Home | History | Annotate | Download | only in Lex 
      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 "clang/Lex/MacroArgs.h"
     15 #include "clang/Lex/LexDiagnostic.h"
     16 #include "clang/Lex/MacroInfo.h"
     17 #include "clang/Lex/Preprocessor.h"
     18 #include "llvm/ADT/SmallString.h"
     19 #include "llvm/Support/SaveAndRestore.h"
     20 #include <algorithm>
     21 
     22 using namespace clang;
     23 
     24 /// MacroArgs ctor function - This destroys the vector passed in.
     25 MacroArgs *MacroArgs::create(const MacroInfo *MI,
     26                              ArrayRef<Token> UnexpArgTokens,
     27                              bool VarargsElided, Preprocessor &PP) {
     28   assert(MI->isFunctionLike() &&
     29          "Can't have args for an object-like macro!");
     30   MacroArgs **ResultEnt = 0;
     31   unsigned ClosestMatch = ~0U;
     32 
     33   // See if we have an entry with a big enough argument list to reuse on the
     34   // free list.  If so, reuse it.
     35   for (MacroArgs **Entry = &PP.MacroArgCache; *Entry;
     36        Entry = &(*Entry)->ArgCache)
     37     if ((*Entry)->NumUnexpArgTokens >= UnexpArgTokens.size() &&
     38         (*Entry)->NumUnexpArgTokens < ClosestMatch) {
     39       ResultEnt = Entry;
     40 
     41       // If we have an exact match, use it.
     42       if ((*Entry)->NumUnexpArgTokens == UnexpArgTokens.size())
     43         break;
     44       // Otherwise, use the best fit.
     45       ClosestMatch = (*Entry)->NumUnexpArgTokens;
     46     }
     47 
     48   MacroArgs *Result;
     49   if (ResultEnt == 0) {
     50     // Allocate memory for a MacroArgs object with the lexer tokens at the end.
     51     Result = (MacroArgs*)malloc(sizeof(MacroArgs) +
     52                                 UnexpArgTokens.size() * sizeof(Token));
     53     // Construct the MacroArgs object.
     54     new (Result) MacroArgs(UnexpArgTokens.size(), VarargsElided);
     55   } else {
     56     Result = *ResultEnt;
     57     // Unlink this node from the preprocessors singly linked list.
     58     *ResultEnt = Result->ArgCache;
     59     Result->NumUnexpArgTokens = UnexpArgTokens.size();
     60     Result->VarargsElided = VarargsElided;
     61   }
     62 
     63   // Copy the actual unexpanded tokens to immediately after the result ptr.
     64   if (!UnexpArgTokens.empty())
     65     std::copy(UnexpArgTokens.begin(), UnexpArgTokens.end(),
     66               const_cast<Token*>(Result->getUnexpArgument(0)));
     67 
     68   return Result;
     69 }
     70 
     71 /// destroy - Destroy and deallocate the memory for this object.
     72 ///
     73 void MacroArgs::destroy(Preprocessor &PP) {
     74   StringifiedArgs.clear();
     75 
     76   // Don't clear PreExpArgTokens, just clear the entries.  Clearing the entries
     77   // would deallocate the element vectors.
     78   for (unsigned i = 0, e = PreExpArgTokens.size(); i != e; ++i)
     79     PreExpArgTokens[i].clear();
     80 
     81   // Add this to the preprocessor's free list.
     82   ArgCache = PP.MacroArgCache;
     83   PP.MacroArgCache = this;
     84 }
     85 
     86 /// deallocate - This should only be called by the Preprocessor when managing
     87 /// its freelist.
     88 MacroArgs *MacroArgs::deallocate() {
     89   MacroArgs *Next = ArgCache;
     90 
     91   // Run the dtor to deallocate the vectors.
     92   this->~MacroArgs();
     93   // Release the memory for the object.
     94   free(this);
     95 
     96   return Next;
     97 }
     98 
     99 
    100 /// getArgLength - Given a pointer to an expanded or unexpanded argument,
    101 /// return the number of tokens, not counting the EOF, that make up the
    102 /// argument.
    103 unsigned MacroArgs::getArgLength(const Token *ArgPtr) {
    104   unsigned NumArgTokens = 0;
    105   for (; ArgPtr->isNot(tok::eof); ++ArgPtr)
    106     ++NumArgTokens;
    107   return NumArgTokens;
    108 }
    109 
    110 
    111 /// getUnexpArgument - Return the unexpanded tokens for the specified formal.
    112 ///
    113 const Token *MacroArgs::getUnexpArgument(unsigned Arg) const {
    114   // The unexpanded argument tokens start immediately after the MacroArgs object
    115   // in memory.
