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   assert(MI->isFunctionLike() &&
25          "Can't have args for an object-like macro!");
26 
27   // Allocate memory for the MacroArgs object with the lexer tokens at the end.
28   MacroArgs *Result = (MacroArgs*)malloc(sizeof(MacroArgs) +
29                                          NumToks*sizeof(Token));
30   // Construct the macroargs object.
31   new (Result) MacroArgs(NumToks, VarargsElided);
32 
33   // Copy the actual unexpanded tokens to immediately after the result ptr.
34   if (NumToks)
35     memcpy(const_cast<Token*>(Result->getUnexpArgument(0)),
36            UnexpArgTokens, NumToks*sizeof(Token));
37 
38   return Result;
39 }
40 
41 /// destroy - Destroy and deallocate the memory for this object.
42 ///
43 void MacroArgs::destroy() {
44   // Run the dtor to deallocate the vectors.
45   this->~MacroArgs();
46   // Release the memory for the object.
47   free(this);
48 }
49 
50 
51 /// getArgLength - Given a pointer to an expanded or unexpanded argument,
52 /// return the number of tokens, not counting the EOF, that make up the
53 /// argument.
54 unsigned MacroArgs::getArgLength(const Token *ArgPtr) {
55   unsigned NumArgTokens = 0;
56   for (; ArgPtr->isNot(tok::eof); ++ArgPtr)
57     ++NumArgTokens;
58   return NumArgTokens;
59 }
60 
61 
62 /// getUnexpArgument - Return the unexpanded tokens for the specified formal.
63 ///
64 const Token *MacroArgs::getUnexpArgument(unsigned Arg) const {
65   // The unexpanded argument tokens start immediately after the MacroArgs object
66   // in memory.
67   const Token *Start = (const Token *)(this+1);
68   const Token *Result = Start;
69   // Scan to find Arg.
70   for (; Arg; ++Result) {
71     assert(Result < Start+NumUnexpArgTokens && "Invalid arg #");
72     if (Result->is(tok::eof))
73       --Arg;
74   }
75   assert(Result < Start+NumUnexpArgTokens && "Invalid arg #");
76   return Result;
77 }
78 
79 
80 /// ArgNeedsPreexpansion - If we can prove that the argument won't be affected
81 /// by pre-expansion, return false.  Otherwise, conservatively return true.
82 bool MacroArgs::ArgNeedsPreexpansion(const Token *ArgTok,
83                                      Preprocessor &PP) const {
84   // If there are no identifiers in the argument list, or if the identifiers are
85   // known to not be macros, pre-expansion won't modify it.
86   for (; ArgTok->isNot(tok::eof); ++ArgTok)
87     if (IdentifierInfo *II = ArgTok->getIdentifierInfo()) {
88       if (II->hasMacroDefinition() && PP.getMacroInfo(II)->isEnabled())
89         // Return true even though the macro could be a function-like macro
90         // without a following '(' token.
91         return true;
92     }
93   return false;
94 }
95 
96 /// getPreExpArgument - Return the pre-expanded form of the specified
97 /// argument.
98 const std::vector<Token> &
99 MacroArgs::getPreExpArgument(unsigned Arg, Preprocessor &PP) {
100   assert(Arg < NumUnexpArgTokens && "Invalid argument number!");
101 
102   // If we have already computed this, return it.
103   if (PreExpArgTokens.empty())
104     PreExpArgTokens.resize(NumUnexpArgTokens);
105 
106   std::vector<Token> &Result = PreExpArgTokens[Arg];
107   if (!Result.empty()) return Result;
108 
109   const Token *AT = getUnexpArgument(Arg);
110   unsigned NumToks = getArgLength(AT)+1;  // Include the EOF.
111 
112   // Otherwise, we have to pre-expand this argument, populating Result.  To do
113   // this, we set up a fake TokenLexer to lex from the unexpanded argument
114   // list.  With this installed, we lex expanded tokens until we hit the EOF
115   // token at the end of the unexp list.
116   PP.EnterTokenStream(AT, NumToks, false /*disable expand*/,
117                       false /*owns tokens*/);
118 
119   // Lex all of the macro-expanded tokens into Result.
120   do {
121     Result.push_back(Token());
122     Token &Tok = Result.back();
123     PP.Lex(Tok);
124   } while (Result.back().isNot(tok::eof));
125 
126   // Pop the token stream off the top of the stack.  We know that the internal
127   // pointer inside of it is to the "end" of the token stream, but the stack
128   // will not otherwise be popped until the next token is lexed.  The problem is
129   // that the token may be lexed sometime after the vector of tokens itself is
130   // destroyed, which would be badness.
