#if defined(HAVE_CONFIG_H)

#include "config.h"

#endif

#include <ctype.h>

#include <limits.h>

#include <stdio.h>

#include "resip/stack/HeaderTypes.hxx"

#include "resip/stack/SipMessage.hxx"

#include "resip/stack/MsgHeaderScanner.hxx"

#include "rutil/WinLeakCheck.hxx"

namespace resip 

{

///////////////////////////////////////////////////////////////////////////////

//   Any character could be used as the chunk terminating sentinel, as long as

//   it would otherwise be character category "other".  The null character

//   was chosen because it is unlikely to occur naturally -- but it's OK if it

//   does.

enum { chunkTermSentinelChar = '\0' };

enum CharCategoryEnum

{

   ccChunkTermSentinel,

   ccOther,

   ccFieldName,

   ccWhitespace,

   ccColon,

   ccDoubleQuotationMark,

   ccLeftAngleBracket,

   ccRightAngleBracket,

   ccBackslash,

   ccComma,

   ccCarriageReturn,

   ccLineFeed,

   numCharCategories

};

typedef char CharCategory;

char* 

MsgHeaderScanner::allocateBuffer(int size)

{

   return new char[size + MaxNumCharsChunkOverflow];

}

struct CharInfo

{

      CharCategory category;

      MsgHeaderScanner::TextPropBitMask textPropBitMask;

};

static CharInfo charInfoArray[UCHAR_MAX+1];

static inline int c2i(unsigned char c)

{

   return static_cast<int>(c); 

}

static void initCharInfoArray()

{

   for(unsigned int charIndex = 0; charIndex <= UCHAR_MAX; ++charIndex) 

   {

      charInfoArray[charIndex].category = ccOther;

      charInfoArray[charIndex].textPropBitMask = 0;

   }

   for(const char *charPtr = "abcdefghijklmnopqrstuvwxyz"

          "ABCDEFGHIJKLMNOPQRSTUVWXYZ0123456789-.!%*_+`'~";

       *charPtr;

       ++charPtr)

   {

      charInfoArray[c2i(*charPtr)].category = ccFieldName;

   }

   charInfoArray[c2i(' ')].category  = ccWhitespace;

   charInfoArray[c2i('\t')].category = ccWhitespace;

   charInfoArray[c2i(':')].category  = ccColon;

   charInfoArray[c2i('"')].category  = ccDoubleQuotationMark;

   charInfoArray[c2i('<')].category  = ccLeftAngleBracket;

   charInfoArray[c2i('>')].category  = ccRightAngleBracket;

   charInfoArray[c2i('\\')].category  = ccBackslash;

   charInfoArray[c2i(',')].category  = ccComma;

   charInfoArray[c2i('\r')].category = ccCarriageReturn;

   charInfoArray[c2i('\n')].category = ccLineFeed;

   // Assert: "chunkTermSentinelChar"'s category is still the default "ccOther".

   charInfoArray[c2i(chunkTermSentinelChar)].category = ccChunkTermSentinel;

   // Init text property bit masks.

   charInfoArray[c2i('\r')].textPropBitMask =

      MsgHeaderScanner::tpbmContainsLineBreak;

   charInfoArray[c2i('\n')].textPropBitMask =

      MsgHeaderScanner::tpbmContainsLineBreak;

   charInfoArray[c2i(' ')].textPropBitMask =

      MsgHeaderScanner::tpbmContainsWhitespace;

   charInfoArray[c2i('\t')].textPropBitMask =

      MsgHeaderScanner::tpbmContainsWhitespace;

   charInfoArray[c2i('\\')].textPropBitMask =

      MsgHeaderScanner::tpbmContainsBackslash;

   charInfoArray[c2i('%')].textPropBitMask =

      MsgHeaderScanner::tpbmContainsPercent;

   charInfoArray[c2i(';')].textPropBitMask =

      MsgHeaderScanner::tpbmContainsSemicolon;

   charInfoArray[c2i('(')].textPropBitMask =

      MsgHeaderScanner::tpbmContainsParen;

   charInfoArray[c2i(')')].textPropBitMask =

      MsgHeaderScanner::tpbmContainsParen;

}

///////////////////////////////////////////////////////////////////////////////

//   States marked '1' scan normal values.  States marked 'N' scan multi-values.

