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(); |
| 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; |
Chris Lattner | b99bdad | 2009-01-26 03:46:22 +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(); |
Reid Spencer | 5f016e2 | 2007-07-11 17:01:13 +0000 | [diff] [blame] | 43 | |
| 44 | // If this is a function-like macro, expand the arguments and change |
Chris Lattner | 8d89643 | 2008-03-09 02:07:49 +0000 | [diff] [blame] | 45 | // Tokens to point to the expanded tokens. |
Reid Spencer | 5f016e2 | 2007-07-11 17:01:13 +0000 | [diff] [blame] | 46 | if (Macro->isFunctionLike() && Macro->getNumArgs()) |
| 47 | ExpandFunctionArguments(); |
| 48 | |
| 49 | // Mark the macro as currently disabled, so that it is not recursively |
| 50 | // expanded. The macro must be disabled only after argument pre-expansion of |
| 51 | // function-like macro arguments occurs. |
| 52 | Macro->DisableMacro(); |
| 53 | } |
| 54 | |
Chris Lattner | 9594acf | 2007-07-15 00:25:26 +0000 | [diff] [blame] | 55 | |
| 56 | |
Chris Lattner | 1543e9c | 2008-03-09 02:18:51 +0000 | [diff] [blame] | 57 | /// Create a TokenLexer for the specified token stream. This does not |
Reid Spencer | 5f016e2 | 2007-07-11 17:01:13 +0000 | [diff] [blame] | 58 | /// take ownership of the specified token vector. |
Chris Lattner | 6b88450 | 2008-03-10 06:06:04 +0000 | [diff] [blame] | 59 | void TokenLexer::Init(const Token *TokArray, unsigned NumToks, |
| 60 | bool disableMacroExpansion, bool ownsTokens) { |
Chris Lattner | 1543e9c | 2008-03-09 02:18:51 +0000 | [diff] [blame] | 61 | // 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] | 62 | // associated with it. |
| 63 | destroy(); |
| 64 | |
| 65 | Macro = 0; |
| 66 | ActualArgs = 0; |
Chris Lattner | 8d89643 | 2008-03-09 02:07:49 +0000 | [diff] [blame] | 67 | Tokens = TokArray; |
Chris Lattner | 6b88450 | 2008-03-10 06:06:04 +0000 | [diff] [blame] | 68 | OwnsTokens = ownsTokens; |
| 69 | DisableMacroExpansion = disableMacroExpansion; |
Chris Lattner | 8d89643 | 2008-03-09 02:07:49 +0000 | [diff] [blame] | 70 | NumTokens = NumToks; |
Chris Lattner | 9594acf | 2007-07-15 00:25:26 +0000 | [diff] [blame] | 71 | CurToken = 0; |
Chris Lattner | e7fb484 | 2009-02-15 20:52:18 +0000 | [diff] [blame] | 72 | InstantiateLocStart = InstantiateLocEnd = SourceLocation(); |
Chris Lattner | 9594acf | 2007-07-15 00:25:26 +0000 | [diff] [blame] | 73 | AtStartOfLine = false; |
| 74 | HasLeadingSpace = false; |
Reid Spencer | 5f016e2 | 2007-07-11 17:01:13 +0000 | [diff] [blame] | 75 | |
| 76 | // Set HasLeadingSpace/AtStartOfLine so that the first token will be |
| 77 | // returned unmodified. |
| 78 | if (NumToks != 0) { |
| 79 | AtStartOfLine = TokArray[0].isAtStartOfLine(); |
| 80 | HasLeadingSpace = TokArray[0].hasLeadingSpace(); |
| 81 | } |
| 82 | } |
| 83 | |
| 84 | |
Chris Lattner | 1543e9c | 2008-03-09 02:18:51 +0000 | [diff] [blame] | 85 | void TokenLexer::destroy() { |
Reid Spencer | 5f016e2 | 2007-07-11 17:01:13 +0000 | [diff] [blame] | 86 | // If this was a function-like macro that actually uses its arguments, delete |
| 87 | // the expanded tokens. |
Chris Lattner | 8d89643 | 2008-03-09 02:07:49 +0000 | [diff] [blame] | 88 | if (OwnsTokens) { |
| 89 | delete [] Tokens; |
| 90 | Tokens = 0; |
Chris Lattner | 32fca72 | 2009-03-04 06:50:57 +0000 | [diff] [blame] | 91 | OwnsTokens = false; |
