Chris Lattner | 1543e9c | 2008-03-09 02:18:51 +0000 | [diff] [blame] | 1 | //===--- TokenLexer.cpp - Lex from a token stream -------------------------===// |
Reid Spencer | 5f016e2 | 2007-07-11 17:01:13 +0000 | [diff] [blame] | 2 | // |
| 3 | // The LLVM Compiler Infrastructure |
| 4 | // |
Chris Lattner | 0bc735f | 2007-12-29 19:59:25 +0000 | [diff] [blame] | 5 | // This file is distributed under the University of Illinois Open Source |
| 6 | // License. See LICENSE.TXT for details. |
Reid Spencer | 5f016e2 | 2007-07-11 17:01:13 +0000 | [diff] [blame] | 7 | // |
| 8 | //===----------------------------------------------------------------------===// |
| 9 | // |
Chris Lattner | 1543e9c | 2008-03-09 02:18:51 +0000 | [diff] [blame] | 10 | // This file implements the TokenLexer interface. |
Reid Spencer | 5f016e2 | 2007-07-11 17:01:13 +0000 | [diff] [blame] | 11 | // |
| 12 | //===----------------------------------------------------------------------===// |
| 13 | |
Chris Lattner | 5d75de0 | 2008-03-09 02:22:57 +0000 | [diff] [blame] | 14 | #include "clang/Lex/TokenLexer.h" |
Chris Lattner | e5c8ffe | 2008-03-09 02:55:12 +0000 | [diff] [blame] | 15 | #include "MacroArgs.h" |
Reid Spencer | 5f016e2 | 2007-07-11 17:01:13 +0000 | [diff] [blame] | 16 | #include "clang/Lex/MacroInfo.h" |
| 17 | #include "clang/Lex/Preprocessor.h" |
| 18 | #include "clang/Basic/SourceManager.h" |
Chris Lattner | 500d329 | 2009-01-29 05:15:15 +0000 | [diff] [blame] | 19 | #include "clang/Lex/LexDiagnostic.h" |
Reid Spencer | 5f016e2 | 2007-07-11 17:01:13 +0000 | [diff] [blame] | 20 | #include "llvm/ADT/SmallVector.h" |
| 21 | using namespace clang; |
| 22 | |
Reid Spencer | 5f016e2 | 2007-07-11 17:01:13 +0000 | [diff] [blame] | 23 | |
Chris Lattner | 1543e9c | 2008-03-09 02:18:51 +0000 | [diff] [blame] | 24 | /// Create a TokenLexer for the specified macro with the specified actual |
Reid Spencer | 5f016e2 | 2007-07-11 17:01:13 +0000 | [diff] [blame] | 25 | /// arguments. Note that this ctor takes ownership of the ActualArgs pointer. |
Chris Lattner | e7fb484 | 2009-02-15 20:52:18 +0000 | [diff] [blame] | 26 | void TokenLexer::Init(Token &Tok, SourceLocation ILEnd, MacroArgs *Actuals) { |
Chris Lattner | 1543e9c | 2008-03-09 02:18:51 +0000 | [diff] [blame] | 27 | // If the client is reusing a TokenLexer, make sure to free any memory |
Chris Lattner | 9594acf | 2007-07-15 00:25:26 +0000 | [diff] [blame] | 28 | // associated with it. |
| 29 | destroy(); |
Mike Stump | 1eb4433 | 2009-09-09 15:08:12 +0000 | [diff] [blame] | 30 | |
Chris Lattner | cc1a875 | 2007-10-07 08:44:20 +0000 | [diff] [blame] | 31 | Macro = PP.getMacroInfo(Tok.getIdentifierInfo()); |
Chris Lattner | 9594acf | 2007-07-15 00:25:26 +0000 | [diff] [blame] | 32 | ActualArgs = Actuals; |
| 33 | CurToken = 0; |
Mike Stump | 1eb4433 | 2009-09-09 15:08:12 +0000 | [diff] [blame] | 34 | |
Chris Lattner | e7fb484 | 2009-02-15 20:52:18 +0000 | [diff] [blame] | 35 | InstantiateLocStart = Tok.getLocation(); |
| 36 | InstantiateLocEnd = ILEnd; |
Chris Lattner | 9594acf | 2007-07-15 00:25:26 +0000 | [diff] [blame] | 37 | AtStartOfLine = Tok.isAtStartOfLine(); |
| 38 | HasLeadingSpace = Tok.hasLeadingSpace(); |
Chris Lattner | 8d89643 | 2008-03-09 02:07:49 +0000 | [diff] [blame] | 39 | Tokens = &*Macro->tokens_begin(); |
| 40 | OwnsTokens = false; |
Chris Lattner | 6b88450 | 2008-03-10 06:06:04 +0000 | [diff] [blame] | 41 | DisableMacroExpansion = false; |
Chris Lattner | 8d89643 | 2008-03-09 02:07:49 +0000 | [diff] [blame] | 42 | NumTokens = Macro->tokens_end()-Macro->tokens_begin(); |
Argyrios Kyrtzidis | b73377e | 2011-07-07 03:40:34 +0000 | [diff] [blame^] | 43 | MacroExpansionStart = SourceLocation(); |
| 44 | |
| 45 | SourceManager &SM = PP.getSourceManager(); |
| 46 | MacroStartSLocOffset = SM.getNextOffset(); |
| 47 | |
| 48 | if (NumTokens > 0) { |
| 49 | assert(Tokens[0].getLocation().isValid()); |
| 50 | assert((Tokens[0].getLocation().isFileID() || Tokens[0].is(tok::comment)) && |
| 51 | "Macro defined in macro?"); |
| 52 | assert(InstantiateLocStart.isValid()); |
| 53 | |
| 54 | // Reserve a source location entry chunk for the length of the macro |
| 55 | // definition. Tokens that get lexed directly from the definition will |
| 56 | // have their locations pointing inside this chunk. This is to avoid |
| 57 | // creating separate source location entries for each token. |
| 58 | SourceLocation macroStart = SM.getInstantiationLoc(Tokens[0].getLocation()); |
| 59 | MacroDefStartInfo = SM.getDecomposedLoc(macroStart); |
| 60 | MacroExpansionStart = SM.createInstantiationLoc(macroStart, |
| 61 | InstantiateLocStart, |
| 62 | InstantiateLocEnd, |
| 63 | Macro->getDefinitionLength(SM)); |
| 64 | } |
Reid Spencer | 5f016e2 | 2007-07-11 17:01:13 +0000 | [diff] [blame] | 65 | |
| 66 | // If this is a function-like macro, expand the arguments and change |
Chris Lattner | 8d89643 | 2008-03-09 02:07:49 +0000 | [diff] [blame] | 67 | // Tokens to point to the expanded tokens. |
Reid Spencer | 5f016e2 | 2007-07-11 17:01:13 +0000 | [diff] [blame] | 68 | if (Macro->isFunctionLike() && Macro->getNumArgs()) |
| 69 | ExpandFunctionArguments(); |
Mike Stump | 1eb4433 | 2009-09-09 15:08:12 +0000 | [diff] [blame] | 70 | |
Reid Spencer | 5f016e2 | 2007-07-11 17:01:13 +0000 | [diff] [blame] | 71 | // Mark the macro as currently disabled, so that it is not recursively |
| 72 | // expanded. The macro must be disabled only after argument pre-expansion of |
| 73 | // function-like macro arguments occurs. |
| 74 | Macro->DisableMacro(); |
| 75 | } |
| 76 | |
Chris Lattner | 9594acf | 2007-07-15 00:25:26 +0000 | [diff] [blame] | 77 | |
| 78 | |
Chris Lattner | 1543e9c | 2008-03-09 02:18:51 +0000 | [diff] [blame] | 79 | /// Create a TokenLexer for the specified token stream. This does not |
Reid Spencer | 5f016e2 | 2007-07-11 17:01:13 +0000 | [diff] [blame] | 80 | /// take ownership of the specified token vector. |
Chris Lattner | 6b88450 | 2008-03-10 06:06:04 +0000 | [diff] [blame] | 81 | void TokenLexer::Init(const Token *TokArray, unsigned NumToks, |
| 82 | bool disableMacroExpansion, bool ownsTokens) { |
Chris Lattner | 1543e9c | 2008-03-09 02:18:51 +0000 | [diff] [blame] | 83 | // If the client is reusing a TokenLexer, make sure to free any memory |
Chris Lattner | 9594acf | 2007-07-15 00:25:26 +0000 | [diff] [blame] | 84 | // associated with it. |
| 85 | destroy(); |