enum StateEnum

{

   sMsgStart,

   sHalfLineBreakAtMsgStart,

   sScanStatusLine,

   sHalfLineBreakAfterStatusLine,

   sAfterLineBreakAfterStatusLine,

   sScanFieldName,

   sScanWhitespaceAfter1FieldName,

   sScanWhitespaceAfterNFieldName,

   sScanWhitespaceOr1Value,

   sScanWhitespaceOrNValue,

   sHalfLineBreakInWhitespaceBefore1Value,

   sHalfLineBreakInWhitespaceBeforeNValue,

   sAfterLineBreakInWhitespaceBefore1Value,

   sAfterLineBreakInWhitespaceBeforeNValue,

   sScan1Value,

   sScanNValue,

   sHalfLineBreakIn1Value,

   sHalfLineBreakInNValue,

   sAfterLineBreakIn1Value,

   sAfterLineBreakInNValue,

   sScanNValueInQuotes,

   sAfterEscCharInQuotesInNValue,

   sHalfLineBreakInQuotesInNValue,

   sAfterLineBreakInQuotesInNValue,

   sScanNValueInAngles,

   sHalfLineBreakInAnglesInNValue,

   sAfterLineBreakInAnglesInNValue,

   sHalfLineBreakAfterLineBreak,

   numStates

};

typedef char State;

// For each '1' state, the 'N' state is "deltaOfNStateFrom1State" larger.

enum { deltaOfNStateFrom1State = 1 };

/////

enum TransitionActionEnum {

   taNone,

   taTermStatusLine,       // The current character terminates the status

   //     line.

   taTermFieldName,        // The current character terminates a field name.

   //     If the field supports multi-values, shift

   //     the state machine into multi-value scanning.

   taBeyondEmptyValue,     // The current character terminates an empty value.

   //     Implies taStartText.

   taTermValueAfterLineBreak, 

   // The previous two characters are a linebreak

   //      terminating a value.  Implies taStartText.

   taTermValue,            // The current character terminates a value.

   taStartText,            // The current character starts a text unit.

   //     (The status line, a field name, or a value.)

   taEndHeader,            // The current character mEnds_ the header.

   taChunkTermSentinel,    // Either the current character terminates the

   //    current chunk or it is an ordinary character.

   taError                 // The input is erroneous.

};

typedef char TransitionAction;

struct TransitionInfo

{

      TransitionAction  action;

      State             nextState;

};

static TransitionInfo stateMachine[numStates][numCharCategories];

inline void specTransition(State state,

                           CharCategory charCategory,

                           TransitionAction action,

                           State nextState)

{

   stateMachine[c2i(state)][c2i(charCategory)].action = action;

   stateMachine[c2i(state)][c2i(charCategory)].nextState = nextState;

}

static void specDefaultTransition(State state,

                                  TransitionAction action,

                                  State nextState)

{

   for (int charCategory = 0;

        charCategory < numCharCategories;

        ++charCategory) 

   {

      specTransition(state, charCategory, action, nextState);

   }

   specTransition(state, ccCarriageReturn, taError, state);

   specTransition(state, ccLineFeed, taError, state);

   specTransition(state, ccChunkTermSentinel, taChunkTermSentinel, state);

}

static void specHalfLineBreakState(State halfLineBreakState,

                                   State  afterLineBreakState)

{

   specDefaultTransition(halfLineBreakState, taError, halfLineBreakState);

   specTransition(halfLineBreakState, ccLineFeed, taNone, afterLineBreakState);

}

//   Single-value (1) scanning and multi-value (N) scanning involves several nearly

//   identical states.

//   "stateDelta" is either 0 or "deltaOfNStateFrom1State".

static void specXValueStates(int  stateDelta)

{

   specDefaultTransition(sScanWhitespaceAfter1FieldName + stateDelta,

                         taError,

                         sScanWhitespaceAfter1FieldName + stateDelta);

   specTransition(sScanWhitespaceAfter1FieldName + stateDelta,

                  ccWhitespace,

                  taNone,

                  sScanWhitespaceAfter1FieldName + stateDelta);

   specTransition(sScanWhitespaceAfter1FieldName + stateDelta,

                  ccColon,

                  taNone,

                  sScanWhitespaceOr1Value + stateDelta);

   specDefaultTransition(sScanWhitespaceOr1Value + stateDelta,

                         taStartText,

                         sScan1Value + stateDelta);

   specTransition(sScanWhitespaceOr1Value + stateDelta,

                  ccWhitespace,

                  taNone,

                  sScanWhitespaceOr1Value + stateDelta);

   if (stateDelta == deltaOfNStateFrom1State)

   {

      specTransition(sScanWhitespaceOr1Value + stateDelta,

                     ccComma,

                     taError,

                     sScanWhitespaceOr1Value + stateDelta);

      specTransition(sScanWhitespaceOr1Value + stateDelta,

                     ccLeftAngleBracket,

                     taStartText,

                     sScanNValueInAngles);

      specTransition(sScanWhitespaceOr1Value + stateDelta,

                     ccDoubleQuotationMark,

                     taStartText,

                     sScanNValueInQuotes);