Chris Lattner | 9c68306 | 2007-07-22 01:16:55 +0000 | [diff] [blame] | 92 | } |
Reid Spencer | 5f016e2 | 2007-07-11 17:01:13 +0000 | [diff] [blame] | 93 | |
Chris Lattner | 1543e9c | 2008-03-09 02:18:51 +0000 | [diff] [blame] | 94 | // TokenLexer owns its formal arguments. |
Reid Spencer | 5f016e2 | 2007-07-11 17:01:13 +0000 | [diff] [blame] | 95 | if (ActualArgs) ActualArgs->destroy(); |
| 96 | } |
| 97 | |
| 98 | /// 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] | 99 | /// return preexpanded tokens from Tokens. |
Chris Lattner | 1543e9c | 2008-03-09 02:18:51 +0000 | [diff] [blame] | 100 | void TokenLexer::ExpandFunctionArguments() { |
Chris Lattner | d217773 | 2007-07-20 16:59:19 +0000 | [diff] [blame] | 101 | llvm::SmallVector<Token, 128> ResultToks; |
Reid Spencer | 5f016e2 | 2007-07-11 17:01:13 +0000 | [diff] [blame] | 102 | |
Chris Lattner | 8d89643 | 2008-03-09 02:07:49 +0000 | [diff] [blame] | 103 | // Loop through 'Tokens', expanding them into ResultToks. Keep |
Reid Spencer | 5f016e2 | 2007-07-11 17:01:13 +0000 | [diff] [blame] | 104 | // 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] | 105 | // we install the newly expanded sequence as the new 'Tokens' list. |
Reid Spencer | 5f016e2 | 2007-07-11 17:01:13 +0000 | [diff] [blame] | 106 | bool MadeChange = false; |
| 107 | |
| 108 | // NextTokGetsSpace - When this is true, the next token appended to the |
| 109 | // output list will get a leading space, regardless of whether it had one to |
| 110 | // begin with or not. This is used for placemarker support. |
| 111 | bool NextTokGetsSpace = false; |
| 112 | |
Chris Lattner | 8d89643 | 2008-03-09 02:07:49 +0000 | [diff] [blame] | 113 | for (unsigned i = 0, e = NumTokens; i != e; ++i) { |
Reid Spencer | 5f016e2 | 2007-07-11 17:01:13 +0000 | [diff] [blame] | 114 | // If we found the stringify operator, get the argument stringified. The |
| 115 | // preprocessor already verified that the following token is a macro name |
| 116 | // when the #define was parsed. |
Chris Lattner | 8d89643 | 2008-03-09 02:07:49 +0000 | [diff] [blame] | 117 | const Token &CurTok = Tokens[i]; |
Chris Lattner | 22f6bbc | 2007-10-09 18:02:16 +0000 | [diff] [blame] | 118 | if (CurTok.is(tok::hash) || CurTok.is(tok::hashat)) { |
Chris Lattner | 8d89643 | 2008-03-09 02:07:49 +0000 | [diff] [blame] | 119 | int ArgNo = Macro->getArgumentNum(Tokens[i+1].getIdentifierInfo()); |
Reid Spencer | 5f016e2 | 2007-07-11 17:01:13 +0000 | [diff] [blame] | 120 | assert(ArgNo != -1 && "Token following # is not an argument?"); |
| 121 | |
Chris Lattner | d217773 | 2007-07-20 16:59:19 +0000 | [diff] [blame] | 122 | Token Res; |
Chris Lattner | 22f6bbc | 2007-10-09 18:02:16 +0000 | [diff] [blame] | 123 | if (CurTok.is(tok::hash)) // Stringify |
Reid Spencer | 5f016e2 | 2007-07-11 17:01:13 +0000 | [diff] [blame] | 124 | Res = ActualArgs->getStringifiedArgument(ArgNo, PP); |
| 125 | else { |
| 126 | // 'charify': don't bother caching these. |
Chris Lattner | e5c8ffe | 2008-03-09 02:55:12 +0000 | [diff] [blame] | 127 | Res = MacroArgs::StringifyArgument(ActualArgs->getUnexpArgument(ArgNo), |
| 128 | PP, true); |
Reid Spencer | 5f016e2 | 2007-07-11 17:01:13 +0000 | [diff] [blame] | 129 | } |
| 130 | |
| 131 | // The stringified/charified string leading space flag gets set to match |