Mike Stump | 1eb4433 | 2009-09-09 15:08:12 +0000 | [diff] [blame] | 86 | |
Chris Lattner | 9594acf | 2007-07-15 00:25:26 +0000 | [diff] [blame] | 87 | Macro = 0; |
| 88 | ActualArgs = 0; |
Chris Lattner | 8d89643 | 2008-03-09 02:07:49 +0000 | [diff] [blame] | 89 | Tokens = TokArray; |
Chris Lattner | 6b88450 | 2008-03-10 06:06:04 +0000 | [diff] [blame] | 90 | OwnsTokens = ownsTokens; |
| 91 | DisableMacroExpansion = disableMacroExpansion; |
Chris Lattner | 8d89643 | 2008-03-09 02:07:49 +0000 | [diff] [blame] | 92 | NumTokens = NumToks; |
Chris Lattner | 9594acf | 2007-07-15 00:25:26 +0000 | [diff] [blame] | 93 | CurToken = 0; |
Chris Lattner | e7fb484 | 2009-02-15 20:52:18 +0000 | [diff] [blame] | 94 | InstantiateLocStart = InstantiateLocEnd = SourceLocation(); |
Chris Lattner | 9594acf | 2007-07-15 00:25:26 +0000 | [diff] [blame] | 95 | AtStartOfLine = false; |
| 96 | HasLeadingSpace = false; |
Argyrios Kyrtzidis | b73377e | 2011-07-07 03:40:34 +0000 | [diff] [blame^] | 97 | MacroExpansionStart = SourceLocation(); |
Mike Stump | 1eb4433 | 2009-09-09 15:08:12 +0000 | [diff] [blame] | 98 | |
Reid Spencer | 5f016e2 | 2007-07-11 17:01:13 +0000 | [diff] [blame] | 99 | // Set HasLeadingSpace/AtStartOfLine so that the first token will be |
| 100 | // returned unmodified. |
| 101 | if (NumToks != 0) { |
| 102 | AtStartOfLine = TokArray[0].isAtStartOfLine(); |
| 103 | HasLeadingSpace = TokArray[0].hasLeadingSpace(); |
| 104 | } |
| 105 | } |
| 106 | |
| 107 | |
Chris Lattner | 1543e9c | 2008-03-09 02:18:51 +0000 | [diff] [blame] | 108 | void TokenLexer::destroy() { |
Reid Spencer | 5f016e2 | 2007-07-11 17:01:13 +0000 | [diff] [blame] | 109 | // If this was a function-like macro that actually uses its arguments, delete |
| 110 | // the expanded tokens. |
Chris Lattner | 8d89643 | 2008-03-09 02:07:49 +0000 | [diff] [blame] | 111 | if (OwnsTokens) { |
| 112 | delete [] Tokens; |
| 113 | Tokens = 0; |
Chris Lattner | 32fca72 | 2009-03-04 06:50:57 +0000 | [diff] [blame] | 114 | OwnsTokens = false; |
Chris Lattner | 9c68306 | 2007-07-22 01:16:55 +0000 | [diff] [blame] | 115 | } |
Mike Stump | 1eb4433 | 2009-09-09 15:08:12 +0000 | [diff] [blame] | 116 | |
Chris Lattner | 1543e9c | 2008-03-09 02:18:51 +0000 | [diff] [blame] | 117 | // TokenLexer owns its formal arguments. |
Chris Lattner | 561395b | 2009-12-14 22:12:52 +0000 | [diff] [blame] | 118 | if (ActualArgs) ActualArgs->destroy(PP); |
Reid Spencer | 5f016e2 | 2007-07-11 17:01:13 +0000 | [diff] [blame] | 119 | } |
| 120 | |
| 121 | /// Expand the arguments of a function-like macro so that we can quickly |
Chris Lattner | 8d89643 | 2008-03-09 02:07:49 +0000 | [diff] [blame] | 122 | /// return preexpanded tokens from Tokens. |
Chris Lattner | 1543e9c | 2008-03-09 02:18:51 +0000 | [diff] [blame] | 123 | void TokenLexer::ExpandFunctionArguments() { |
Chris Lattner | d217773 | 2007-07-20 16:59:19 +0000 | [diff] [blame] | 124 | llvm::SmallVector<Token, 128> ResultToks; |
Mike Stump | 1eb4433 | 2009-09-09 15:08:12 +0000 | [diff] [blame] | 125 | |
Chris Lattner | 8d89643 | 2008-03-09 02:07:49 +0000 | [diff] [blame] | 126 | // Loop through 'Tokens', expanding them into ResultToks. Keep |
Reid Spencer | 5f016e2 | 2007-07-11 17:01:13 +0000 | [diff] [blame] | 127 | // track of whether we change anything. If not, no need to keep them. If so, |
Chris Lattner | 8d89643 | 2008-03-09 02:07:49 +0000 | [diff] [blame] | 128 | // we install the newly expanded sequence as the new 'Tokens' list. |
Reid Spencer | 5f016e2 | 2007-07-11 17:01:13 +0000 | [diff] [blame] | 129 | bool MadeChange = false; |
Mike Stump | 1eb4433 | 2009-09-09 15:08:12 +0000 | [diff] [blame] | 130 | |
Reid Spencer | 5f016e2 | 2007-07-11 17:01:13 +0000 | [diff] [blame] | 131 | // NextTokGetsSpace - When this is true, the next token appended to the |
| 132 | // output list will get a leading space, regardless of whether it had one to |
| 133 | // begin with or not. This is used for placemarker support. |
| 134 | bool NextTokGetsSpace = false; |
Mike Stump | 1eb4433 | 2009-09-09 15:08:12 +0000 | [diff] [blame] | 135 | |
Chris Lattner | 8d89643 | 2008-03-09 02:07:49 +0000 | [diff] [blame] | 136 | for (unsigned i = 0, e = NumTokens; i != e; ++i) { |
Reid Spencer | 5f016e2 | 2007-07-11 17:01:13 +0000 | [diff] [blame] | 137 | // If we found the stringify operator, get the argument stringified. The |
| 138 | // preprocessor already verified that the following token is a macro name |
| 139 | // when the #define was parsed. |
Chris Lattner | 8d89643 | 2008-03-09 02:07:49 +0000 | [diff] [blame] | 140 | const Token &CurTok = Tokens[i]; |
Chris Lattner | 22f6bbc | 2007-10-09 18:02:16 +0000 | [diff] [blame] | 141 | if (CurTok.is(tok::hash) || CurTok.is(tok::hashat)) { |
Chris Lattner | 8d89643 | 2008-03-09 02:07:49 +0000 | [diff] [blame] | 142 | int ArgNo = Macro->getArgumentNum(Tokens[i+1].getIdentifierInfo()); |
Reid Spencer | 5f016e2 | 2007-07-11 17:01:13 +0000 | [diff] [blame] | 143 | assert(ArgNo != -1 && "Token following # is not an argument?"); |
Mike Stump | 1eb4433 | 2009-09-09 15:08:12 +0000 | [diff] [blame] | 144 | |
Argyrios Kyrtzidis | b73377e | 2011-07-07 03:40:34 +0000 | [diff] [blame^] | 145 | SourceLocation hashInstLoc; |
| 146 | if(InstantiateLocStart.isValid()) { |
| 147 | hashInstLoc = getMacroExpansionLocation(CurTok.getLocation()); |
| 148 | assert(hashInstLoc.isValid() && "Expected '#' to come from definition"); |
| 149 | } |
| 150 | |
Chris Lattner | d217773 | 2007-07-20 16:59:19 +0000 | [diff] [blame] | 151 | Token Res; |
Chris Lattner | 22f6bbc | 2007-10-09 18:02:16 +0000 | [diff] [blame] | 152 | if (CurTok.is(tok::hash)) // Stringify |
Argyrios Kyrtzidis | b73377e | 2011-07-07 03:40:34 +0000 | [diff] [blame^] | 153 | Res = ActualArgs->getStringifiedArgument(ArgNo, PP, hashInstLoc); |
Reid Spencer | 5f016e2 | 2007-07-11 17:01:13 +0000 | [diff] [blame] | 154 | else { |
| 155 | // 'charify': don't bother caching these. |
Chris Lattner | e5c8ffe | 2008-03-09 02:55:12 +0000 | [diff] [blame] | 156 | Res = MacroArgs::StringifyArgument(ActualArgs->getUnexpArgument(ArgNo), |
Argyrios Kyrtzidis | b73377e | 2011-07-07 03:40:34 +0000 | [diff] [blame^] | 157 | PP, true, hashInstLoc); |
Reid Spencer | 5f016e2 | 2007-07-11 17:01:13 +0000 | [diff] [blame] | 158 | } |
Mike Stump | 1eb4433 | 2009-09-09 15:08:12 +0000 | [diff] [blame] | 159 | |
