lib/Lex/MacroArgs.cpp

   1 //===--- TokenLexer.cpp - Lex from a token stream -------------------------===//
   2 //
   3 //                     The LLVM Compiler Infrastructure
   4 //
   5 // This file is distributed under the University of Illinois Open Source
   6 // License. See LICENSE.TXT for details.
   7 //
   8 //===----------------------------------------------------------------------===//
   9 //
  10 // This file implements the TokenLexer interface.
  11 //
  12 //===----------------------------------------------------------------------===//
  13
  14 #include "MacroArgs.h"
  15 #include "clang/Lex/MacroInfo.h"
  16 #include "clang/Lex/Preprocessor.h"
  17 #include "clang/Lex/LexDiagnostic.h"
  18 using namespace clang;
  19
  20 /// MacroArgs ctor function - This destroys the vector passed in.
  21 MacroArgs *MacroArgs::create(const MacroInfo *MI,
  22                              const Token *UnexpArgTokens,
  23                              unsigned NumToks, bool VarargsElided,
  24                              Preprocessor &PP) {
  25   assert(MI->isFunctionLike() &&
  26          "Can't have args for an object-like macro!");
  27   MacroArgs **ResultEnt = 0;
  28   unsigned ClosestMatch = ~0U;
  29
  30   // See if we have an entry with a big enough argument list to reuse on the
  31   // free list.  If so, reuse it.
  32   for (MacroArgs **Entry = &PP.MacroArgCache; *Entry;
  33        Entry = &(*Entry)->ArgCache)
  34     if ((*Entry)->NumUnexpArgTokens >= NumToks &&
  35         (*Entry)->NumUnexpArgTokens < ClosestMatch) {
  36       ResultEnt = Entry;
  37
  38       // If we have an exact match, use it.
  39       if ((*Entry)->NumUnexpArgTokens == NumToks)
  40         break;
  41       // Otherwise, use the best fit.
  42       ClosestMatch = (*Entry)->NumUnexpArgTokens;
  43     }
  44
  45   MacroArgs *Result;
  46   if (ResultEnt == 0) {
  47     // Allocate memory for a MacroArgs object with the lexer tokens at the end.
  48     Result = (MacroArgs*)malloc(sizeof(MacroArgs) + NumToks*sizeof(Token));
  49     // Construct the MacroArgs object.
  50     new (Result) MacroArgs(NumToks, VarargsElided);
  51   } else {
  52     Result = *ResultEnt;
  53     // Unlink this node from the preprocessors singly linked list.
  54     *ResultEnt = Result->ArgCache;
  55     Result->NumUnexpArgTokens = NumToks;
  56     Result->VarargsElided = VarargsElided;
  57   }
  58
  59   // Copy the actual unexpanded tokens to immediately after the result ptr.
  60   if (NumToks)
  61     memcpy(const_cast<Token*>(Result->getUnexpArgument(0)),
  62            UnexpArgTokens, NumToks*sizeof(Token));
  63
  64   return Result;
  65 }
  66
  67 /// destroy - Destroy and deallocate the memory for this object.
  68 ///
  69 void MacroArgs::destroy(Preprocessor &PP) {
  70   StringifiedArgs.clear();
  71
  72   // Don't clear PreExpArgTokens, just clear the entries.  Clearing the entries
  73   // would deallocate the element vectors.
  74   for (unsigned i = 0, e = PreExpArgTokens.size(); i != e; ++i)
  75     PreExpArgTokens[i].clear();
  76
  77   // Add this to the preprocessor's free list.
  78   ArgCache = PP.MacroArgCache;
  79   PP.MacroArgCache = this;
  80 }
  81
  82 /// deallocate - This should only be called by the Preprocessor when managing
  83 /// its freelist.
  84 MacroArgs *MacroArgs::deallocate() {
  85   MacroArgs *Next = ArgCache;
  86
  87   // Run the dtor to deallocate the vectors.
  88   this->~MacroArgs();
  89   // Release the memory for the object.
  90   free(this);
  91
  92   return Next;
  93 }
  94
  95
  96 /// getArgLength - Given a pointer to an expanded or unexpanded argument,
  97 /// return the number of tokens, not counting the EOF, that make up the
  98 /// argument.
  99 unsigned MacroArgs::getArgLength(const Token *ArgPtr) {
 100   unsigned NumArgTokens = 0;
 101   for (; ArgPtr->isNot(tok::eof); ++ArgPtr)
 102     ++NumArgTokens;
 103   return NumArgTokens;
 104 }
 105
 106
 107 /// getUnexpArgument - Return the unexpanded tokens for the specified formal.