| 132 | // the #/#@ operator. |
| 133 | if (CurTok.hasLeadingSpace() || NextTokGetsSpace) |
Chris Lattner | d217773 | 2007-07-20 16:59:19 +0000 | [diff] [blame] | 134 | Res.setFlag(Token::LeadingSpace); |
Reid Spencer | 5f016e2 | 2007-07-11 17:01:13 +0000 | [diff] [blame] | 135 | |
| 136 | ResultToks.push_back(Res); |
| 137 | MadeChange = true; |
| 138 | ++i; // Skip arg name. |
| 139 | NextTokGetsSpace = false; |
| 140 | continue; |
| 141 | } |
| 142 | |
| 143 | // Otherwise, if this is not an argument token, just add the token to the |
| 144 | // output buffer. |
| 145 | IdentifierInfo *II = CurTok.getIdentifierInfo(); |
| 146 | int ArgNo = II ? Macro->getArgumentNum(II) : -1; |
| 147 | if (ArgNo == -1) { |
| 148 | // This isn't an argument, just add it. |
| 149 | ResultToks.push_back(CurTok); |
| 150 | |
| 151 | if (NextTokGetsSpace) { |
Chris Lattner | d217773 | 2007-07-20 16:59:19 +0000 | [diff] [blame] | 152 | ResultToks.back().setFlag(Token::LeadingSpace); |
Reid Spencer | 5f016e2 | 2007-07-11 17:01:13 +0000 | [diff] [blame] | 153 | NextTokGetsSpace = false; |
| 154 | } |
| 155 | continue; |
| 156 | } |
| 157 | |
| 158 | // An argument is expanded somehow, the result is different than the |
| 159 | // input. |
| 160 | MadeChange = true; |
| 161 | |
| 162 | // Otherwise, this is a use of the argument. Find out if there is a paste |
| 163 | // (##) operator before or after the argument. |
| 164 | bool PasteBefore = |
Chris Lattner | 22f6bbc | 2007-10-09 18:02:16 +0000 | [diff] [blame] | 165 | !ResultToks.empty() && ResultToks.back().is(tok::hashhash); |
Chris Lattner | 8d89643 | 2008-03-09 02:07:49 +0000 | [diff] [blame] | 166 | bool PasteAfter = i+1 != e && Tokens[i+1].is(tok::hashhash); |
Reid Spencer | 5f016e2 | 2007-07-11 17:01:13 +0000 | [diff] [blame] | 167 | |
| 168 | // If it is not the LHS/RHS of a ## operator, we must pre-expand the |
| 169 | // argument and substitute the expanded tokens into the result. This is |
| 170 | // C99 6.10.3.1p1. |
| 171 | if (!PasteBefore && !PasteAfter) { |
Chris Lattner | d217773 | 2007-07-20 16:59:19 +0000 | [diff] [blame] | 172 | const Token *ResultArgToks; |
Reid Spencer | 5f016e2 | 2007-07-11 17:01:13 +0000 | [diff] [blame] | 173 | |
| 174 | // Only preexpand the argument if it could possibly need it. This |
| 175 | // avoids some work in common cases. |
Chris Lattner | d217773 | 2007-07-20 16:59:19 +0000 | [diff] [blame] | 176 | const Token *ArgTok = ActualArgs->getUnexpArgument(ArgNo); |
Chris Lattner | cc1a875 | 2007-10-07 08:44:20 +0000 | [diff] [blame] | 177 | if (ActualArgs->ArgNeedsPreexpansion(ArgTok, PP)) |
Reid Spencer | 5f016e2 | 2007-07-11 17:01:13 +0000 | [diff] [blame] | 178 | ResultArgToks = &ActualArgs->getPreExpArgument(ArgNo, PP)[0]; |
| 179 | else |
| 180 | ResultArgToks = ArgTok; // Use non-preexpanded tokens. |
| 181 | |
| 182 | // If the arg token expanded into anything, append it. |
Chris Lattner | 22f6bbc | 2007-10-09 18:02:16 +0000 | [diff] [blame] | 183 | if (ResultArgToks->isNot(tok::eof)) { |
Reid Spencer | 5f016e2 | 2007-07-11 17:01:13 +0000 | [diff] [blame] | 184 | unsigned FirstResult = ResultToks.size(); |
| 185 | unsigned NumToks = MacroArgs::getArgLength(ResultArgToks); |
| 186 | ResultToks.append(ResultArgToks, ResultArgToks+NumToks); |
| 187 | |
