--- parser3/src/main/pa_charset.C	2003/09/29 09:42:12	1.38
+++ parser3/src/main/pa_charset.C	2004/02/03 13:09:52	1.46
@@ -5,7 +5,7 @@
 	Author: Alexander Petrosyan<paf@design.ru>(http://paf.design.ru)
 */
 
-static const char* IDENT_CHARSET_C="$Date: 2003/09/29 09:42:12 $";
+static const char * const IDENT_CHARSET_C="$Date: 2004/02/03 13:09:52 $";
 
 #include "pa_charset.h"
 #include "pa_charsets.h"
@@ -14,6 +14,8 @@ static const char* IDENT_CHARSET_C="$Dat
 #include "libxml/encoding.h"
 #endif
 
+//#define PA_PATCHED_LIBXML_BACKWARD
+
 // globals
 
 Charset::UTF8CaseTable::Rec UTF8CaseToUpperRecords[]={
@@ -117,10 +119,6 @@ void Charset::load_definition(Request_ch
 
 	// charset
 	memset(&tables, 0, sizeof(tables));
-	// strangly vital
-	tables.toTable[tables.toTableSize].intCh=0;
-	tables.toTable[tables.toTableSize].extCh=(XMLByte)0;
-	tables.toTableSize++;
 
 	// loading text
 	char *data=file_read_text(charsets, afile_spec);
@@ -130,7 +128,7 @@ void Charset::load_definition(Request_ch
 
 	// parse cells
 	char *row;
-	while(row=getrow(&data)) {
+	while((row=getrow(&data))) {
 		// remove empty&comment lines
 		if(!*row || *row=='#')
 			continue;
@@ -138,7 +136,7 @@ void Charset::load_definition(Request_ch
 		// char	white-space	digit	hex-digit	letter	word	lowercase	unicode1	unicode2	
 		unsigned int c=0;
 		char *cell;
-		for(int column=0; cell=lsplit(&row, '\t'); column++) {
+		for(int column=0; (cell=lsplit(&row, '\t')); column++) {
 			switch(column) {
 			case 0: c=to_wchar_code(cell); break;
 			// pcre_tables
@@ -192,22 +190,20 @@ void Charset::sort_ToTable() {
 static XMLByte xlatOneTo(const XMLCh toXlat,
 			 const Charset::Tables& tables,
 			 XMLByte not_found) {
-	unsigned int    lowOfs = 0;
-	unsigned int    hiOfs = tables.toTableSize - 1;
-	do {
+	int    lo = 0;
+	int    hi = tables.toTableSize - 1;
+	while(lo<=hi) {
 		// Calc the mid point of the low and high offset.
-		const unsigned int midOfs =((hiOfs - lowOfs) / 2)+lowOfs;
-		
-		//  If our test char is greater than the mid point char, then
-		//  we move up to the upper half. Else we move to the lower
-		//  half. If its equal, then its our guy.
-		if(toXlat>tables.toTable[midOfs].intCh)
-			lowOfs = midOfs;
-		else if(toXlat<tables.toTable[midOfs].intCh)
-			hiOfs = midOfs;
+		const unsigned int i = (lo + hi) / 2;
+
+		XMLCh cur=tables.toTable[i].intCh;
+		if(toXlat==cur)
+			return tables.toTable[i].extCh;
+		if(toXlat>cur)
+			lo = i+1;
 		else
-			return tables.toTable[midOfs].extCh;
-	} while(lowOfs+1<hiOfs);
+			hi = i-1;
+	}
 	
 	return not_found;
 }
@@ -445,7 +441,7 @@ const String::C Charset::transcodeToUTF8
 		(XMLByte *)src.str, src_length,
 		dest_body, dest_length,
 		tables)<0)
-		throw(0, 0,
+		throw Exception(0,
 			0,
 			"Charset::transcodeToUTF8 buffer overflow");
 
@@ -454,23 +450,22 @@ const String::C Charset::transcodeToUTF8
 }
 
 static XMLCh change_case_UTF8(const XMLCh src, const Charset::UTF8CaseTable& table) {
-	unsigned int    lowOfs = 0;
-	unsigned int    hiOfs = table.size - 1;
-	do {
+	int    lo = 0;
+	int    hi = table.size - 1;
+	while(lo<=hi) {
 		// Calc the mid point of the low and high offset.
-		const unsigned int midOfs =((hiOfs - lowOfs) / 2)+lowOfs;
-		
-		//  If our test char is greater than the mid point char, then
-		//  we move up to the upper half. Else we move to the lower
-		//  half. If its equal, then its our guy.
-		if(src>table.records[midOfs].from)
-			lowOfs = midOfs;
-		else if(src<table.records[midOfs].from)
-			hiOfs = midOfs;
+		const unsigned int i = (lo + hi) / 2;
+
+		XMLCh cur=table.records[i].from;
+		if(src==cur)
+			return table.records[i].to;
+		if(src>cur)
+			lo = i+1;
 		else
-			return table.records[midOfs].to;
-	} while(lowOfs+1<hiOfs);
-	
+			hi = i-1;
+	}
+
+	// not found
 	return src;
 }
 