Reid Spencer | 5f016e2 | 2007-07-11 17:01:13 +0000 | [diff] [blame] | 160 | // The stringified/charified string leading space flag gets set to match |
| 161 | // the #/#@ operator. |
| 162 | if (CurTok.hasLeadingSpace() || NextTokGetsSpace) |
Chris Lattner | d217773 | 2007-07-20 16:59:19 +0000 | [diff] [blame] | 163 | Res.setFlag(Token::LeadingSpace); |
Mike Stump | 1eb4433 | 2009-09-09 15:08:12 +0000 | [diff] [blame] | 164 | |
Reid Spencer | 5f016e2 | 2007-07-11 17:01:13 +0000 | [diff] [blame] | 165 | ResultToks.push_back(Res); |
| 166 | MadeChange = true; |
| 167 | ++i; // Skip arg name. |
| 168 | NextTokGetsSpace = false; |
| 169 | continue; |
| 170 | } |
Mike Stump | 1eb4433 | 2009-09-09 15:08:12 +0000 | [diff] [blame] | 171 | |
Reid Spencer | 5f016e2 | 2007-07-11 17:01:13 +0000 | [diff] [blame] | 172 | // Otherwise, if this is not an argument token, just add the token to the |
| 173 | // output buffer. |
| 174 | IdentifierInfo *II = CurTok.getIdentifierInfo(); |
| 175 | int ArgNo = II ? Macro->getArgumentNum(II) : -1; |
| 176 | if (ArgNo == -1) { |
| 177 | // This isn't an argument, just add it. |
| 178 | ResultToks.push_back(CurTok); |
| 179 | |
| 180 | if (NextTokGetsSpace) { |
Chris Lattner | d217773 | 2007-07-20 16:59:19 +0000 | [diff] [blame] | 181 | ResultToks.back().setFlag(Token::LeadingSpace); |
Reid Spencer | 5f016e2 | 2007-07-11 17:01:13 +0000 | [diff] [blame] | 182 | NextTokGetsSpace = false; |
| 183 | } |
| 184 | continue; |
| 185 | } |
Mike Stump | 1eb4433 | 2009-09-09 15:08:12 +0000 | [diff] [blame] | 186 | |
Reid Spencer | 5f016e2 | 2007-07-11 17:01:13 +0000 | [diff] [blame] | 187 | // An argument is expanded somehow, the result is different than the |
| 188 | // input. |
| 189 | MadeChange = true; |
| 190 | |
| 191 | // Otherwise, this is a use of the argument. Find out if there is a paste |
| 192 | // (##) operator before or after the argument. |
Mike Stump | 1eb4433 | 2009-09-09 15:08:12 +0000 | [diff] [blame] | 193 | bool PasteBefore = |
Chris Lattner | 22f6bbc | 2007-10-09 18:02:16 +0000 | [diff] [blame] | 194 | !ResultToks.empty() && ResultToks.back().is(tok::hashhash); |
Chris Lattner | 8d89643 | 2008-03-09 02:07:49 +0000 | [diff] [blame] | 195 | bool PasteAfter = i+1 != e && Tokens[i+1].is(tok::hashhash); |
Mike Stump | 1eb4433 | 2009-09-09 15:08:12 +0000 | [diff] [blame] | 196 | |
Reid Spencer | 5f016e2 | 2007-07-11 17:01:13 +0000 | [diff] [blame] | 197 | // If it is not the LHS/RHS of a ## operator, we must pre-expand the |
| 198 | // argument and substitute the expanded tokens into the result. This is |
| 199 | // C99 6.10.3.1p1. |
| 200 | if (!PasteBefore && !PasteAfter) { |
Chris Lattner | d217773 | 2007-07-20 16:59:19 +0000 | [diff] [blame] | 201 | const Token *ResultArgToks; |
Reid Spencer | 5f016e2 | 2007-07-11 17:01:13 +0000 | [diff] [blame] | 202 | |
| 203 | // Only preexpand the argument if it could possibly need it. This |
| 204 | // avoids some work in common cases. |
Chris Lattner | d217773 | 2007-07-20 16:59:19 +0000 | [diff] [blame] | 205 | const Token *ArgTok = ActualArgs->getUnexpArgument(ArgNo); |
Chris Lattner | cc1a875 | 2007-10-07 08:44:20 +0000 | [diff] [blame] | 206 | if (ActualArgs->ArgNeedsPreexpansion(ArgTok, PP)) |
Chris Lattner | f5809a7 | 2009-12-28 06:17:16 +0000 | [diff] [blame] | 207 | ResultArgToks = &ActualArgs->getPreExpArgument(ArgNo, Macro, PP)[0]; |
Reid Spencer | 5f016e2 | 2007-07-11 17:01:13 +0000 | [diff] [blame] | 208 | else |
| 209 | ResultArgToks = ArgTok; // Use non-preexpanded tokens. |
Mike Stump | 1eb4433 | 2009-09-09 15:08:12 +0000 | [diff] [blame] | 210 | |
Reid Spencer | 5f016e2 | 2007-07-11 17:01:13 +0000 | [diff] [blame] | 211 | // If the arg token expanded into anything, append it. |
Chris Lattner | 22f6bbc | 2007-10-09 18:02:16 +0000 | [diff] [blame] | 212 | if (ResultArgToks->isNot(tok::eof)) { |
Reid Spencer | 5f016e2 | 2007-07-11 17:01:13 +0000 | [diff] [blame] | 213 | unsigned FirstResult = ResultToks.size(); |
| 214 | unsigned NumToks = MacroArgs::getArgLength(ResultArgToks); |
| 215 | ResultToks.append(ResultArgToks, ResultArgToks+NumToks); |
Mike Stump | 1eb4433 | 2009-09-09 15:08:12 +0000 | [diff] [blame] | 216 | |
Argyrios Kyrtzidis | b73377e | 2011-07-07 03:40:34 +0000 | [diff] [blame^] | 217 | if(InstantiateLocStart.isValid()) { |
| 218 | SourceManager &SM = PP.getSourceManager(); |
| 219 | SourceLocation curInst = |
| 220 | getMacroExpansionLocation(CurTok.getLocation()); |
| 221 | assert(curInst.isValid() && |
| 222 | "Expected arg identifier to come from definition"); |
| 223 | for (unsigned i = FirstResult, e = ResultToks.size(); i != e; ++i) { |
| 224 | Token &Tok = ResultToks[i]; |
| 225 | Tok.setLocation(SM.createInstantiationLoc(Tok.getLocation(), |
| 226 | curInst, curInst, |
| 227 | Tok.getLength())); |
| 228 | } |
| 229 | } |
| 230 | |
Reid Spencer | 5f016e2 | 2007-07-11 17:01:13 +0000 | [diff] [blame] | 231 | // If any tokens were substituted from the argument, the whitespace |
| 232 | // before the first token should match the whitespace of the arg |
| 233 | // identifier. |
Chris Lattner | d217773 | 2007-07-20 16:59:19 +0000 | [diff] [blame] | 234 | ResultToks[FirstResult].setFlagValue(Token::LeadingSpace, |
Reid Spencer | 5f016e2 | 2007-07-11 17:01:13 +0000 | [diff] [blame] | 235 | CurTok.hasLeadingSpace() || |
| 236 | NextTokGetsSpace); |
| 237 | NextTokGetsSpace = false; |
| 238 | } else { |
| 239 | // If this is an empty argument, and if there was whitespace before the |
| 240 | // formal token, make sure the next token gets whitespace before it. |
| 241 | NextTokGetsSpace = CurTok.hasLeadingSpace(); |
| 242 | } |
| 243 | continue; |
| 244 | } |
Mike Stump | 1eb4433 | 2009-09-09 15:08:12 +0000 | [diff] [blame] | 245 | |
Reid Spencer | 5f016e2 | 2007-07-11 17:01:13 +0000 | [diff] [blame] | 246 | // Okay, we have a token that is either the LHS or RHS of a paste (##) |
| 247 | // argument. It gets substituted as its non-pre-expanded tokens. |
Chris Lattner | d217773 | 2007-07-20 16:59:19 +0000 | [diff] [blame] | 248 | const Token *ArgToks = ActualArgs->getUnexpArgument(ArgNo); |
Reid Spencer | 5f016e2 | 2007-07-11 17:01:13 +0000 | [diff] [blame] | 249 | unsigned NumToks = MacroArgs::getArgLength(ArgToks); |
| 250 | if (NumToks) { // Not an empty argument? |
Chris Lattner | 71a3a8d | 2008-01-29 07:54:23 +0000 | [diff] [blame] | 251 | // If this is the GNU ", ## __VA_ARG__" extension, and we just learned |