 108 ///
 109 const Token *MacroArgs::getUnexpArgument(unsigned Arg) const {
 110   // The unexpanded argument tokens start immediately after the MacroArgs object
 111   // in memory.
 112   const Token *Start = (const Token *)(this+1);
 113   const Token *Result = Start;
 114   // Scan to find Arg.
 115   for (; Arg; ++Result) {
 116     assert(Result < Start+NumUnexpArgTokens && "Invalid arg #");
 117     if (Result->is(tok::eof))
 118       --Arg;
 119   }
 120   assert(Result < Start+NumUnexpArgTokens && "Invalid arg #");
 121   return Result;
 122 }
 123
 124
 125 /// ArgNeedsPreexpansion - If we can prove that the argument won't be affected
 126 /// by pre-expansion, return false.  Otherwise, conservatively return true.
 127 bool MacroArgs::ArgNeedsPreexpansion(const Token *ArgTok,
 128                                      Preprocessor &PP) const {
 129   // If there are no identifiers in the argument list, or if the identifiers are
 130   // known to not be macros, pre-expansion won't modify it.
 131   for (; ArgTok->isNot(tok::eof); ++ArgTok)
 132     if (IdentifierInfo *II = ArgTok->getIdentifierInfo()) {
 133       if (II->hasMacroDefinition() && PP.getMacroInfo(II)->isEnabled())
 134         // Return true even though the macro could be a function-like macro
 135         // without a following '(' token.
 136         return true;
 137     }
 138   return false;
 139 }
 140
 141 /// getPreExpArgument - Return the pre-expanded form of the specified
 142 /// argument.
 143 const std::vector<Token> &
 144 MacroArgs::getPreExpArgument(unsigned Arg, const MacroInfo *MI,
 145                              Preprocessor &PP) {
 146   assert(Arg < MI->getNumArgs() && "Invalid argument number!");
 147
 148   // If we have already computed this, return it.
 149   if (PreExpArgTokens.size() < MI->getNumArgs())
 150     PreExpArgTokens.resize(MI->getNumArgs());
 151
 152   std::vector<Token> &Result = PreExpArgTokens[Arg];
 153   if (!Result.empty()) return Result;
 154
 155   const Token *AT = getUnexpArgument(Arg);
 156   unsigned NumToks = getArgLength(AT)+1;  // Include the EOF.
 157
 158   // Otherwise, we have to pre-expand this argument, populating Result.  To do
 159   // this, we set up a fake TokenLexer to lex from the unexpanded argument
 160   // list.  With this installed, we lex expanded tokens until we hit the EOF
 161   // token at the end of the unexp list.
 162   PP.EnterTokenStream(AT, NumToks, false /*disable expand*/,
 163                       false /*owns tokens*/);
 164
 165   // Lex all of the macro-expanded tokens into Result.
 166   do {
 167     Result.push_back(Token());
 168     Token &Tok = Result.back();
 169     PP.Lex(Tok);
 170   } while (Result.back().isNot(tok::eof));
 171
 172   // Pop the token stream off the top of the stack.  We know that the internal
 173   // pointer inside of it is to the "end" of the token stream, but the stack
 174   // will not otherwise be popped until the next token is lexed.  The problem is
 175   // that the token may be lexed sometime after the vector of tokens itself is
 176   // destroyed, which would be badness.
 177   PP.RemoveTopOfLexerStack();
 178   return Result;
 179 }
 180
 181
 182 /// StringifyArgument - Implement C99 6.10.3.2p2, converting a sequence of
 183 /// tokens into the literal string token that should be produced by the C #
 184 /// preprocessor operator.  If Charify is true, then it should be turned into
 185 /// a character literal for the Microsoft charize (#@) extension.
 186 ///
 187 Token MacroArgs::StringifyArgument(const Token *ArgToks,
 188                                    Preprocessor &PP, bool Charify) {
 189   Token Tok;
 190   Tok.startToken();
 191   Tok.setKind(Charify ? tok::char_constant : tok::string_literal);
 192
 193   const Token *ArgTokStart = ArgToks;
 194
 195   // Stringify all the tokens.