    116   const Token *Start = (const Token *)(this+1);
    117   const Token *Result = Start;
    118   // Scan to find Arg.
    119   for (; Arg; ++Result) {
    120     assert(Result < Start+NumUnexpArgTokens && "Invalid arg #");
    121     if (Result->is(tok::eof))
    122       --Arg;
    123   }
    124   assert(Result < Start+NumUnexpArgTokens && "Invalid arg #");
    125   return Result;
    126 }
    127 
    128 
    129 /// ArgNeedsPreexpansion - If we can prove that the argument won't be affected
    130 /// by pre-expansion, return false.  Otherwise, conservatively return true.
    131 bool MacroArgs::ArgNeedsPreexpansion(const Token *ArgTok,
    132                                      Preprocessor &PP) const {
    133   // If there are no identifiers in the argument list, or if the identifiers are
    134   // known to not be macros, pre-expansion won't modify it.
    135   for (; ArgTok->isNot(tok::eof); ++ArgTok)
    136     if (IdentifierInfo *II = ArgTok->getIdentifierInfo()) {
    137       if (II->hasMacroDefinition() && PP.getMacroInfo(II)->isEnabled())
    138         // Return true even though the macro could be a function-like macro
    139         // without a following '(' token.
    140         return true;
    141     }
    142   return false;
    143 }
    144 
    145 /// getPreExpArgument - Return the pre-expanded form of the specified
    146 /// argument.
    147 const std::vector<Token> &
    148 MacroArgs::getPreExpArgument(unsigned Arg, const MacroInfo *MI,
    149                              Preprocessor &PP) {
    150   assert(Arg < MI->getNumArgs() && "Invalid argument number!");
    151 
    152   // If we have already computed this, return it.
    153   if (PreExpArgTokens.size() < MI->getNumArgs())
    154     PreExpArgTokens.resize(MI->getNumArgs());
    155 
    156   std::vector<Token> &Result = PreExpArgTokens[Arg];
    157   if (!Result.empty()) return Result;
    158 
    159   SaveAndRestore<bool> PreExpandingMacroArgs(PP.InMacroArgPreExpansion, true);
    160 
    161   const Token *AT = getUnexpArgument(Arg);
    162   unsigned NumToks = getArgLength(AT)+1;  // Include the EOF.
    163 
    164   // Otherwise, we have to pre-expand this argument, populating Result.  To do
    165   // this, we set up a fake TokenLexer to lex from the unexpanded argument
    166   // list.  With this installed, we lex expanded tokens until we hit the EOF
    167   // token at the end of the unexp list.
    168   PP.EnterTokenStream(AT, NumToks, false /*disable expand*/,
    169                       false /*owns tokens*/);
    170 
    171   // Lex all of the macro-expanded tokens into Result.
    172   do {
    173     Result.push_back(Token());
    174     Token &Tok = Result.back();
    175     PP.Lex(Tok);
    176   } while (Result.back().isNot(tok::eof));
    177 
    178   // Pop the token stream off the top of the stack.  We know that the internal
    179   // pointer inside of it is to the "end" of the token stream, but the stack
    180   // will not otherwise be popped until the next token is lexed.  The problem is
    181   // that the token may be lexed sometime after the vector of tokens itself is
    182   // destroyed, which would be badness.
    183   if (PP.InCachingLexMode())
    184     PP.ExitCachingLexMode();
    185   PP.RemoveTopOfLexerStack();
    186   return Result;
    187 }
    188 
    189 
    190 /// StringifyArgument - Implement C99 6.10.3.2p2, converting a sequence of
    191 /// tokens into the literal string token that should be produced by the C #
    192 /// preprocessor operator.  If Charify is true, then it should be turned into
    193 /// a character literal for the Microsoft charize (#@) extension.
    194 ///
    195 Token MacroArgs::StringifyArgument(const Token *ArgToks,
    196                                    Preprocessor &PP, bool Charify,
    197                                    SourceLocation ExpansionLocStart,
    198                                    SourceLocation ExpansionLocEnd) {
    199   Token Tok;
    200   Tok.startToken();
    201   Tok.setKind(Charify ? tok::char_constant : tok::string_literal);
    202 
    203   const Token *ArgTokStart = ArgToks;
    204 
    205   // Stringify all the tokens.
    206   SmallString<128> Result;
    207   Result += "\"";
    208 
    209   bool isFirst = true;
    210   for (; ArgToks->isNot(tok::eof); ++ArgToks) {
    211     const Token &Tok = *ArgToks;
    212     if (!isFirst && (Tok.hasLeadingSpace() || Tok.isAtStartOfLine()))
    213       Result += ' ';
    214     isFirst = false;
    215 
    216     // If this is a string or character constant, escape the token as specified
    217     // by 6.10.3.2p2.