| 252 | // that __VA_ARG__ expands to multiple tokens, avoid a pasting error when |
| 253 | // the expander trys to paste ',' with the first token of the __VA_ARG__ |
| 254 | // expansion. |
| 255 | if (PasteBefore && ResultToks.size() >= 2 && |
| 256 | ResultToks[ResultToks.size()-2].is(tok::comma) && |
| 257 | (unsigned)ArgNo == Macro->getNumArgs()-1 && |
| 258 | Macro->isVariadic()) { |
| 259 | // Remove the paste operator, report use of the extension. |
| 260 | PP.Diag(ResultToks.back().getLocation(), diag::ext_paste_comma); |
| 261 | ResultToks.pop_back(); |
| 262 | } |
Mike Stump | 1eb4433 | 2009-09-09 15:08:12 +0000 | [diff] [blame] | 263 | |
Reid Spencer | 5f016e2 | 2007-07-11 17:01:13 +0000 | [diff] [blame] | 264 | ResultToks.append(ArgToks, ArgToks+NumToks); |
Mike Stump | 1eb4433 | 2009-09-09 15:08:12 +0000 | [diff] [blame] | 265 | |
Argyrios Kyrtzidis | b73377e | 2011-07-07 03:40:34 +0000 | [diff] [blame^] | 266 | if(InstantiateLocStart.isValid()) { |
| 267 | SourceManager &SM = PP.getSourceManager(); |
| 268 | SourceLocation curInst = |
| 269 | getMacroExpansionLocation(CurTok.getLocation()); |
| 270 | assert(curInst.isValid() && |
| 271 | "Expected arg identifier to come from definition"); |
| 272 | for (unsigned i = ResultToks.size() - NumToks, e = ResultToks.size(); |
| 273 | i != e; ++i) { |
| 274 | Token &Tok = ResultToks[i]; |
| 275 | Tok.setLocation(SM.createInstantiationLoc(Tok.getLocation(), |
| 276 | curInst, curInst, |
| 277 | Tok.getLength())); |
| 278 | } |
| 279 | } |
| 280 | |
Chris Lattner | 442a661 | 2009-05-25 16:23:08 +0000 | [diff] [blame] | 281 | // If this token (the macro argument) was supposed to get leading |
| 282 | // whitespace, transfer this information onto the first token of the |
| 283 | // expansion. |
| 284 | // |
| 285 | // Do not do this if the paste operator occurs before the macro argument, |
| 286 | // as in "A ## MACROARG". In valid code, the first token will get |
| 287 | // smooshed onto the preceding one anyway (forming AMACROARG). In |
| 288 | // assembler-with-cpp mode, invalid pastes are allowed through: in this |
| 289 | // case, we do not want the extra whitespace to be added. For example, |
| 290 | // we want ". ## foo" -> ".foo" not ". foo". |
| 291 | if ((CurTok.hasLeadingSpace() || NextTokGetsSpace) && |
| 292 | !PasteBefore) |
| 293 | ResultToks[ResultToks.size()-NumToks].setFlag(Token::LeadingSpace); |
Mike Stump | 1eb4433 | 2009-09-09 15:08:12 +0000 | [diff] [blame] | 294 | |
Chris Lattner | 442a661 | 2009-05-25 16:23:08 +0000 | [diff] [blame] | 295 | NextTokGetsSpace = false; |
Reid Spencer | 5f016e2 | 2007-07-11 17:01:13 +0000 | [diff] [blame] | 296 | continue; |
| 297 | } |
Mike Stump | 1eb4433 | 2009-09-09 15:08:12 +0000 | [diff] [blame] | 298 | |
Reid Spencer | 5f016e2 | 2007-07-11 17:01:13 +0000 | [diff] [blame] | 299 | // If an empty argument is on the LHS or RHS of a paste, the standard (C99 |
| 300 | // 6.10.3.3p2,3) calls for a bunch of placemarker stuff to occur. We |
| 301 | // implement this by eating ## operators when a LHS or RHS expands to |
| 302 | // empty. |
| 303 | NextTokGetsSpace |= CurTok.hasLeadingSpace(); |
| 304 | if (PasteAfter) { |
| 305 | // Discard the argument token and skip (don't copy to the expansion |
| 306 | // buffer) the paste operator after it. |
Chris Lattner | 8d89643 | 2008-03-09 02:07:49 +0000 | [diff] [blame] | 307 | NextTokGetsSpace |= Tokens[i+1].hasLeadingSpace(); |
Reid Spencer | 5f016e2 | 2007-07-11 17:01:13 +0000 | [diff] [blame] | 308 | ++i; |
| 309 | continue; |
| 310 | } |
Mike Stump | 1eb4433 | 2009-09-09 15:08:12 +0000 | [diff] [blame] | 311 | |
Reid Spencer | 5f016e2 | 2007-07-11 17:01:13 +0000 | [diff] [blame] | 312 | // If this is on the RHS of a paste operator, we've already copied the |
| 313 | // paste operator to the ResultToks list. Remove it. |
Chris Lattner | 22f6bbc | 2007-10-09 18:02:16 +0000 | [diff] [blame] | 314 | assert(PasteBefore && ResultToks.back().is(tok::hashhash)); |
Reid Spencer | 5f016e2 | 2007-07-11 17:01:13 +0000 | [diff] [blame] | 315 | NextTokGetsSpace |= ResultToks.back().hasLeadingSpace(); |
| 316 | ResultToks.pop_back(); |
Mike Stump | 1eb4433 | 2009-09-09 15:08:12 +0000 | [diff] [blame] | 317 | |
Reid Spencer | 5f016e2 | 2007-07-11 17:01:13 +0000 | [diff] [blame] | 318 | // If this is the __VA_ARGS__ token, and if the argument wasn't provided, |
| 319 | // and if the macro had at least one real argument, and if the token before |
| 320 | // the ## was a comma, remove the comma. |
| 321 | if ((unsigned)ArgNo == Macro->getNumArgs()-1 && // is __VA_ARGS__ |
| 322 | ActualArgs->isVarargsElidedUse() && // Argument elided. |
Chris Lattner | 22f6bbc | 2007-10-09 18:02:16 +0000 | [diff] [blame] | 323 | !ResultToks.empty() && ResultToks.back().is(tok::comma)) { |
Reid Spencer | 5f016e2 | 2007-07-11 17:01:13 +0000 | [diff] [blame] | 324 | // Never add a space, even if the comma, ##, or arg had a space. |
| 325 | NextTokGetsSpace = false; |
Chris Lattner | 71a3a8d | 2008-01-29 07:54:23 +0000 | [diff] [blame] | 326 | // Remove the paste operator, report use of the extension. |
| 327 | PP.Diag(ResultToks.back().getLocation(), diag::ext_paste_comma); |
Reid Spencer | 5f016e2 | 2007-07-11 17:01:13 +0000 | [diff] [blame] | 328 | ResultToks.pop_back(); |
Chris Lattner | 496af57 | 2010-08-21 00:27:00 +0000 | [diff] [blame] | 329 | |
| 330 | // If the comma was right after another paste (e.g. "X##,##__VA_ARGS__"), |
| 331 | // then removal of the comma should produce a placemarker token (in C99 |
| 332 | // terms) which we model by popping off the previous ##, giving us a plain |
| 333 | // "X" when __VA_ARGS__ is empty. |
| 334 | if (!ResultToks.empty() && ResultToks.back().is(tok::hashhash)) |
| 335 | ResultToks.pop_back(); |
Reid Spencer | 5f016e2 | 2007-07-11 17:01:13 +0000 | [diff] [blame] | 336 | } |
| 337 | continue; |
| 338 | } |
Mike Stump | 1eb4433 | 2009-09-09 15:08:12 +0000 | [diff] [blame] | 339 | |
Chris Lattner | 8d89643 | 2008-03-09 02:07:49 +0000 | [diff] [blame] | 340 | // If anything changed, install this as the new Tokens list. |
Reid Spencer | 5f016e2 | 2007-07-11 17:01:13 +0000 | [diff] [blame] | 341 | if (MadeChange) { |
Chris Lattner | 32fca72 | 2009-03-04 06:50:57 +0000 | [diff] [blame] | 342 | assert(!OwnsTokens && "This would leak if we already own the token list"); |
Reid Spencer | 5f016e2 | 2007-07-11 17:01:13 +0000 | [diff] [blame] | 343 | // This is deleted in the dtor. |