@@ -526,13 +521,19 @@ static void change_case_UTF8(XMLCh src,
 						const Charset::UTF8CaseTable& table) {
 	store_UTF8(change_case_UTF8(src, table), outPtr);
 };
-void change_case_UTF8(const XMLByte* srcData, XMLByte* toFill, 
-							const Charset::UTF8CaseTable& table) {
+void change_case_UTF8(const XMLByte* srcData, size_t srcLen,
+					  XMLByte* toFill, size_t toFillLen,
+					  const Charset::UTF8CaseTable& table) {
 	const XMLByte* srcPtr=srcData;
+	const XMLByte* srcEnd=srcData+srcLen;
 	XMLByte* outPtr=toFill;
+	XMLByte* outEnd=toFill+toFillLen;
+
+	//  We now loop until we either run out of input data, or room to store
+	while ((srcPtr < srcEnd) && (outPtr < outEnd)) {
+		// Get the next leading byte out
+		const XMLByte firstByte =* srcPtr;
 
-	// Get the next leading byte out
-	while (const XMLByte firstByte = *srcPtr) {
 		if(firstByte<= 127) {
 			change_case_UTF8(firstByte, outPtr, table);
 			srcPtr++;
@@ -590,7 +591,7 @@ const String::C Charset::transcodeFromUT
 		(XMLByte *)src.str, src_length,
 		dest_body, dest_length,
 		tables)<0)
-		throw(0, 0,
+		throw Exception(0, 
 			0,
 			"Charset::transcodeFromUTF8 buffer overflow");
 
@@ -625,6 +626,28 @@ const String::C Charset::transcodeToChar
 
 static const Charset::Tables* tables[MAX_CHARSETS];
 
+#ifdef PA_PATCHED_LIBXML_BACKWARD
+
+#define declareXml256ioFuncs(i) \
+	static int xml256CharEncodingInputFunc##i( \
+		unsigned char *out, int *outlen, \
+		const unsigned char *in, int *inlen, void*) { \
+		return transcodeToUTF8( \
+			in, *(size_t*)inlen, \
+			out, *(size_t*)outlen, \
+			*tables[i]); \
+	} \
+	static int xml256CharEncodingOutputFunc##i( \
+		unsigned char *out, int *outlen, \
+		const unsigned char *in, int *inlen, void*) { \
+		return transcodeFromUTF8( \
+			in, *(size_t*)inlen, \
+			out, *(size_t*)outlen, \
+			*tables[i]); \
+	}
+
+#else
+
 #define declareXml256ioFuncs(i) \
 	static int xml256CharEncodingInputFunc##i( \
 		unsigned char *out, int *outlen, \
@@ -643,6 +666,9 @@ static const Charset::Tables* tables[MAX
 			*tables[i]); \
 	}
 
+#endif
+
+
 declareXml256ioFuncs(0)	declareXml256ioFuncs(1)
 declareXml256ioFuncs(2)	declareXml256ioFuncs(3)
 declareXml256ioFuncs(4)	declareXml256ioFuncs(5)
@@ -672,7 +698,7 @@ void Charset::addEncoding(char *name_cst
 			"already allocated %d handlers, no space for new encoding '%s'",
 				MAX_CHARSETS, name_cstr);
 
-	xmlCharEncodingHandler* handler=new(PointerFreeGC) xmlCharEncodingHandler;
+	xmlCharEncodingHandler* handler=new(UseGC) xmlCharEncodingHandler;
 	{
 		handler->name=name_cstr;
 		handler->input=inputFuncs[handlers_count]; 
@@ -713,7 +739,11 @@ String::C Charset::transcode_cstr(xmlCha
 	if(xmlCharEncodingOutputFunc output=transcoder(FNAME).output) {
 		error=output(
 			(unsigned char*)out, &outlen,
-			(const unsigned char*)s, &inlen);
+			(const unsigned char*)s, &inlen
+#ifdef PA_PATCHED_LIBXML_BACKWARD
+			,0
+#endif
+			);
 	} else {
 		memcpy(out, s, outlen=inlen);
 		error=0;
@@ -752,10 +782,14 @@ xmlChar* Charset::transcode_buf2xchar(co
 #ifndef NDEBUG
 		saved_outlen=outlen;
 #endif
-		out=(xmlChar*)xmlMalloc(outlen+1);
+		out=(xmlChar*)g_malloc(outlen+1);
 		error=input(
 			out, &outlen,
-			(const unsigned char*)buf, (int*)&buf_size);
+			(const unsigned char*)buf, (int*)&buf_size
+#ifdef PA_PATCHED_LIBXML_BACKWARD
+			,0
+#endif
+			);
 	} else {
 		outlen=buf_size;
 #ifndef NDEBUG
@@ -825,7 +859,7 @@ struct Transcode_pair_info {
 	const Charset* dest_transcoder;
 };
 #endif
-static void transcode_pair(const String::Body akey, 
+static void transcode_pair(const String::Body /*akey*/, 
 			 String::Body& avalue, 
 			 Transcode_pair_info* info) {
 	avalue=Charset::transcode(avalue,