   }

   specTransition(sScanWhitespaceOr1Value + stateDelta,

                  ccCarriageReturn,

                  taNone,

                  sHalfLineBreakInWhitespaceBefore1Value + stateDelta);

   specHalfLineBreakState(sHalfLineBreakInWhitespaceBefore1Value + stateDelta,

                          sAfterLineBreakInWhitespaceBefore1Value + stateDelta);

   specDefaultTransition(sAfterLineBreakInWhitespaceBefore1Value + stateDelta,

                         taError,

                         sAfterLineBreakInWhitespaceBefore1Value + stateDelta);

   specTransition(sAfterLineBreakInWhitespaceBefore1Value + stateDelta,

                  ccFieldName,

                  taBeyondEmptyValue,

                  sScanFieldName);

   specTransition(sAfterLineBreakInWhitespaceBefore1Value + stateDelta,

                  ccWhitespace,

                  taNone,

                  sScanWhitespaceOr1Value + stateDelta);

   specTransition(sAfterLineBreakInWhitespaceBefore1Value + stateDelta,

                  ccCarriageReturn,

                  taBeyondEmptyValue,

                  sHalfLineBreakAfterLineBreak);

   specDefaultTransition(sScan1Value + stateDelta,

                         taNone,

                         sScan1Value + stateDelta);

   if (stateDelta == deltaOfNStateFrom1State)

   {

      specTransition(sScan1Value + stateDelta,

                     ccComma,

                     taTermValue,

                     sScanWhitespaceOr1Value + stateDelta);

      specTransition(sScan1Value + stateDelta,

                     ccLeftAngleBracket,

                     taNone,

                     sScanNValueInAngles);

      specTransition(sScan1Value + stateDelta,

                     ccDoubleQuotationMark,

                     taNone,

                     sScanNValueInQuotes);

   }

   specTransition(sScan1Value + stateDelta,

                  ccCarriageReturn,

                  taNone,

                  sHalfLineBreakIn1Value + stateDelta);

   specHalfLineBreakState(sHalfLineBreakIn1Value + stateDelta,

                          sAfterLineBreakIn1Value + stateDelta);

   specDefaultTransition(sAfterLineBreakIn1Value + stateDelta,

                         taError,

                         sAfterLineBreakIn1Value + stateDelta);

   specTransition(sAfterLineBreakIn1Value + stateDelta,

                  ccFieldName,

                  taTermValueAfterLineBreak,

                  sScanFieldName);

   specTransition(sAfterLineBreakIn1Value + stateDelta,

                  ccWhitespace,

                  taNone,

                  sScan1Value + stateDelta);

   specTransition(sAfterLineBreakIn1Value + stateDelta,

                  ccCarriageReturn,

                  taTermValueAfterLineBreak,

                  sHalfLineBreakAfterLineBreak);

}

static void initStateMachine()

{

   // By convention, error transitions maintain the same state.

   specDefaultTransition(sMsgStart, taStartText, sScanStatusLine);

   specTransition(sMsgStart,

                  ccCarriageReturn,

                  taNone,

                  sHalfLineBreakAtMsgStart);

   specTransition(sMsgStart, ccLineFeed, taError, sMsgStart);

   specHalfLineBreakState(sHalfLineBreakAtMsgStart, sMsgStart);

   specDefaultTransition(sScanStatusLine, taNone, sScanStatusLine);

   specTransition(sScanStatusLine,

                  ccCarriageReturn,

                  taTermStatusLine,

                  sHalfLineBreakAfterStatusLine);

   specHalfLineBreakState(sHalfLineBreakAfterStatusLine,

                          sAfterLineBreakAfterStatusLine);

   specDefaultTransition(sAfterLineBreakAfterStatusLine,

                         taError,

                         sAfterLineBreakAfterStatusLine);

   specTransition(sAfterLineBreakAfterStatusLine,

                  ccFieldName,

                  taStartText,

                  sScanFieldName);

   specTransition(sAfterLineBreakAfterStatusLine,

                  ccWhitespace,

                  taError,

                  sAfterLineBreakAfterStatusLine);

   specTransition(sAfterLineBreakAfterStatusLine,

                  ccCarriageReturn,

                  taNone,

                  sHalfLineBreakAfterLineBreak);

   specDefaultTransition(sScanFieldName, taError, sScanFieldName);

   specTransition(sScanFieldName, ccFieldName, taNone, sScanFieldName);

   specTransition(sScanFieldName,

                  ccWhitespace,

                  taTermFieldName,

                  sScanWhitespaceAfter1FieldName);

   specTransition(sScanFieldName,

                  ccColon,

                  taTermFieldName,

                  sScanWhitespaceOr1Value);

   specXValueStates(0);