Chris Lattner | 8d89643 | 2008-03-09 02:07:49 +0000 | [diff] [blame] | 344 | NumTokens = ResultToks.size(); |
Argyrios Kyrtzidis | 5b3284a | 2011-06-29 22:20:11 +0000 | [diff] [blame] | 345 | // The tokens will be added to Preprocessor's cache and will be removed |
| 346 | // when this TokenLexer finishes lexing them. |
| 347 | Tokens = PP.cacheMacroExpandedTokens(this, ResultToks); |
Mike Stump | 1eb4433 | 2009-09-09 15:08:12 +0000 | [diff] [blame] | 348 | |
Argyrios Kyrtzidis | 5b3284a | 2011-06-29 22:20:11 +0000 | [diff] [blame] | 349 | // The preprocessor cache of macro expanded tokens owns these tokens,not us. |
Chris Lattner | 32fca72 | 2009-03-04 06:50:57 +0000 | [diff] [blame] | 350 | OwnsTokens = false; |
Reid Spencer | 5f016e2 | 2007-07-11 17:01:13 +0000 | [diff] [blame] | 351 | } |
| 352 | } |
| 353 | |
| 354 | /// Lex - Lex and return a token from this macro stream. |
| 355 | /// |
Chris Lattner | 1543e9c | 2008-03-09 02:18:51 +0000 | [diff] [blame] | 356 | void TokenLexer::Lex(Token &Tok) { |
Reid Spencer | 5f016e2 | 2007-07-11 17:01:13 +0000 | [diff] [blame] | 357 | // Lexing off the end of the macro, pop this macro off the expansion stack. |
| 358 | if (isAtEnd()) { |
| 359 | // If this is a macro (not a token stream), mark the macro enabled now |
| 360 | // that it is no longer being expanded. |
| 361 | if (Macro) Macro->EnableMacro(); |
| 362 | |
| 363 | // Pop this context off the preprocessors lexer stack and get the next |
| 364 | // token. This will delete "this" so remember the PP instance var. |
| 365 | Preprocessor &PPCache = PP; |
Chris Lattner | fde2bf9 | 2008-03-09 03:04:16 +0000 | [diff] [blame] | 366 | if (PP.HandleEndOfTokenLexer(Tok)) |
Reid Spencer | 5f016e2 | 2007-07-11 17:01:13 +0000 | [diff] [blame] | 367 | return; |
| 368 | |
Chris Lattner | fde2bf9 | 2008-03-09 03:04:16 +0000 | [diff] [blame] | 369 | // HandleEndOfTokenLexer may not return a token. If it doesn't, lex |
| 370 | // whatever is next. |
Reid Spencer | 5f016e2 | 2007-07-11 17:01:13 +0000 | [diff] [blame] | 371 | return PPCache.Lex(Tok); |
| 372 | } |
Mike Stump | 1eb4433 | 2009-09-09 15:08:12 +0000 | [diff] [blame] | 373 | |
Reid Spencer | 5f016e2 | 2007-07-11 17:01:13 +0000 | [diff] [blame] | 374 | // If this is the first token of the expanded result, we inherit spacing |
| 375 | // properties later. |
| 376 | bool isFirstToken = CurToken == 0; |
Mike Stump | 1eb4433 | 2009-09-09 15:08:12 +0000 | [diff] [blame] | 377 | |
Reid Spencer | 5f016e2 | 2007-07-11 17:01:13 +0000 | [diff] [blame] | 378 | // Get the next token to return. |
Chris Lattner | 8d89643 | 2008-03-09 02:07:49 +0000 | [diff] [blame] | 379 | Tok = Tokens[CurToken++]; |
Mike Stump | 1eb4433 | 2009-09-09 15:08:12 +0000 | [diff] [blame] | 380 | |
Chris Lattner | 332ee08 | 2009-04-19 20:29:42 +0000 | [diff] [blame] | 381 | bool TokenIsFromPaste = false; |
Mike Stump | 1eb4433 | 2009-09-09 15:08:12 +0000 | [diff] [blame] | 382 | |
Reid Spencer | 5f016e2 | 2007-07-11 17:01:13 +0000 | [diff] [blame] | 383 | // If this token is followed by a token paste (##) operator, paste the tokens! |
Chris Lattner | b11e43c | 2011-06-14 18:19:37 +0000 | [diff] [blame] | 384 | // Note that ## is a normal token when not expanding a macro. |
| 385 | if (!isAtEnd() && Tokens[CurToken].is(tok::hashhash) && Macro) { |
Chris Lattner | 04f2bd8 | 2009-12-04 06:14:03 +0000 | [diff] [blame] | 386 | // When handling the microsoft /##/ extension, the final token is |
| 387 | // returned by PasteTokens, not the pasted token. |
| 388 | if (PasteTokens(Tok)) |
Chris Lattner | 3f1cc83 | 2008-02-07 06:03:59 +0000 | [diff] [blame] | 389 | return; |
Kovarththanan Rajaratnam | 65cc1e8 | 2010-03-13 08:53:33 +0000 | [diff] [blame] | 390 | |
Chris Lattner | 04f2bd8 | 2009-12-04 06:14:03 +0000 | [diff] [blame] | 391 | TokenIsFromPaste = true; |
Mike Stump | 1eb4433 | 2009-09-09 15:08:12 +0000 | [diff] [blame] | 392 | } |
Reid Spencer | 5f016e2 | 2007-07-11 17:01:13 +0000 | [diff] [blame] | 393 | |
Argyrios Kyrtzidis | b73377e | 2011-07-07 03:40:34 +0000 | [diff] [blame^] | 394 | SourceManager &SM = PP.getSourceManager(); |
Reid Spencer | 5f016e2 | 2007-07-11 17:01:13 +0000 | [diff] [blame] | 395 | // The token's current location indicate where the token was lexed from. We |
| 396 | // need this information to compute the spelling of the token, but any |
| 397 | // diagnostics for the expanded token should appear as if they came from |
| 398 | // InstantiationLoc. Pull this information together into a new SourceLocation |
| 399 | // that captures all of this. |
Argyrios Kyrtzidis | b73377e | 2011-07-07 03:40:34 +0000 | [diff] [blame^] | 400 | if (InstantiateLocStart.isValid() && // Don't do this for token streams. |
| 401 | // Check that the token's location was not already set properly. |
| 402 | SM.isBeforeInSourceLocationOffset(Tok.getLocation(), |
| 403 | MacroStartSLocOffset)) { |
| 404 | SourceLocation instLoc; |
| 405 | if (Tok.is(tok::comment)) { |
| 406 | instLoc = SM.createInstantiationLoc(Tok.getLocation(), |
| 407 | InstantiateLocStart, |
| 408 | InstantiateLocEnd, |
| 409 | Tok.getLength()); |
| 410 | } else { |
| 411 | instLoc = getMacroExpansionLocation(Tok.getLocation()); |
| 412 | assert(instLoc.isValid() && |
| 413 | "Location for token not coming from definition was not set!"); |
| 414 | } |
| 415 | |
| 416 | Tok.setLocation(instLoc); |
Reid Spencer | 5f016e2 | 2007-07-11 17:01:13 +0000 | [diff] [blame] | 417 | } |
Mike Stump | 1eb4433 | 2009-09-09 15:08:12 +0000 | [diff] [blame] | 418 | |
Reid Spencer | 5f016e2 | 2007-07-11 17:01:13 +0000 | [diff] [blame] | 419 | // If this is the first token, set the lexical properties of the token to |
| 420 | // match the lexical properties of the macro identifier. |
| 421 | if (isFirstToken) { |
Chris Lattner | d217773 | 2007-07-20 16:59:19 +0000 | [diff] [blame] | 422 | Tok.setFlagValue(Token::StartOfLine , AtStartOfLine); |
| 423 | Tok.setFlagValue(Token::LeadingSpace, HasLeadingSpace); |
Reid Spencer | 5f016e2 | 2007-07-11 17:01:13 +0000 | [diff] [blame] | 424 | } |
Mike Stump | 1eb4433 | 2009-09-09 15:08:12 +0000 | [diff] [blame] | 425 | |
Reid Spencer | 5f016e2 | 2007-07-11 17:01:13 +0000 | [diff] [blame] | 426 | // Handle recursive expansion! |
Argyrios Kyrtzidis | 6f421f9 | 2009-05-22 21:09:31 +0000 | [diff] [blame] | 427 | if (!Tok.isAnnotation() && Tok.getIdentifierInfo() != 0) { |