 196   llvm::SmallString<128> Result;
 197   Result += "\"";
 198
 199   bool isFirst = true;
 200   for (; ArgToks->isNot(tok::eof); ++ArgToks) {
 201     const Token &Tok = *ArgToks;
 202     if (!isFirst && (Tok.hasLeadingSpace() || Tok.isAtStartOfLine()))
 203       Result += ' ';
 204     isFirst = false;
 205
 206     // If this is a string or character constant, escape the token as specified
 207     // by 6.10.3.2p2.
 208     if (Tok.is(tok::string_literal) ||       // "foo"
 209         Tok.is(tok::wide_string_literal) ||  // L"foo"
 210         Tok.is(tok::char_constant)) {        // 'x' and L'x'.
 211       bool Invalid = false;
 212       std::string TokStr = PP.getSpelling(Tok, &Invalid);
 213       if (!Invalid) {
 214         std::string Str = Lexer::Stringify(TokStr);
 215         Result.append(Str.begin(), Str.end());
 216       }
 217     } else {
 218       // Otherwise, just append the token.  Do some gymnastics to get the token
 219       // in place and avoid copies where possible.
 220       unsigned CurStrLen = Result.size();
 221       Result.resize(CurStrLen+Tok.getLength());
 222       const char *BufPtr = &Result[CurStrLen];
 223       bool Invalid = false;
 224       unsigned ActualTokLen = PP.getSpelling(Tok, BufPtr, &Invalid);
 225
 226       if (!Invalid) {
 227         // If getSpelling returned a pointer to an already uniqued version of
 228         // the string instead of filling in BufPtr, memcpy it onto our string.
 229         if (BufPtr != &Result[CurStrLen])
 230           memcpy(&Result[CurStrLen], BufPtr, ActualTokLen);
 231
 232         // If the token was dirty, the spelling may be shorter than the token.
 233         if (ActualTokLen != Tok.getLength())
 234           Result.resize(CurStrLen+ActualTokLen);
 235       }
 236     }
 237   }
 238
 239   // If the last character of the string is a \, and if it isn't escaped, this
 240   // is an invalid string literal, diagnose it as specified in C99.
 241   if (Result.back() == '\\') {
 242     // Count the number of consequtive \ characters.  If even, then they are
 243     // just escaped backslashes, otherwise it's an error.
 244     unsigned FirstNonSlash = Result.size()-2;
 245     // Guaranteed to find the starting " if nothing else.
 246     while (Result[FirstNonSlash] == '\\')
 247       --FirstNonSlash;
 248     if ((Result.size()-1-FirstNonSlash) & 1) {
 249       // Diagnose errors for things like: #define F(X) #X   /   F(\)
 250       PP.Diag(ArgToks[-1], diag::pp_invalid_string_literal);
 251       Result.pop_back();  // remove one of the \'s.
 252     }
 253   }
 254   Result += '"';
 255
 256   // If this is the charify operation and the result is not a legal character
 257   // constant, diagnose it.
 258   if (Charify) {
 259     // First step, turn double quotes into single quotes:
 260     Result[0] = '\'';
 261     Result[Result.size()-1] = '\'';
 262
 263     // Check for bogus character.
 264     bool isBad = false;
 265     if (Result.size() == 3)
 266       isBad = Result[1] == '\'';   // ''' is not legal. '\' already fixed above.
 267     else
 268       isBad = (Result.size() != 4 || Result[1] != '\\');  // Not '\x'
 269
 270     if (isBad) {
 271       PP.Diag(ArgTokStart[0], diag::err_invalid_character_to_charify);
 272       Result = "' '";  // Use something arbitrary, but legal.
 273     }
 274   }
 275
 276   PP.CreateString(&Result[0], Result.size(), Tok);
 277   return Tok;
 278 }
 279
 280 /// getStringifiedArgument - Compute, cache, and return the specified argument
 281 /// that has been 'stringified' as required by the # operator.
 282 const Token &MacroArgs::getStringifiedArgument(unsigned ArgNo,
 283                                                Preprocessor &PP) {
 284   assert(ArgNo < NumUnexpArgTokens && "Invalid argument number!");
 285   if (StringifiedArgs.empty()) {
 286     StringifiedArgs.resize(getNumArguments());
 287     memset(&StringifiedArgs[0], 0,
 288            sizeof(StringifiedArgs[0])*getNumArguments());
 289   }
 290   if (StringifiedArgs[ArgNo].isNot(tok::string_literal))
 291     StringifiedArgs[ArgNo] = StringifyArgument(getUnexpArgument(ArgNo), PP);
 292   return StringifiedArgs[ArgNo];
 293 }