   specXValueStates(deltaOfNStateFrom1State);

   specDefaultTransition(sScanNValueInQuotes, taNone, sScanNValueInQuotes);

   specTransition(sScanNValueInQuotes,

                  ccDoubleQuotationMark,

                  taNone,

                  sScanNValue);

   specTransition(sScanNValueInQuotes,

                  ccBackslash,

                  taNone,

                  sAfterEscCharInQuotesInNValue);

   specTransition(sScanNValueInQuotes,

                  ccCarriageReturn,

                  taNone,

                  sHalfLineBreakInQuotesInNValue);

   specDefaultTransition(sAfterEscCharInQuotesInNValue,

                         taNone,

                         sScanNValueInQuotes);

   specHalfLineBreakState(sHalfLineBreakInQuotesInNValue,

                          sAfterLineBreakInQuotesInNValue);

   specDefaultTransition(sAfterLineBreakInQuotesInNValue,

                         taError,

                         sAfterLineBreakInQuotesInNValue);

   specTransition(sAfterLineBreakInQuotesInNValue,

                  ccWhitespace,

                  taNone,

                  sScanNValueInQuotes);

   specDefaultTransition(sScanNValueInAngles, taNone, sScanNValueInAngles);

   specTransition(sScanNValueInAngles,

                  ccRightAngleBracket,

                  taNone,

                  sScanNValue);

   specTransition(sScanNValueInAngles,

                  ccCarriageReturn,

                  taNone,

                  sHalfLineBreakInAnglesInNValue);

   specHalfLineBreakState(sHalfLineBreakInAnglesInNValue,

                          sAfterLineBreakInAnglesInNValue);

   specDefaultTransition(sAfterLineBreakInAnglesInNValue,

                         taError,

                         sAfterLineBreakInAnglesInNValue);

   specTransition(sAfterLineBreakInAnglesInNValue,

                  ccWhitespace,

                  taNone,

                  sScanNValueInAngles);

   specHalfLineBreakState(sHalfLineBreakAfterLineBreak, sMsgStart);

   // Most half-line-break states do nothing when they read a line feed,

   // but sHalfLineBreakAfterLineBreak must end the message header scanning.

   specTransition(sHalfLineBreakAfterLineBreak,

                  ccLineFeed,

                  taEndHeader,

                  sMsgStart); // Arbitrary but possibly handy.

}

// Debug follows

#if defined(RESIP_MSG_HEADER_SCANNER_DEBUG)  

static void printText(const char *  text,

                      unsigned int  textLength)

{

   const char *charPtr = text;

   for (unsigned int counter = 0; counter < textLength; ++charPtr, ++counter)

   {

      char c = *charPtr;

      switch (c)

      {

         case '\\': printf("\\\\");

            break;

         case '\r': printf("\\r");

            break;

         case '\n': printf("\\n");

            break;

         case '\t': printf("\\t");

            break;

         case '\0': printf("\\0");

            break;

         default:   putchar(c);

      }

   }

}

static const char *

categorySymbol(CharCategory c)

{

   switch(c)

   {

      case ccChunkTermSentinel: return "TERM";

      case ccOther: return "*";

      case ccFieldName: return "FName";

      case ccWhitespace: return "WS";

      case ccColon: return "\\\":\\\"";

      case ccDoubleQuotationMark: return "\\\"";

      case ccLeftAngleBracket: return "\\\"<\\\"";

      case ccRightAngleBracket: return "\\\">\\\"";

      case ccBackslash: return "\\\"\\\\\\\"";

      case ccComma: return "\\\",\\\"";

      case ccCarriageReturn: return "CR";

      case ccLineFeed: return "LF";

   }

   return "??CC??";

}

static const char *

categoryName(CharCategory c)

{

   switch(c)

   {

      case ccChunkTermSentinel: return "ccChunkTermSentinel";

      case ccOther: return "ccOther";

      case ccFieldName: return "ccFieldName";

      case ccWhitespace: return "ccWhitespace";

      case ccColon: return "ccColon";

      case ccDoubleQuotationMark: return "ccDoubleQuotationMark";

      case ccLeftAngleBracket: return "ccLeftAngleBracket";

      case ccRightAngleBracket: return "ccRightAngleBracket";

      case ccBackslash: return "ccBackslash";

      case ccComma: return "ccComma";

      case ccCarriageReturn: return "ccCarriageReturn";

      case ccLineFeed: return "ccLineFeed";

   }

   return "UNKNOWNCC";

}

static const char *

cleanName(const char * name)

{

   // Remove leading type-noise from name

   static char *leaders[] = {

      "cc",

      "s",

      "taChunkTerm", // hack to make ChunkTermSentinel smaller

      "ta"