Chris Lattner | 863c486 | 2009-01-23 18:35:48 +0000 | [diff] [blame] | 428 | // Change the kind of this identifier to the appropriate token kind, e.g. |
| 429 | // turning "for" into a keyword. |
Argyrios Kyrtzidis | 6f421f9 | 2009-05-22 21:09:31 +0000 | [diff] [blame] | 430 | IdentifierInfo *II = Tok.getIdentifierInfo(); |
Chris Lattner | 863c486 | 2009-01-23 18:35:48 +0000 | [diff] [blame] | 431 | Tok.setKind(II->getTokenID()); |
Mike Stump | 1eb4433 | 2009-09-09 15:08:12 +0000 | [diff] [blame] | 432 | |
Chris Lattner | 332ee08 | 2009-04-19 20:29:42 +0000 | [diff] [blame] | 433 | // If this identifier was poisoned and from a paste, emit an error. This |
| 434 | // won't be handled by Preprocessor::HandleIdentifier because this is coming |
| 435 | // from a macro expansion. |
| 436 | if (II->isPoisoned() && TokenIsFromPaste) { |
John Wiegley | 28bbe4b | 2011-04-28 01:08:34 +0000 | [diff] [blame] | 437 | PP.HandlePoisonedIdentifier(Tok); |
Chris Lattner | 332ee08 | 2009-04-19 20:29:42 +0000 | [diff] [blame] | 438 | } |
Mike Stump | 1eb4433 | 2009-09-09 15:08:12 +0000 | [diff] [blame] | 439 | |
Chris Lattner | 863c486 | 2009-01-23 18:35:48 +0000 | [diff] [blame] | 440 | if (!DisableMacroExpansion && II->isHandleIdentifierCase()) |
| 441 | PP.HandleIdentifier(Tok); |
| 442 | } |
Reid Spencer | 5f016e2 | 2007-07-11 17:01:13 +0000 | [diff] [blame] | 443 | |
| 444 | // Otherwise, return a normal token. |
| 445 | } |
| 446 | |
| 447 | /// PasteTokens - Tok is the LHS of a ## operator, and CurToken is the ## |
| 448 | /// operator. Read the ## and RHS, and paste the LHS/RHS together. If there |
Nico Weber | 48002c8 | 2008-09-29 00:25:48 +0000 | [diff] [blame] | 449 | /// are more ## after it, chomp them iteratively. Return the result as Tok. |
Chris Lattner | 3f1cc83 | 2008-02-07 06:03:59 +0000 | [diff] [blame] | 450 | /// If this returns true, the caller should immediately return the token. |
Chris Lattner | 1543e9c | 2008-03-09 02:18:51 +0000 | [diff] [blame] | 451 | bool TokenLexer::PasteTokens(Token &Tok) { |
Kovarththanan Rajaratnam | 1935754 | 2010-03-13 10:17:05 +0000 | [diff] [blame] | 452 | llvm::SmallString<128> Buffer; |
Chris Lattner | 47246be | 2009-01-26 19:29:26 +0000 | [diff] [blame] | 453 | const char *ResultTokStrPtr = 0; |
Argyrios Kyrtzidis | b73377e | 2011-07-07 03:40:34 +0000 | [diff] [blame^] | 454 | SourceLocation PasteOpLoc; |
Reid Spencer | 5f016e2 | 2007-07-11 17:01:13 +0000 | [diff] [blame] | 455 | do { |
| 456 | // Consume the ## operator. |
Argyrios Kyrtzidis | b73377e | 2011-07-07 03:40:34 +0000 | [diff] [blame^] | 457 | PasteOpLoc = Tokens[CurToken].getLocation(); |
Reid Spencer | 5f016e2 | 2007-07-11 17:01:13 +0000 | [diff] [blame] | 458 | ++CurToken; |
| 459 | assert(!isAtEnd() && "No token on the RHS of a paste operator!"); |
Mike Stump | 1eb4433 | 2009-09-09 15:08:12 +0000 | [diff] [blame] | 460 | |
Reid Spencer | 5f016e2 | 2007-07-11 17:01:13 +0000 | [diff] [blame] | 461 | // Get the RHS token. |
Chris Lattner | 8d89643 | 2008-03-09 02:07:49 +0000 | [diff] [blame] | 462 | const Token &RHS = Tokens[CurToken]; |
Mike Stump | 1eb4433 | 2009-09-09 15:08:12 +0000 | [diff] [blame] | 463 | |
Reid Spencer | 5f016e2 | 2007-07-11 17:01:13 +0000 | [diff] [blame] | 464 | // Allocate space for the result token. This is guaranteed to be enough for |
Chris Lattner | 1fa4953 | 2009-03-08 08:08:45 +0000 | [diff] [blame] | 465 | // the two tokens. |
| 466 | Buffer.resize(Tok.getLength() + RHS.getLength()); |
Mike Stump | 1eb4433 | 2009-09-09 15:08:12 +0000 | [diff] [blame] | 467 | |
Reid Spencer | 5f016e2 | 2007-07-11 17:01:13 +0000 | [diff] [blame] | 468 | // Get the spelling of the LHS token in Buffer. |
| 469 | const char *BufPtr = &Buffer[0]; |
Douglas Gregor | 453091c | 2010-03-16 22:30:13 +0000 | [diff] [blame] | 470 | bool Invalid = false; |
| 471 | unsigned LHSLen = PP.getSpelling(Tok, BufPtr, &Invalid); |
Reid Spencer | 5f016e2 | 2007-07-11 17:01:13 +0000 | [diff] [blame] | 472 | if (BufPtr != &Buffer[0]) // Really, we want the chars in Buffer! |
| 473 | memcpy(&Buffer[0], BufPtr, LHSLen); |
Douglas Gregor | 453091c | 2010-03-16 22:30:13 +0000 | [diff] [blame] | 474 | if (Invalid) |
| 475 | return true; |
| 476 | |
Reid Spencer | 5f016e2 | 2007-07-11 17:01:13 +0000 | [diff] [blame] | 477 | BufPtr = &Buffer[LHSLen]; |
Douglas Gregor | 453091c | 2010-03-16 22:30:13 +0000 | [diff] [blame] | 478 | unsigned RHSLen = PP.getSpelling(RHS, BufPtr, &Invalid); |
| 479 | if (Invalid) |
| 480 | return true; |
Reid Spencer | 5f016e2 | 2007-07-11 17:01:13 +0000 | [diff] [blame] | 481 | if (BufPtr != &Buffer[LHSLen]) // Really, we want the chars in Buffer! |
| 482 | memcpy(&Buffer[LHSLen], BufPtr, RHSLen); |
Mike Stump | 1eb4433 | 2009-09-09 15:08:12 +0000 | [diff] [blame] | 483 | |
Reid Spencer | 5f016e2 | 2007-07-11 17:01:13 +0000 | [diff] [blame] | 484 | // Trim excess space. |
Chris Lattner | 1fa4953 | 2009-03-08 08:08:45 +0000 | [diff] [blame] | 485 | Buffer.resize(LHSLen+RHSLen); |
Mike Stump | 1eb4433 | 2009-09-09 15:08:12 +0000 | [diff] [blame] | 486 | |
Reid Spencer | 5f016e2 | 2007-07-11 17:01:13 +0000 | [diff] [blame] | 487 | // Plop the pasted result (including the trailing newline and null) into a |
| 488 | // scratch buffer where we can lex it. |
Chris Lattner | 47246be | 2009-01-26 19:29:26 +0000 | [diff] [blame] | 489 | Token ResultTokTmp; |
| 490 | ResultTokTmp.startToken(); |
Mike Stump | 1eb4433 | 2009-09-09 15:08:12 +0000 | [diff] [blame] | 491 | |
Chris Lattner | 47246be | 2009-01-26 19:29:26 +0000 | [diff] [blame] | 492 | // Claim that the tmp token is a string_literal so that we can get the |
Chris Lattner | 917f143 | 2009-12-23 21:29:53 +0000 | [diff] [blame] | 493 | // character pointer back from CreateString in getLiteralData(). |
Chris Lattner | 47246be | 2009-01-26 19:29:26 +0000 | [diff] [blame] | 494 | ResultTokTmp.setKind(tok::string_literal); |
| 495 | PP.CreateString(&Buffer[0], Buffer.size(), ResultTokTmp); |
| 496 | SourceLocation ResultTokLoc = ResultTokTmp.getLocation(); |
| 497 | ResultTokStrPtr = ResultTokTmp.getLiteralData(); |
| 498 | |
Reid Spencer | 5f016e2 | 2007-07-11 17:01:13 +0000 | [diff] [blame] | 499 | // Lex the resultant pasted token into Result. |
Chris Lattner | d217773 | 2007-07-20 16:59:19 +0000 | [diff] [blame] | 500 | Token Result; |
Mike Stump | 1eb4433 | 2009-09-09 15:08:12 +0000 | [diff] [blame] | 501 | |