| 188 | // If any tokens were substituted from the argument, the whitespace |
| 189 | // before the first token should match the whitespace of the arg |
| 190 | // identifier. |
Chris Lattner | d217773 | 2007-07-20 16:59:19 +0000 | [diff] [blame] | 191 | ResultToks[FirstResult].setFlagValue(Token::LeadingSpace, |
Reid Spencer | 5f016e2 | 2007-07-11 17:01:13 +0000 | [diff] [blame] | 192 | CurTok.hasLeadingSpace() || |
| 193 | NextTokGetsSpace); |
| 194 | NextTokGetsSpace = false; |
| 195 | } else { |
| 196 | // If this is an empty argument, and if there was whitespace before the |
| 197 | // formal token, make sure the next token gets whitespace before it. |
| 198 | NextTokGetsSpace = CurTok.hasLeadingSpace(); |
| 199 | } |
| 200 | continue; |
| 201 | } |
| 202 | |
| 203 | // Okay, we have a token that is either the LHS or RHS of a paste (##) |
| 204 | // argument. It gets substituted as its non-pre-expanded tokens. |
Chris Lattner | d217773 | 2007-07-20 16:59:19 +0000 | [diff] [blame] | 205 | const Token *ArgToks = ActualArgs->getUnexpArgument(ArgNo); |
Reid Spencer | 5f016e2 | 2007-07-11 17:01:13 +0000 | [diff] [blame] | 206 | unsigned NumToks = MacroArgs::getArgLength(ArgToks); |
| 207 | if (NumToks) { // Not an empty argument? |
Chris Lattner | 71a3a8d | 2008-01-29 07:54:23 +0000 | [diff] [blame] | 208 | // If this is the GNU ", ## __VA_ARG__" extension, and we just learned |
| 209 | // that __VA_ARG__ expands to multiple tokens, avoid a pasting error when |
| 210 | // the expander trys to paste ',' with the first token of the __VA_ARG__ |
| 211 | // expansion. |
| 212 | if (PasteBefore && ResultToks.size() >= 2 && |
| 213 | ResultToks[ResultToks.size()-2].is(tok::comma) && |
| 214 | (unsigned)ArgNo == Macro->getNumArgs()-1 && |
| 215 | Macro->isVariadic()) { |
| 216 | // Remove the paste operator, report use of the extension. |
| 217 | PP.Diag(ResultToks.back().getLocation(), diag::ext_paste_comma); |
| 218 | ResultToks.pop_back(); |
| 219 | } |
| 220 | |
Reid Spencer | 5f016e2 | 2007-07-11 17:01:13 +0000 | [diff] [blame] | 221 | ResultToks.append(ArgToks, ArgToks+NumToks); |
| 222 | |
| 223 | // If the next token was supposed to get leading whitespace, ensure it has |
| 224 | // it now. |
Eli Friedman | d9b856f | 2009-05-03 19:16:00 +0000 | [diff] [blame] | 225 | if (CurTok.hasLeadingSpace() || NextTokGetsSpace) { |
Chris Lattner | d217773 | 2007-07-20 16:59:19 +0000 | [diff] [blame] | 226 | ResultToks[ResultToks.size()-NumToks].setFlag(Token::LeadingSpace); |
Reid Spencer | 5f016e2 | 2007-07-11 17:01:13 +0000 | [diff] [blame] | 227 | NextTokGetsSpace = false; |
| 228 | } |
| 229 | continue; |
| 230 | } |
| 231 | |
| 232 | // If an empty argument is on the LHS or RHS of a paste, the standard (C99 |
| 233 | // 6.10.3.3p2,3) calls for a bunch of placemarker stuff to occur. We |
| 234 | // implement this by eating ## operators when a LHS or RHS expands to |
| 235 | // empty. |
| 236 | NextTokGetsSpace |= CurTok.hasLeadingSpace(); |
| 237 | if (PasteAfter) { |
| 238 | // Discard the argument token and skip (don't copy to the expansion |
| 239 | // buffer) the paste operator after it. |
Chris Lattner | 8d89643 | 2008-03-09 02:07:49 +0000 | [diff] [blame] | 240 | NextTokGetsSpace |= Tokens[i+1].hasLeadingSpace(); |
Reid Spencer | 5f016e2 | 2007-07-11 17:01:13 +0000 | [diff] [blame] | 241 | ++i; |