   };

   const int nLeaders = sizeof(leaders)/sizeof(*leaders);

   int offset = 0;

   for(int i = 0 ; i < nLeaders ; i++)

   {

      unsigned int l = strlen(leaders[i]);

      if (strstr(name,leaders[i]) == name &&

          strlen(name) > l && 

          isupper(name[l]))

      {

         offset = l;

         break;

      }

   }

   return &name[offset];

}

static const char * 

stateName(State state)

{

   const char *stateName;

   switch (state) 

   {

      case sMsgStart:

         stateName = "sMsgStart";

         break;

      case sHalfLineBreakAtMsgStart:

         stateName = "sHalfLineBreakAtMsgStart";

         break;

      case sScanStatusLine:

         stateName = "sScanStatusLine";

         break;

      case sHalfLineBreakAfterStatusLine:

         stateName = "sHalfLineBreakAfterStatusLine";

         break;

      case sAfterLineBreakAfterStatusLine:

         stateName = "sAfterLineBreakAfterStatusLine";

         break;

      case sScanFieldName:

         stateName = "sScanFieldName";

         break;

      case sScanWhitespaceAfter1FieldName:

         stateName = "sScanWhitespaceAfter1FieldName";

         break;

      case sScanWhitespaceAfterNFieldName:

         stateName = "sScanWhitespaceAfterNFieldName";

         break;

      case sScanWhitespaceOr1Value:

         stateName = "sScanWhitespaceOr1Value";

         break;

      case sScanWhitespaceOrNValue:

         stateName = "sScanWhitespaceOrNValue";

         break;

      case sHalfLineBreakInWhitespaceBefore1Value:

         stateName = "sHalfLineBreakInWhitespaceBefore1Value";

         break;

      case sHalfLineBreakInWhitespaceBeforeNValue:

         stateName = "sHalfLineBreakInWhitespaceBeforeNValue";

         break;

      case sAfterLineBreakInWhitespaceBefore1Value:

         stateName = "sAfterLineBreakInWhitespaceBefore1Value";

         break;

      case sAfterLineBreakInWhitespaceBeforeNValue:

         stateName = "sAfterLineBreakInWhitespaceBeforeNValue";

         break;

      case sScan1Value:

         stateName = "sScan1Value";

         break;

      case sScanNValue:

         stateName = "sScanNValue";

         break;

      case sHalfLineBreakIn1Value:

         stateName = "sHalfLineBreakIn1Value";

         break;

      case sHalfLineBreakInNValue:

         stateName = "sHalfLineBreakInNValue";

         break;

      case sAfterLineBreakIn1Value:

         stateName = "sAfterLineBreakIn1Value";

         break;

      case sAfterLineBreakInNValue:

         stateName = "sAfterLineBreakInNValue";

         break;

      case sScanNValueInQuotes:

         stateName = "sScanNValueInQuotes";

         break;

      case sAfterEscCharInQuotesInNValue:

         stateName = "sAfterEscCharInQuotesInNValue";

         break;

      case sHalfLineBreakInQuotesInNValue:

         stateName = "sHalfLineBreakInQuotesInNValue";

         break;

      case sAfterLineBreakInQuotesInNValue:

         stateName = "sAfterLineBreakInQuotesInNValue";

         break;

      case sScanNValueInAngles:

         stateName = "sScanNValueInAngles";

         break;

      case sHalfLineBreakInAnglesInNValue:

         stateName = "sHalfLineBreakInAnglesInNValue";

         break;

      case sAfterLineBreakInAnglesInNValue:

         stateName = "sAfterLineBreakInAnglesInNValue";

         break;

      case sHalfLineBreakAfterLineBreak:

         stateName = "sHalfLineBreakAfterLineBreak";

         break;

      default:

         stateName = "<unknown>";

   }//switch

   return stateName;

}

static const char *

trActionName(TransitionAction transitionAction)

{  

   const char *transitionActionName;

   switch (transitionAction)

   {

      case taNone:

         transitionActionName = "taNone";

         break;

      case taTermStatusLine:

         transitionActionName = "taTermStatusLine";

         break;

      case taTermFieldName:

         transitionActionName = "taTermFieldName";

         break;

      case taBeyondEmptyValue:

         transitionActionName = "taBeyondEmptyValue";

         break;

      case taTermValueAfterLineBreak:

         transitionActionName = "taTermValueAfterLineBreak";

         break;

      case taTermValue:

         transitionActionName = "taTermValue";

         break;

      case taStartText:

         transitionActionName = "taStartText";

         break;

      case taEndHeader:

         transitionActionName = "taEndHeader";

         break;

      case taChunkTermSentinel:

         transitionActionName = "taChunkTermSentinel";

         break;

      case taError:

         transitionActionName = "taError";

         break;

      default:

         transitionActionName = "<unknown>";