Abramo Bagnara | c4bf2b9 | 2010-12-22 08:23:18 +0000 | [diff] [blame] | 502 | if (Tok.isAnyIdentifier() && RHS.isAnyIdentifier()) { |
Reid Spencer | 5f016e2 | 2007-07-11 17:01:13 +0000 | [diff] [blame] | 503 | // Common paste case: identifier+identifier = identifier. Avoid creating |
| 504 | // a lexer and other overhead. |
| 505 | PP.IncrementPasteCounter(true); |
| 506 | Result.startToken(); |
Abramo Bagnara | c4bf2b9 | 2010-12-22 08:23:18 +0000 | [diff] [blame] | 507 | Result.setKind(tok::raw_identifier); |
| 508 | Result.setRawIdentifierData(ResultTokStrPtr); |
Reid Spencer | 5f016e2 | 2007-07-11 17:01:13 +0000 | [diff] [blame] | 509 | Result.setLocation(ResultTokLoc); |
| 510 | Result.setLength(LHSLen+RHSLen); |
| 511 | } else { |
| 512 | PP.IncrementPasteCounter(false); |
Mike Stump | 1eb4433 | 2009-09-09 15:08:12 +0000 | [diff] [blame] | 513 | |
Chris Lattner | bcc2a67 | 2009-01-19 06:46:35 +0000 | [diff] [blame] | 514 | assert(ResultTokLoc.isFileID() && |
| 515 | "Should be a raw location into scratch buffer"); |
Reid Spencer | 5f016e2 | 2007-07-11 17:01:13 +0000 | [diff] [blame] | 516 | SourceManager &SourceMgr = PP.getSourceManager(); |
Chris Lattner | 47246be | 2009-01-26 19:29:26 +0000 | [diff] [blame] | 517 | FileID LocFileID = SourceMgr.getFileID(ResultTokLoc); |
Mike Stump | 1eb4433 | 2009-09-09 15:08:12 +0000 | [diff] [blame] | 518 | |
Douglas Gregor | f715ca1 | 2010-03-16 00:06:06 +0000 | [diff] [blame] | 519 | bool Invalid = false; |
Douglas Gregor | aea67db | 2010-03-15 22:54:52 +0000 | [diff] [blame] | 520 | const char *ScratchBufStart |
Benjamin Kramer | f6ac97b | 2010-03-16 14:14:31 +0000 | [diff] [blame] | 521 | = SourceMgr.getBufferData(LocFileID, &Invalid).data(); |
Douglas Gregor | f715ca1 | 2010-03-16 00:06:06 +0000 | [diff] [blame] | 522 | if (Invalid) |
Douglas Gregor | aea67db | 2010-03-15 22:54:52 +0000 | [diff] [blame] | 523 | return false; |
Mike Stump | 1eb4433 | 2009-09-09 15:08:12 +0000 | [diff] [blame] | 524 | |
Chris Lattner | bcc2a67 | 2009-01-19 06:46:35 +0000 | [diff] [blame] | 525 | // Make a lexer to lex this string from. Lex just this one token. |
Reid Spencer | 5f016e2 | 2007-07-11 17:01:13 +0000 | [diff] [blame] | 526 | // Make a lexer object so that we lex and expand the paste result. |
Chris Lattner | 47246be | 2009-01-26 19:29:26 +0000 | [diff] [blame] | 527 | Lexer TL(SourceMgr.getLocForStartOfFile(LocFileID), |
| 528 | PP.getLangOptions(), ScratchBufStart, |
Chris Lattner | 1fa4953 | 2009-03-08 08:08:45 +0000 | [diff] [blame] | 529 | ResultTokStrPtr, ResultTokStrPtr+LHSLen+RHSLen); |
Mike Stump | 1eb4433 | 2009-09-09 15:08:12 +0000 | [diff] [blame] | 530 | |
Reid Spencer | 5f016e2 | 2007-07-11 17:01:13 +0000 | [diff] [blame] | 531 | // Lex a token in raw mode. This way it won't look up identifiers |
| 532 | // automatically, lexing off the end will return an eof token, and |
| 533 | // warnings are disabled. This returns true if the result token is the |
| 534 | // entire buffer. |
Chris Lattner | 1fa4953 | 2009-03-08 08:08:45 +0000 | [diff] [blame] | 535 | bool isInvalid = !TL.LexFromRawLexer(Result); |
Mike Stump | 1eb4433 | 2009-09-09 15:08:12 +0000 | [diff] [blame] | 536 | |
Reid Spencer | 5f016e2 | 2007-07-11 17:01:13 +0000 | [diff] [blame] | 537 | // If we got an EOF token, we didn't form even ONE token. For example, we |
| 538 | // did "/ ## /" to get "//". |
Chris Lattner | 1fa4953 | 2009-03-08 08:08:45 +0000 | [diff] [blame] | 539 | isInvalid |= Result.is(tok::eof); |
Mike Stump | 1eb4433 | 2009-09-09 15:08:12 +0000 | [diff] [blame] | 540 | |
Chris Lattner | 1fa4953 | 2009-03-08 08:08:45 +0000 | [diff] [blame] | 541 | // If pasting the two tokens didn't form a full new token, this is an |
| 542 | // error. This occurs with "x ## +" and other stuff. Return with Tok |
| 543 | // unmodified and with RHS as the next token to lex. |
| 544 | if (isInvalid) { |
| 545 | // Test for the Microsoft extension of /##/ turning into // here on the |
| 546 | // error path. |
Mike Stump | 1eb4433 | 2009-09-09 15:08:12 +0000 | [diff] [blame] | 547 | if (PP.getLangOptions().Microsoft && Tok.is(tok::slash) && |
Chris Lattner | 1fa4953 | 2009-03-08 08:08:45 +0000 | [diff] [blame] | 548 | RHS.is(tok::slash)) { |
| 549 | HandleMicrosoftCommentPaste(Tok); |
| 550 | return true; |
| 551 | } |
Mike Stump | 1eb4433 | 2009-09-09 15:08:12 +0000 | [diff] [blame] | 552 | |
Chris Lattner | 920bf45 | 2010-07-17 16:24:30 +0000 | [diff] [blame] | 553 | // Do not emit the error when preprocessing assembler code. |
Chris Lattner | 3c17b8a | 2009-05-28 05:39:39 +0000 | [diff] [blame] | 554 | if (!PP.getLangOptions().AsmPreprocessor) { |
| 555 | // Explicitly convert the token location to have proper instantiation |
| 556 | // information so that the user knows where it came from. |
| 557 | SourceManager &SM = PP.getSourceManager(); |
Chris Lattner | b11e43c | 2011-06-14 18:19:37 +0000 | [diff] [blame] | 558 | SourceLocation Loc = |
| 559 | SM.createInstantiationLoc(PasteOpLoc, InstantiateLocStart, |
| 560 | InstantiateLocEnd, 2); |
Chris Lattner | 920bf45 | 2010-07-17 16:24:30 +0000 | [diff] [blame] | 561 | // If we're in microsoft extensions mode, downgrade this from a hard |
| 562 | // error to a warning that defaults to an error. This allows |
| 563 | // disabling it. |
| 564 | PP.Diag(Loc, |
| 565 | PP.getLangOptions().Microsoft ? diag::err_pp_bad_paste_ms |
| 566 | : diag::err_pp_bad_paste) |
Benjamin Kramer | 476d8b8 | 2010-08-11 14:47:12 +0000 | [diff] [blame] | 567 | << Buffer.str(); |
Chris Lattner | 3c17b8a | 2009-05-28 05:39:39 +0000 | [diff] [blame] | 568 | } |
Mike Stump | 1eb4433 | 2009-09-09 15:08:12 +0000 | [diff] [blame] | 569 | |
Chris Lattner | f1538de | 2009-03-23 17:32:45 +0000 | [diff] [blame] | 570 | // Do not consume the RHS. |
| 571 | --CurToken; |
Chris Lattner | 1fa4953 | 2009-03-08 08:08:45 +0000 | [diff] [blame] | 572 | } |
Mike Stump | 1eb4433 | 2009-09-09 15:08:12 +0000 | [diff] [blame] | 573 | |
Chris Lattner | 1fa4953 | 2009-03-08 08:08:45 +0000 | [diff] [blame] | 574 | // Turn ## into 'unknown' to avoid # ## # from looking like a paste |
| 575 | // operator. |
| 576 | if (Result.is(tok::hashhash)) |
| 577 | Result.setKind(tok::unknown); |
Reid Spencer | 5f016e2 | 2007-07-11 17:01:13 +0000 | [diff] [blame] | 578 | } |
Mike Stump | 1eb4433 | 2009-09-09 15:08:12 +0000 | [diff] [blame] | 579 | |