    218     if (tok::isStringLiteral(Tok.getKind()) || // "foo", u8R"x(foo)x"_bar, etc.
    219         Tok.is(tok::char_constant) ||          // 'x'
    220         Tok.is(tok::wide_char_constant) ||     // L'x'.
    221         Tok.is(tok::utf16_char_constant) ||    // u'x'.
    222         Tok.is(tok::utf32_char_constant)) {    // U'x'.
    223       bool Invalid = false;
    224       std::string TokStr = PP.getSpelling(Tok, &Invalid);
    225       if (!Invalid) {
    226         std::string Str = Lexer::Stringify(TokStr);
    227         Result.append(Str.begin(), Str.end());
    228       }
    229     } else if (Tok.is(tok::code_completion)) {
    230       PP.CodeCompleteNaturalLanguage();
    231     } else {
    232       // Otherwise, just append the token.  Do some gymnastics to get the token
    233       // in place and avoid copies where possible.
    234       unsigned CurStrLen = Result.size();
    235       Result.resize(CurStrLen+Tok.getLength());
    236       const char *BufPtr = &Result[CurStrLen];
    237       bool Invalid = false;
    238       unsigned ActualTokLen = PP.getSpelling(Tok, BufPtr, &Invalid);
    239 
    240       if (!Invalid) {
    241         // If getSpelling returned a pointer to an already uniqued version of
    242         // the string instead of filling in BufPtr, memcpy it onto our string.
    243         if (BufPtr != &Result[CurStrLen])
    244           memcpy(&Result[CurStrLen], BufPtr, ActualTokLen);
    245 
    246         // If the token was dirty, the spelling may be shorter than the token.
    247         if (ActualTokLen != Tok.getLength())
    248           Result.resize(CurStrLen+ActualTokLen);
    249       }
    250     }
    251   }
    252 
    253   // If the last character of the string is a \, and if it isn't escaped, this
    254   // is an invalid string literal, diagnose it as specified in C99.
    255   if (Result.back() == '\\') {
    256     // Count the number of consequtive \ characters.  If even, then they are
    257     // just escaped backslashes, otherwise it's an error.
    258     unsigned FirstNonSlash = Result.size()-2;
    259     // Guaranteed to find the starting " if nothing else.
    260     while (Result[FirstNonSlash] == '\\')
    261       --FirstNonSlash;
    262     if ((Result.size()-1-FirstNonSlash) & 1) {
    263       // Diagnose errors for things like: #define F(X) #X   /   F(\)
    264       PP.Diag(ArgToks[-1], diag::pp_invalid_string_literal);
    265       Result.pop_back();  // remove one of the \'s.
    266     }
    267   }
    268   Result += '"';
    269 
    270   // If this is the charify operation and the result is not a legal character
    271   // constant, diagnose it.
    272   if (Charify) {
    273     // First step, turn double quotes into single quotes:
    274     Result[0] = '\'';
    275     Result[Result.size()-1] = '\'';
    276 
    277     // Check for bogus character.
    278     bool isBad = false;
    279     if (Result.size() == 3)
    280       isBad = Result[1] == '\'';   // ''' is not legal. '\' already fixed above.
    281     else
    282       isBad = (Result.size() != 4 || Result[1] != '\\');  // Not '\x'
    283 
    284     if (isBad) {
    285       PP.Diag(ArgTokStart[0], diag::err_invalid_character_to_charify);
    286       Result = "' '";  // Use something arbitrary, but legal.
    287     }
    288   }
    289 
    290   PP.CreateString(Result, Tok,
    291                   ExpansionLocStart, ExpansionLocEnd);
    292   return Tok;
    293 }
    294 
    295 /// getStringifiedArgument - Compute, cache, and return the specified argument
    296 /// that has been 'stringified' as required by the # operator.
    297 const Token &MacroArgs::getStringifiedArgument(unsigned ArgNo,
    298                                                Preprocessor &PP,
    299                                                SourceLocation ExpansionLocStart,
    300                                                SourceLocation ExpansionLocEnd) {
    301   assert(ArgNo < NumUnexpArgTokens && "Invalid argument number!");
    302   if (StringifiedArgs.empty()) {
    303     StringifiedArgs.resize(getNumArguments());
    304     memset((void*)&StringifiedArgs[0], 0,
    305            sizeof(StringifiedArgs[0])*getNumArguments());
    306   }
    307   if (StringifiedArgs[ArgNo].isNot(tok::string_literal))
    308     StringifiedArgs[ArgNo] = StringifyArgument(getUnexpArgument(ArgNo), PP,
    309                                                /*Charify=*/false,
    310                                                ExpansionLocStart,
    311                                                ExpansionLocEnd);
    312   return StringifiedArgs[ArgNo];
    313 }
    314