| 242 | continue; |
| 243 | } |
| 244 | |
| 245 | // If this is on the RHS of a paste operator, we've already copied the |
| 246 | // paste operator to the ResultToks list. Remove it. |
Chris Lattner | 22f6bbc | 2007-10-09 18:02:16 +0000 | [diff] [blame] | 247 | assert(PasteBefore && ResultToks.back().is(tok::hashhash)); |
Reid Spencer | 5f016e2 | 2007-07-11 17:01:13 +0000 | [diff] [blame] | 248 | NextTokGetsSpace |= ResultToks.back().hasLeadingSpace(); |
| 249 | ResultToks.pop_back(); |
| 250 | |
| 251 | // If this is the __VA_ARGS__ token, and if the argument wasn't provided, |
| 252 | // and if the macro had at least one real argument, and if the token before |
| 253 | // the ## was a comma, remove the comma. |
| 254 | if ((unsigned)ArgNo == Macro->getNumArgs()-1 && // is __VA_ARGS__ |
| 255 | ActualArgs->isVarargsElidedUse() && // Argument elided. |
Chris Lattner | 22f6bbc | 2007-10-09 18:02:16 +0000 | [diff] [blame] | 256 | !ResultToks.empty() && ResultToks.back().is(tok::comma)) { |
Reid Spencer | 5f016e2 | 2007-07-11 17:01:13 +0000 | [diff] [blame] | 257 | // Never add a space, even if the comma, ##, or arg had a space. |
| 258 | NextTokGetsSpace = false; |
Chris Lattner | 71a3a8d | 2008-01-29 07:54:23 +0000 | [diff] [blame] | 259 | // Remove the paste operator, report use of the extension. |
| 260 | PP.Diag(ResultToks.back().getLocation(), diag::ext_paste_comma); |
Reid Spencer | 5f016e2 | 2007-07-11 17:01:13 +0000 | [diff] [blame] | 261 | ResultToks.pop_back(); |
| 262 | } |
| 263 | continue; |
| 264 | } |
| 265 | |
Chris Lattner | 8d89643 | 2008-03-09 02:07:49 +0000 | [diff] [blame] | 266 | // If anything changed, install this as the new Tokens list. |
Reid Spencer | 5f016e2 | 2007-07-11 17:01:13 +0000 | [diff] [blame] | 267 | if (MadeChange) { |
Chris Lattner | 32fca72 | 2009-03-04 06:50:57 +0000 | [diff] [blame] | 268 | assert(!OwnsTokens && "This would leak if we already own the token list"); |
Reid Spencer | 5f016e2 | 2007-07-11 17:01:13 +0000 | [diff] [blame] | 269 | // This is deleted in the dtor. |
Chris Lattner | 8d89643 | 2008-03-09 02:07:49 +0000 | [diff] [blame] | 270 | NumTokens = ResultToks.size(); |
Chris Lattner | 32fca72 | 2009-03-04 06:50:57 +0000 | [diff] [blame] | 271 | llvm::BumpPtrAllocator &Alloc = PP.getPreprocessorAllocator(); |
| 272 | Token *Res = |
| 273 | static_cast<Token *>(Alloc.Allocate(sizeof(Token)*ResultToks.size(), |
| 274 | llvm::alignof<Token>())); |
Chris Lattner | 8d89643 | 2008-03-09 02:07:49 +0000 | [diff] [blame] | 275 | if (NumTokens) |
| 276 | memcpy(Res, &ResultToks[0], NumTokens*sizeof(Token)); |
| 277 | Tokens = Res; |
Chris Lattner | 32fca72 | 2009-03-04 06:50:57 +0000 | [diff] [blame] | 278 | |
| 279 | // The preprocessor bump pointer owns these tokens, not us. |
| 280 | OwnsTokens = false; |
Reid Spencer | 5f016e2 | 2007-07-11 17:01:13 +0000 | [diff] [blame] | 281 | } |
| 282 | } |
| 283 | |
| 284 | /// Lex - Lex and return a token from this macro stream. |
| 285 | /// |
Chris Lattner | 1543e9c | 2008-03-09 02:18:51 +0000 | [diff] [blame] | 286 | void TokenLexer::Lex(Token &Tok) { |
Reid Spencer | 5f016e2 | 2007-07-11 17:01:13 +0000 | [diff] [blame] | 287 | // Lexing off the end of the macro, pop this macro off the expansion stack. |
| 288 | if (isAtEnd()) { |
| 289 | // If this is a macro (not a token stream), mark the macro enabled now |