   }

   return transitionActionName;

}

static void

printStateTransition(State state,

                     char character,

                     TransitionAction transitionAction)

{

   printf("                %s['", cleanName(stateName(state)));

   printText(&character, 1);

   printf("']: %s\n", cleanName(trActionName(transitionAction)));

}

#if !defined(RESIP_MSG_HEADER_SCANNER_DEBUG)

static const char* stateName(const char*)

{ return "RECOMPILE_WITH_SCANNER_DEBUG"; }

static const char* trActionName(const char*)

{ return stateName(0); }

#endif

/// START OF MEMBER METHODS

int

MsgHeaderScanner::dumpStateMachine(int fd)

{

   FILE *fp = fdopen(fd,"w");

   if (!fp) 

   {

      fprintf(stderr,"MsgHeaderScanner:: unable to open output file\n");

      return -1;

   }

   // Force instance so things are initialized -- YUCK! 

   MsgHeaderScanner scanner;(void)scanner;

   fprintf(fp,"digraph MsgHeaderScannerFSM {\n");

   fprintf(fp,"\tnode[shape=record\n\t\tfontsize=8\n\t\tfontname=\"Helvetica\"\n\t]\n");

   fprintf(fp,"\tedge [ fontsize=6 fontname=\"Helvetica\"]\n");

   fprintf(fp,"\tgraph [ ratio=0.8\n\t\tfontsize=6 compound=true ]");

   for(int state  = 0 ; state < numStates; ++state)

   {

      fprintf(fp,

              "  %s [ label = \"%d|%s\" ]\n",

              cleanName(stateName(state)),

              state,

              cleanName(stateName(state))

         );

      for(int category = 0 ; category < numCharCategories; ++category)

      {

         // Skip Verbose Error or Empty Transitions

         if (stateMachine[state][category].nextState == state &&

             (stateMachine[state][category].action == taError ||

              stateMachine[state][category].action == taNone

                )) continue;

         fprintf(fp,

                 "    %s -> %s [label=\"%s\\n%s\" ]\n",

                 cleanName(stateName(state)),

                 cleanName(stateName(stateMachine[state][category].nextState)),

                 categorySymbol(category),

                 cleanName(trActionName(stateMachine[state][category].action)));

      }

      fprintf(fp,"\n");

   }

   fprintf(fp,"}\n");

   return 0;

}

#endif //defined(RESIP_MSG_HEADER_SCANNER_DEBUG) 

#if defined(RESIP_MSG_HEADER_SCANNER_DEBUG)  

static const char *const multiValuedFieldNameArray[] = {

   "allow-events",

   "accept-encoding",

   "accept-language",

   "allow",

   "content-language",

   "proxy-require",

   "require",

   "supported",

   "subscription-state",

   "unsupported",

   "security-client",

   "security-server",

   "security-verify",

   "accept",

   "call-info",

   "alert-info",

   "error-info",

   "record-route",

   "route",

   "contact",

   "authorization",

   "proxy-authenticate",

   "proxy-authorization",

   "www-authenticate",

   "via",

   0

};

extern

void

lookupMsgHeaderFieldInfo(

   char *                             fieldName,               //inout

   unsigned int                       *fieldNameLength,        //inout

   MsgHeaderScanner::TextPropBitMask  fieldNameTextPropBitMask,

   int                                *fieldKind,              //out

   bool                               *isMultiValueAllowed)    //out

{

   *isMultiValueAllowed = false;

   const char *const *multiValuedFieldNamePtr = multiValuedFieldNameArray;

   for (;;)

   {

      const char *multiValuedFieldName = *multiValuedFieldNamePtr;

      if (!multiValuedFieldName) 

      {

         break;

      }

      if (strncmp(fieldName, multiValuedFieldName, *fieldNameLength) == 0) 

      {

         *isMultiValueAllowed = true;

         break;

      }

      ++multiValuedFieldNamePtr;

   }//for

}

static

bool

processMsgHeaderStatusLine(

   SipMessage *                       msg,

   char *                             lineText,

   unsigned int                       lineTextLength,

   MsgHeaderScanner::TextPropBitMask  lineTextPropBitMask)

{

   printf("status line: ");

   printText(lineText, lineTextLength);

   printf("\n");

   return true;

}

static

void

processMsgHeaderFieldNameAndValue(

   SipMessage *                       msg,

   int                                fieldKind,

   const char *                       fieldName,

   unsigned int                       fieldNameLength,

   char *                             valueText,

   unsigned int                       valueTextLength,

   MsgHeaderScanner::TextPropBitMask  valueTextPropBitMask)

{

   printText(fieldName, fieldNameLength);

   printf(": [[[[");

   printText(valueText, valueTextLength);

   printf("]]]]\n");

}

#else //!defined(RESIP_MSG_HEADER_SCANNER_DEBUG) } {

//   Determine a field's kind and whether it allows (comma separated) multi-values.