131   PP.RemoveTopOfLexerStack();
132   return Result;
133 }
134 
135 
136 /// StringifyArgument - Implement C99 6.10.3.2p2, converting a sequence of
137 /// tokens into the literal string token that should be produced by the C #
138 /// preprocessor operator.  If Charify is true, then it should be turned into
139 /// a character literal for the Microsoft charize (#@) extension.
140 ///
141 Token MacroArgs::StringifyArgument(const Token *ArgToks,
142                                    Preprocessor &PP, bool Charify) {
143   Token Tok;
144   Tok.startToken();
145   Tok.setKind(tok::string_literal);
146 
147   const Token *ArgTokStart = ArgToks;
148 
149   // Stringify all the tokens.
150   llvm::SmallString<128> Result;
151   Result += "\"";
152 
153   bool isFirst = true;
154   for (; ArgToks->isNot(tok::eof); ++ArgToks) {
155     const Token &Tok = *ArgToks;
156     if (!isFirst && (Tok.hasLeadingSpace() || Tok.isAtStartOfLine()))
157       Result += ' ';
158     isFirst = false;
159 
160     // If this is a string or character constant, escape the token as specified
161     // by 6.10.3.2p2.
162     if (Tok.is(tok::string_literal) ||       // "foo"
163         Tok.is(tok::wide_string_literal) ||  // L"foo"
164         Tok.is(tok::char_constant)) {        // 'x' and L'x'.
165       std::string Str = Lexer::Stringify(PP.getSpelling(Tok));
166       Result.append(Str.begin(), Str.end());
167     } else {
168       // Otherwise, just append the token.  Do some gymnastics to get the token
169       // in place and avoid copies where possible.
170       unsigned CurStrLen = Result.size();
171       Result.resize(CurStrLen+Tok.getLength());
172       const char *BufPtr = &Result[CurStrLen];
173       unsigned ActualTokLen = PP.getSpelling(Tok, BufPtr);
174 
175       // If getSpelling returned a pointer to an already uniqued version of the
176       // string instead of filling in BufPtr, memcpy it onto our string.
177       if (BufPtr != &Result[CurStrLen])
178         memcpy(&Result[CurStrLen], BufPtr, ActualTokLen);
179 
180       // If the token was dirty, the spelling may be shorter than the token.
181       if (ActualTokLen != Tok.getLength())
182         Result.resize(CurStrLen+ActualTokLen);
183     }
184   }
185 
186   // If the last character of the string is a \, and if it isn't escaped, this
187   // is an invalid string literal, diagnose it as specified in C99.
188   if (Result.back() == '\\') {
189     // Count the number of consequtive \ characters.  If even, then they are
190     // just escaped backslashes, otherwise it's an error.
191     unsigned FirstNonSlash = Result.size()-2;
192     // Guaranteed to find the starting " if nothing else.
193     while (Result[FirstNonSlash] == '\\')
194       --FirstNonSlash;
195     if ((Result.size()-1-FirstNonSlash) & 1) {
196       // Diagnose errors for things like: #define F(X) #X   /   F(\)
197       PP.Diag(ArgToks[-1], diag::pp_invalid_string_literal);
198       Result.pop_back();  // remove one of the \'s.
199     }
200   }
201   Result += '"';
202 
203   // If this is the charify operation and the result is not a legal character
204   // constant, diagnose it.
205   if (Charify) {
206     // First step, turn double quotes into single quotes:
207     Result[0] = '\'';
208     Result[Result.size()-1] = '\'';
209 
210     // Check for bogus character.
211     bool isBad = false;
212     if (Result.size() == 3)
213       isBad = Result[1] == '\'';   // ''' is not legal. '\' already fixed above.
214     else
215       isBad = (Result.size() != 4 || Result[1] != '\\');  // Not '\x'
216 
217     if (isBad) {
218       PP.Diag(ArgTokStart[0], diag::err_invalid_character_to_charify);
219       Result = "' '";  // Use something arbitrary, but legal.
220     }
221   }
222 
223   PP.CreateString(&Result[0], Result.size(), Tok);
224   return Tok;
225 }
226 
227 /// getStringifiedArgument - Compute, cache, and return the specified argument
228 /// that has been 'stringified' as required by the # operator.
229 const Token &MacroArgs::getStringifiedArgument(unsigned ArgNo,
230                                                Preprocessor &PP) {
231   assert(ArgNo < NumUnexpArgTokens && "Invalid argument number!");
232   if (StringifiedArgs.empty()) {
233     StringifiedArgs.resize(getNumArguments());
234     memset(&StringifiedArgs[0], 0,
235            sizeof(StringifiedArgs[0])*getNumArguments());
236   }
237   if (StringifiedArgs[ArgNo].isNot(tok::string_literal))
238     StringifiedArgs[ArgNo] = StringifyArgument(getUnexpArgument(ArgNo), PP);
239   return StringifiedArgs[ArgNo];
240 }
241