| 290 | // that it is no longer being expanded. |
| 291 | if (Macro) Macro->EnableMacro(); |
| 292 | |
| 293 | // Pop this context off the preprocessors lexer stack and get the next |
| 294 | // token. This will delete "this" so remember the PP instance var. |
| 295 | Preprocessor &PPCache = PP; |
Chris Lattner | fde2bf9 | 2008-03-09 03:04:16 +0000 | [diff] [blame] | 296 | if (PP.HandleEndOfTokenLexer(Tok)) |
Reid Spencer | 5f016e2 | 2007-07-11 17:01:13 +0000 | [diff] [blame] | 297 | return; |
| 298 | |
Chris Lattner | fde2bf9 | 2008-03-09 03:04:16 +0000 | [diff] [blame] | 299 | // HandleEndOfTokenLexer may not return a token. If it doesn't, lex |
| 300 | // whatever is next. |
Reid Spencer | 5f016e2 | 2007-07-11 17:01:13 +0000 | [diff] [blame] | 301 | return PPCache.Lex(Tok); |
| 302 | } |
| 303 | |
| 304 | // If this is the first token of the expanded result, we inherit spacing |
| 305 | // properties later. |
| 306 | bool isFirstToken = CurToken == 0; |
| 307 | |
| 308 | // Get the next token to return. |
Chris Lattner | 8d89643 | 2008-03-09 02:07:49 +0000 | [diff] [blame] | 309 | Tok = Tokens[CurToken++]; |
Reid Spencer | 5f016e2 | 2007-07-11 17:01:13 +0000 | [diff] [blame] | 310 | |
Chris Lattner | 332ee08 | 2009-04-19 20:29:42 +0000 | [diff] [blame] | 311 | bool TokenIsFromPaste = false; |
| 312 | |
Reid Spencer | 5f016e2 | 2007-07-11 17:01:13 +0000 | [diff] [blame] | 313 | // If this token is followed by a token paste (##) operator, paste the tokens! |
Eli Friedman | 617bd45 | 2009-04-25 22:20:56 +0000 | [diff] [blame] | 314 | if (!isAtEnd() && Tokens[CurToken].is(tok::hashhash)) { |
Chris Lattner | 3f1cc83 | 2008-02-07 06:03:59 +0000 | [diff] [blame] | 315 | if (PasteTokens(Tok)) { |
| 316 | // When handling the microsoft /##/ extension, the final token is |
| 317 | // returned by PasteTokens, not the pasted token. |
| 318 | return; |
Chris Lattner | 332ee08 | 2009-04-19 20:29:42 +0000 | [diff] [blame] | 319 | } else { |
| 320 | TokenIsFromPaste = true; |
Chris Lattner | 3f1cc83 | 2008-02-07 06:03:59 +0000 | [diff] [blame] | 321 | } |
Eli Friedman | 617bd45 | 2009-04-25 22:20:56 +0000 | [diff] [blame] | 322 | } |
Reid Spencer | 5f016e2 | 2007-07-11 17:01:13 +0000 | [diff] [blame] | 323 | |
| 324 | // The token's current location indicate where the token was lexed from. We |
| 325 | // need this information to compute the spelling of the token, but any |
| 326 | // diagnostics for the expanded token should appear as if they came from |
| 327 | // InstantiationLoc. Pull this information together into a new SourceLocation |
| 328 | // that captures all of this. |
Chris Lattner | e7fb484 | 2009-02-15 20:52:18 +0000 | [diff] [blame] | 329 | if (InstantiateLocStart.isValid()) { // Don't do this for token streams. |
| 330 | SourceManager &SM = PP.getSourceManager(); |
| 331 | Tok.setLocation(SM.createInstantiationLoc(Tok.getLocation(), |
| 332 | InstantiateLocStart, |
| 333 | InstantiateLocEnd, |
| 334 | Tok.getLength())); |
Reid Spencer | 5f016e2 | 2007-07-11 17:01:13 +0000 | [diff] [blame] | 335 | } |
| 336 | |
| 337 | // If this is the first token, set the lexical properties of the token to |
| 338 | // match the lexical properties of the macro identifier. |
| 339 | if (isFirstToken) { |