Reid Spencer | 5f016e2 | 2007-07-11 17:01:13 +0000 | [diff] [blame] | 580 | // Transfer properties of the LHS over the the Result. |
Chris Lattner | d217773 | 2007-07-20 16:59:19 +0000 | [diff] [blame] | 581 | Result.setFlagValue(Token::StartOfLine , Tok.isAtStartOfLine()); |
| 582 | Result.setFlagValue(Token::LeadingSpace, Tok.hasLeadingSpace()); |
Argyrios Kyrtzidis | b73377e | 2011-07-07 03:40:34 +0000 | [diff] [blame^] | 583 | |
Reid Spencer | 5f016e2 | 2007-07-11 17:01:13 +0000 | [diff] [blame] | 584 | // Finally, replace LHS with the result, consume the RHS, and iterate. |
| 585 | ++CurToken; |
| 586 | Tok = Result; |
Chris Lattner | 8d89643 | 2008-03-09 02:07:49 +0000 | [diff] [blame] | 587 | } while (!isAtEnd() && Tokens[CurToken].is(tok::hashhash)); |
Mike Stump | 1eb4433 | 2009-09-09 15:08:12 +0000 | [diff] [blame] | 588 | |
Argyrios Kyrtzidis | b73377e | 2011-07-07 03:40:34 +0000 | [diff] [blame^] | 589 | // The token's current location indicate where the token was lexed from. We |
| 590 | // need this information to compute the spelling of the token, but any |
| 591 | // diagnostics for the expanded token should appear as if the token was |
| 592 | // instantiated from the (##) operator. Pull this information together into |
| 593 | // a new SourceLocation that captures all of this. |
| 594 | if (InstantiateLocStart.isValid()) { |
| 595 | SourceManager &SM = PP.getSourceManager(); |
| 596 | SourceLocation pasteLocInst = |
| 597 | getMacroExpansionLocation(PasteOpLoc); |
| 598 | assert(pasteLocInst.isValid() && |
| 599 | "Expected '##' to come from definition"); |
| 600 | |
| 601 | Tok.setLocation(SM.createInstantiationLoc(Tok.getLocation(), |
| 602 | pasteLocInst, |
| 603 | pasteLocInst, |
| 604 | Tok.getLength())); |
| 605 | } |
| 606 | |
Reid Spencer | 5f016e2 | 2007-07-11 17:01:13 +0000 | [diff] [blame] | 607 | // Now that we got the result token, it will be subject to expansion. Since |
| 608 | // token pasting re-lexes the result token in raw mode, identifier information |
| 609 | // isn't looked up. As such, if the result is an identifier, look up id info. |
Abramo Bagnara | c4bf2b9 | 2010-12-22 08:23:18 +0000 | [diff] [blame] | 610 | if (Tok.is(tok::raw_identifier)) { |
Reid Spencer | 5f016e2 | 2007-07-11 17:01:13 +0000 | [diff] [blame] | 611 | // Look up the identifier info for the token. We disabled identifier lookup |
| 612 | // by saying we're skipping contents, so we need to do this manually. |
Abramo Bagnara | c4bf2b9 | 2010-12-22 08:23:18 +0000 | [diff] [blame] | 613 | PP.LookUpIdentifierInfo(Tok); |
Reid Spencer | 5f016e2 | 2007-07-11 17:01:13 +0000 | [diff] [blame] | 614 | } |
Chris Lattner | 3f1cc83 | 2008-02-07 06:03:59 +0000 | [diff] [blame] | 615 | return false; |
Reid Spencer | 5f016e2 | 2007-07-11 17:01:13 +0000 | [diff] [blame] | 616 | } |
| 617 | |
| 618 | /// isNextTokenLParen - If the next token lexed will pop this macro off the |
| 619 | /// expansion stack, return 2. If the next unexpanded token is a '(', return |
| 620 | /// 1, otherwise return 0. |
Chris Lattner | 1543e9c | 2008-03-09 02:18:51 +0000 | [diff] [blame] | 621 | unsigned TokenLexer::isNextTokenLParen() const { |
Reid Spencer | 5f016e2 | 2007-07-11 17:01:13 +0000 | [diff] [blame] | 622 | // Out of tokens? |
| 623 | if (isAtEnd()) |
| 624 | return 2; |
Chris Lattner | 8d89643 | 2008-03-09 02:07:49 +0000 | [diff] [blame] | 625 | return Tokens[CurToken].is(tok::l_paren); |
Reid Spencer | 5f016e2 | 2007-07-11 17:01:13 +0000 | [diff] [blame] | 626 | } |
Chris Lattner | 3f1cc83 | 2008-02-07 06:03:59 +0000 | [diff] [blame] | 627 | |
Peter Collingbourne | b2eb53d | 2011-02-22 13:49:00 +0000 | [diff] [blame] | 628 | /// isParsingPreprocessorDirective - Return true if we are in the middle of a |
| 629 | /// preprocessor directive. |
| 630 | bool TokenLexer::isParsingPreprocessorDirective() const { |
Peter Collingbourne | 8402155 | 2011-02-28 02:37:51 +0000 | [diff] [blame] | 631 | return Tokens[NumTokens-1].is(tok::eod) && !isAtEnd(); |
Peter Collingbourne | b2eb53d | 2011-02-22 13:49:00 +0000 | [diff] [blame] | 632 | } |
Chris Lattner | 3f1cc83 | 2008-02-07 06:03:59 +0000 | [diff] [blame] | 633 | |
| 634 | /// HandleMicrosoftCommentPaste - In microsoft compatibility mode, /##/ pastes |
| 635 | /// together to form a comment that comments out everything in the current |
| 636 | /// macro, other active macros, and anything left on the current physical |
| 637 | /// source line of the instantiated buffer. Handle this by returning the |
| 638 | /// first token on the next line. |
Chris Lattner | 1543e9c | 2008-03-09 02:18:51 +0000 | [diff] [blame] | 639 | void TokenLexer::HandleMicrosoftCommentPaste(Token &Tok) { |
Chris Lattner | 3f1cc83 | 2008-02-07 06:03:59 +0000 | [diff] [blame] | 640 | // We 'comment out' the rest of this macro by just ignoring the rest of the |
| 641 | // tokens that have not been lexed yet, if any. |
Mike Stump | 1eb4433 | 2009-09-09 15:08:12 +0000 | [diff] [blame] | 642 | |
Chris Lattner | 3f1cc83 | 2008-02-07 06:03:59 +0000 | [diff] [blame] | 643 | // Since this must be a macro, mark the macro enabled now that it is no longer |
| 644 | // being expanded. |
| 645 | assert(Macro && "Token streams can't paste comments"); |
| 646 | Macro->EnableMacro(); |
Mike Stump | 1eb4433 | 2009-09-09 15:08:12 +0000 | [diff] [blame] | 647 | |
Chris Lattner | 3f1cc83 | 2008-02-07 06:03:59 +0000 | [diff] [blame] | 648 | PP.HandleMicrosoftCommentPaste(Tok); |
| 649 | } |
Argyrios Kyrtzidis | b73377e | 2011-07-07 03:40:34 +0000 | [diff] [blame^] | 650 | |
| 651 | /// \brief If \arg loc is a FileID and points inside the current macro |
| 652 | /// definition, returns the appropriate source location pointing at the |
| 653 | /// macro expansion source location entry. |
| 654 | SourceLocation TokenLexer::getMacroExpansionLocation(SourceLocation loc) const { |
| 655 | assert(InstantiateLocStart.isValid() && MacroExpansionStart.isValid() && |
| 656 | "Not appropriate for token streams"); |
| 657 | assert(loc.isValid()); |
| 658 | |
| 659 | SourceManager &SM = PP.getSourceManager(); |
| 660 | unsigned relativeOffset; |
| 661 | if (loc.isFileID() && |
| 662 | SM.isInFileID(loc, |
| 663 | MacroDefStartInfo.first, MacroDefStartInfo.second, |
| 664 | Macro->getDefinitionLength(SM), &relativeOffset)) { |
| 665 | return MacroExpansionStart.getFileLocWithOffset(relativeOffset); |
| 666 | } |
| 667 | |
| 668 | return SourceLocation(); |
| 669 | } |