//   "fieldName" is not empty and contains only legal characters.

//   The text in "fieldName" may be canonicalized (eg, translating % escapes),

//   including shrinking it if necessary.

inline void

lookupMsgHeaderFieldInfo(char * fieldName,

                         unsigned int *fieldNameLength,   

                         MsgHeaderScanner::TextPropBitMask fieldNameTextPropBitMask,

                         int *fieldKind,             

                         bool *isMultiValueAllowed)    

{

   //.jacob. Don't ignore fieldNameTextPropBitMask.

   *fieldKind = Headers::getType(fieldName, *fieldNameLength);

   *isMultiValueAllowed =

      Headers::isCommaTokenizing(static_cast<Headers::Type>(*fieldKind));

}

// "lineText" contains no carriage returns and no line feeds.

// Return true on success, false on failure.

inline bool

processMsgHeaderStatusLine(SipMessage * msg,

                           char * lineText,

                           unsigned int lineTextLength,

                           MsgHeaderScanner::TextPropBitMask lineTextPropBitMask)

{

   //.jacob. Don't ignore valueTextPropBitMask, and don't always return true.

   msg->setStartLine(lineText, lineTextLength);

   return true;

}

// This function is called once for a field with one value.  (The value could be

// several values, but separated by something other than commas.)

// This function is called once for a field with 0 comma-separated values, with

// an empty value.

// This function is called N times for a field with N comma-separated values,

// but with the same value of "fieldName" each time.

// "fieldName" is not empty and contains only legal characters.

// "valueText" may be empty, has no leading whitespace, may contain trailing

// whitespace, contains carriage returns and line feeds only in correct pairs

// and followed by whitespace, and, if the field is multi-valued, contains

// balanced '<'/'>' and '"' pairs, contains ',' only within '<'/'>' or '"'

// pairs, and respects '\\'s within '"' pairs.

// The text in "valueText" may be canonicalized (eg, translating % escapes),

// including shrinking it if necessary.

inline void

processMsgHeaderFieldNameAndValue(SipMessage * msg,

                                  int fieldKind,

                                  const char * fieldName,

                                  unsigned int fieldNameLength,

                                  char * valueText,

                                  unsigned int valueTextLength,

                                  MsgHeaderScanner::TextPropBitMask valueTextPropBitMask)

{

   //.jacob. Don't ignore valueTextPropBitMask, particularly for '\r' & '\n'.

   msg->addHeader(static_cast<Headers::Type>(fieldKind),

                  fieldName,

                  fieldNameLength,

                  valueText,

                  valueTextLength);

}

#endif //!defined(RESIP_MSG_HEADER_SCANNER_DEBUG) }

bool MsgHeaderScanner::mInitialized = false;

MsgHeaderScanner::MsgHeaderScanner()

{

   if (!mInitialized)

   {

      mInitialized = true;

      initialize();

   }

}

void

MsgHeaderScanner::prepareForMessage(SipMessage *  msg)

{

   mMsg = msg;

   mState = sMsgStart;

   mPrevScanChunkNumSavedTextChars = 0;

   mNumHeaders=0;

}

void

MsgHeaderScanner::prepareForFrag(SipMessage *  msg, bool hasStartLine)

{

   mMsg = msg;

   if (hasStartLine)

   {

      mState = sMsgStart;

   }

   else

   {

      mState = sAfterLineBreakAfterStatusLine;

   }

   mPrevScanChunkNumSavedTextChars = 0;

   mNumHeaders=0;

}

MsgHeaderScanner::ScanChunkResult

MsgHeaderScanner::scanChunk(char * chunk,

                            unsigned int chunkLength,

                            char ** unprocessedCharPtr)

{

   MsgHeaderScanner::ScanChunkResult result;

   CharInfo* localCharInfoArray = charInfoArray;

   TransitionInfo (*localStateMachine)[numCharCategories] = stateMachine;

   State localState = mState;

   char *charPtr = chunk + mPrevScanChunkNumSavedTextChars;

   char *termCharPtr = chunk + chunkLength;

   char saveChunkTermChar = *termCharPtr;

   *termCharPtr = chunkTermSentinelChar;

   char *textStartCharPtr;

   MsgHeaderScanner::TextPropBitMask localTextPropBitMask = mTextPropBitMask;

   if (mPrevScanChunkNumSavedTextChars == 0)

   {

      textStartCharPtr = 0;

   }

   else

   {

      textStartCharPtr = chunk;

   }

   --charPtr;  // The loop starts by advancing "charPtr", so pre-adjust it.

   for (;;)

   {

      // BEGIN message header character scan block BEGIN

      // The code in this block is executed once per message header character.