Chris Lattner | d217773 | 2007-07-20 16:59:19 +0000 | [diff] [blame] | 340 | Tok.setFlagValue(Token::StartOfLine , AtStartOfLine); |
| 341 | Tok.setFlagValue(Token::LeadingSpace, HasLeadingSpace); |
Reid Spencer | 5f016e2 | 2007-07-11 17:01:13 +0000 | [diff] [blame] | 342 | } |
| 343 | |
| 344 | // Handle recursive expansion! |
Argyrios Kyrtzidis | 6f421f9 | 2009-05-22 21:09:31 +0000 | [diff] [blame^] | 345 | if (!Tok.isAnnotation() && Tok.getIdentifierInfo() != 0) { |
Chris Lattner | 863c486 | 2009-01-23 18:35:48 +0000 | [diff] [blame] | 346 | // Change the kind of this identifier to the appropriate token kind, e.g. |
| 347 | // turning "for" into a keyword. |
Argyrios Kyrtzidis | 6f421f9 | 2009-05-22 21:09:31 +0000 | [diff] [blame^] | 348 | IdentifierInfo *II = Tok.getIdentifierInfo(); |
Chris Lattner | 863c486 | 2009-01-23 18:35:48 +0000 | [diff] [blame] | 349 | Tok.setKind(II->getTokenID()); |
| 350 | |
Chris Lattner | 332ee08 | 2009-04-19 20:29:42 +0000 | [diff] [blame] | 351 | // If this identifier was poisoned and from a paste, emit an error. This |
| 352 | // won't be handled by Preprocessor::HandleIdentifier because this is coming |
| 353 | // from a macro expansion. |
| 354 | if (II->isPoisoned() && TokenIsFromPaste) { |
| 355 | // We warn about __VA_ARGS__ with poisoning. |
| 356 | if (II->isStr("__VA_ARGS__")) |
| 357 | PP.Diag(Tok, diag::ext_pp_bad_vaargs_use); |
| 358 | else |
| 359 | PP.Diag(Tok, diag::err_pp_used_poisoned_id); |
| 360 | } |
| 361 | |
Chris Lattner | 863c486 | 2009-01-23 18:35:48 +0000 | [diff] [blame] | 362 | if (!DisableMacroExpansion && II->isHandleIdentifierCase()) |
| 363 | PP.HandleIdentifier(Tok); |
| 364 | } |
Reid Spencer | 5f016e2 | 2007-07-11 17:01:13 +0000 | [diff] [blame] | 365 | |
| 366 | // Otherwise, return a normal token. |
| 367 | } |
| 368 | |
| 369 | /// PasteTokens - Tok is the LHS of a ## operator, and CurToken is the ## |
| 370 | /// 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] | 371 | /// are more ## after it, chomp them iteratively. Return the result as Tok. |
Chris Lattner | 3f1cc83 | 2008-02-07 06:03:59 +0000 | [diff] [blame] | 372 | /// If this returns true, the caller should immediately return the token. |
Chris Lattner | 1543e9c | 2008-03-09 02:18:51 +0000 | [diff] [blame] | 373 | bool TokenLexer::PasteTokens(Token &Tok) { |
Reid Spencer | 5f016e2 | 2007-07-11 17:01:13 +0000 | [diff] [blame] | 374 | llvm::SmallVector<char, 128> Buffer; |
Chris Lattner | 47246be | 2009-01-26 19:29:26 +0000 | [diff] [blame] | 375 | const char *ResultTokStrPtr = 0; |
Reid Spencer | 5f016e2 | 2007-07-11 17:01:13 +0000 | [diff] [blame] | 376 | do { |
| 377 | // Consume the ## operator. |
Chris Lattner | 8d89643 | 2008-03-09 02:07:49 +0000 | [diff] [blame] | 378 | SourceLocation PasteOpLoc = Tokens[CurToken].getLocation(); |
Reid Spencer | 5f016e2 | 2007-07-11 17:01:13 +0000 | [diff] [blame] | 379 | ++CurToken; |
| 380 | assert(!isAtEnd() && "No token on the RHS of a paste operator!"); |
| 381 | |
| 382 | // Get the RHS token. |
Chris Lattner | 8d89643 | 2008-03-09 02:07:49 +0000 | [diff] [blame] | 383 | const Token &RHS = Tokens[CurToken]; |
Reid Spencer | 5f016e2 | 2007-07-11 17:01:13 +0000 | [diff] [blame] | 384 | |
Reid Spencer | 5f016e2 | 2007-07-11 17:01:13 +0000 | [diff] [blame] | 385 | // 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] | 386 | // the two tokens. |
