaderMgr.getCurrentReader()->isXMLChar(nextCh))
                        {
                            XMLCh tmpBuf[9];
                            XMLString::binToText
                            (
                                nextCh
                                , tmpBuf
                                , 8
                                , 16
                            );
                            emitError(XMLErrs::InvalidCharacter, tmpBuf);
                        }
                    }
                    gotLeadingSurrogate = false;
                }

                // Keep the state machine up to date
                if (!escaped)
                {
                    if (nextCh == chCloseSquare)
                    {
                        if (curState == State_Waiting)
                            curState = State_GotOne;
                        else if (curState == State_GotOne)
                            curState = State_GotTwo;
                    }
                    else if (nextCh == chCloseAngle)
                    {
                        if (curState == State_GotTwo)
                            emitError(XMLErrs::BadSequenceInCharData);
                        curState = State_Waiting;
                    }
                    else
                    {
                        curState = State_Waiting;
                    }
                }
                else
                {
                    curState = State_Waiting;
                }

                // Add this char to the buffer
                toUse.append(nextCh);

                if (secondCh)
                    toUse.append(secondCh);
            }
        }
        catch(const EndOfEntityException& toCatch)
        {
            //  Some entity ended, so we have to send any accumulated
            //  chars and send an end of entity event.
            sendCharData(toUse);
            gotLeadingSurrogate = false;

            if (fDocHandler)
                fDocHandler->endEntityReference(toCatch.getEntity());