      // This entire file is designed specifically to minimize this block's size.

      ++charPtr;

      CharInfo *charInfo = &localCharInfoArray[((unsigned char) (*charPtr))];

      CharCategory charCategory = charInfo->category;

      localTextPropBitMask |= charInfo->textPropBitMask;

     determineTransitionFromCharCategory:

      TransitionInfo *transitionInfo =

         &(localStateMachine[(unsigned)localState][(size_t)charCategory]);

      TransitionAction transitionAction = transitionInfo->action;

#if defined(RESIP_MSG_HEADER_SCANNER_DEBUG)  

      printStateTransition(localState, *charPtr, transitionAction);

#endif

      localState = transitionInfo->nextState;

      if (transitionAction == taNone) continue;

      // END message header character scan block END

      // The loop remainder is executed about 4-5 times per message header line.

      switch (transitionAction)

      {

         case taTermStatusLine:

            if (!processMsgHeaderStatusLine(mMsg,

                                            textStartCharPtr,

                                            (unsigned int)(charPtr - textStartCharPtr),

                                            localTextPropBitMask))

            {

               result = MsgHeaderScanner::scrError;

               *unprocessedCharPtr = charPtr;

               goto endOfFunction;

            }

            textStartCharPtr = 0;

            break;

         case taTermFieldName:

         {

            mFieldNameLength = (unsigned int)(charPtr - textStartCharPtr);

            bool isMultiValueAllowed;

            lookupMsgHeaderFieldInfo(textStartCharPtr,

                                     &mFieldNameLength,

                                     localTextPropBitMask,

                                     &mFieldKind,

                                     &isMultiValueAllowed);

            mFieldName = textStartCharPtr;

            textStartCharPtr = 0;

            if (isMultiValueAllowed) 

            {

               localState += deltaOfNStateFrom1State;

            }

         }

         break;

         case taBeyondEmptyValue:

            processMsgHeaderFieldNameAndValue(mMsg,

                                              mFieldKind,

                                              mFieldName,

                                              mFieldNameLength,

                                              0,

                                              0,

                                              0);

            ++mNumHeaders;

            goto performStartTextAction;

         case taTermValueAfterLineBreak:

            processMsgHeaderFieldNameAndValue(mMsg,

                                              mFieldKind,

                                              mFieldName,

                                              mFieldNameLength,

                                              textStartCharPtr,

                                              (unsigned int)((charPtr - textStartCharPtr) - 2),

                                              localTextPropBitMask);       //^:CRLF

            ++mNumHeaders;

            goto performStartTextAction;

         case taTermValue:

            processMsgHeaderFieldNameAndValue(mMsg,

                                              mFieldKind,

                                              mFieldName,

                                              mFieldNameLength,

                                              textStartCharPtr,

                                              (unsigned int)(charPtr - textStartCharPtr),

                                              localTextPropBitMask);

            textStartCharPtr = 0;

            ++mNumHeaders;

            break;

         case taStartText:

        performStartTextAction:

            textStartCharPtr = charPtr;

            localTextPropBitMask = 0;

            break;

         case taEndHeader:

            // textStartCharPtr is not 0.  Not currently relevant.

            result = MsgHeaderScanner::scrEnd;

            *unprocessedCharPtr = charPtr + 1;  // The current char is processed.

            goto endOfFunction;

            break;

         case taChunkTermSentinel:

            if (charPtr == termCharPtr)

            {

               // The chunk has been consumed.  Save some state and request another.

               mState = localState;

               if (textStartCharPtr == 0) 

               {

                  mPrevScanChunkNumSavedTextChars = 0;

               }

               else

               {

                  mPrevScanChunkNumSavedTextChars = (unsigned int)(termCharPtr - textStartCharPtr);

               }

               mTextPropBitMask = localTextPropBitMask;

               result = MsgHeaderScanner::scrNextChunk;

               *unprocessedCharPtr = termCharPtr - mPrevScanChunkNumSavedTextChars;

               goto endOfFunction;

            }

            else

            {

               // The character is not the sentinel.  Treat it like any other.

               charCategory = ccOther;

               goto determineTransitionFromCharCategory;

            }

            break;

         default:

            result = MsgHeaderScanner::scrError;

            *unprocessedCharPtr = charPtr;

            goto endOfFunction;

      }//switch

   }//for

  endOfFunction:

   *termCharPtr = saveChunkTermChar;

   return result;

}

bool

MsgHeaderScanner::initialize()

{

   initCharInfoArray();

   initStateMachine();

   return true;

}