| 387 | Buffer.resize(Tok.getLength() + RHS.getLength()); |
Reid Spencer | 5f016e2 | 2007-07-11 17:01:13 +0000 | [diff] [blame] | 388 | |
| 389 | // Get the spelling of the LHS token in Buffer. |
| 390 | const char *BufPtr = &Buffer[0]; |
| 391 | unsigned LHSLen = PP.getSpelling(Tok, BufPtr); |
| 392 | if (BufPtr != &Buffer[0]) // Really, we want the chars in Buffer! |
| 393 | memcpy(&Buffer[0], BufPtr, LHSLen); |
| 394 | |
| 395 | BufPtr = &Buffer[LHSLen]; |
| 396 | unsigned RHSLen = PP.getSpelling(RHS, BufPtr); |
| 397 | if (BufPtr != &Buffer[LHSLen]) // Really, we want the chars in Buffer! |
| 398 | memcpy(&Buffer[LHSLen], BufPtr, RHSLen); |
| 399 | |
Reid Spencer | 5f016e2 | 2007-07-11 17:01:13 +0000 | [diff] [blame] | 400 | // Trim excess space. |
Chris Lattner | 1fa4953 | 2009-03-08 08:08:45 +0000 | [diff] [blame] | 401 | Buffer.resize(LHSLen+RHSLen); |
Reid Spencer | 5f016e2 | 2007-07-11 17:01:13 +0000 | [diff] [blame] | 402 | |
| 403 | // Plop the pasted result (including the trailing newline and null) into a |
| 404 | // scratch buffer where we can lex it. |
Chris Lattner | 47246be | 2009-01-26 19:29:26 +0000 | [diff] [blame] | 405 | Token ResultTokTmp; |
| 406 | ResultTokTmp.startToken(); |
Reid Spencer | 5f016e2 | 2007-07-11 17:01:13 +0000 | [diff] [blame] | 407 | |
Chris Lattner | 47246be | 2009-01-26 19:29:26 +0000 | [diff] [blame] | 408 | // Claim that the tmp token is a string_literal so that we can get the |
| 409 | // character pointer back from CreateString. |
| 410 | ResultTokTmp.setKind(tok::string_literal); |
| 411 | PP.CreateString(&Buffer[0], Buffer.size(), ResultTokTmp); |
| 412 | SourceLocation ResultTokLoc = ResultTokTmp.getLocation(); |
| 413 | ResultTokStrPtr = ResultTokTmp.getLiteralData(); |
| 414 | |
Reid Spencer | 5f016e2 | 2007-07-11 17:01:13 +0000 | [diff] [blame] | 415 | // Lex the resultant pasted token into Result. |
Chris Lattner | d217773 | 2007-07-20 16:59:19 +0000 | [diff] [blame] | 416 | Token Result; |
Reid Spencer | 5f016e2 | 2007-07-11 17:01:13 +0000 | [diff] [blame] | 417 | |
Chris Lattner | 0af5742 | 2008-10-12 01:31:51 +0000 | [diff] [blame] | 418 | if (Tok.is(tok::identifier) && RHS.is(tok::identifier)) { |
Reid Spencer | 5f016e2 | 2007-07-11 17:01:13 +0000 | [diff] [blame] | 419 | // Common paste case: identifier+identifier = identifier. Avoid creating |
| 420 | // a lexer and other overhead. |
| 421 | PP.IncrementPasteCounter(true); |
| 422 | Result.startToken(); |
| 423 | Result.setKind(tok::identifier); |
| 424 | Result.setLocation(ResultTokLoc); |
| 425 | Result.setLength(LHSLen+RHSLen); |
| 426 | } else { |
| 427 | PP.IncrementPasteCounter(false); |
| 428 | |
Chris Lattner | bcc2a67 | 2009-01-19 06:46:35 +0000 | [diff] [blame] | 429 | assert(ResultTokLoc.isFileID() && |
| 430 | "Should be a raw location into scratch buffer"); |
Reid Spencer | 5f016e2 | 2007-07-11 17:01:13 +0000 | [diff] [blame] | 431 | SourceManager &SourceMgr = PP.getSourceManager(); |
Chris Lattner | 47246be | 2009-01-26 19:29:26 +0000 | [diff] [blame] | 432 | FileID LocFileID = SourceMgr.getFileID(ResultTokLoc); |
Chris Lattner | bcc2a67 | 2009-01-19 06:46:35 +0000 | [diff] [blame] | 433 | |
Chris Lattner | 47246be | 2009-01-26 19:29:26 +0000 | [diff] [blame] | 434 | const char *ScratchBufStart = SourceMgr.getBufferData(LocFileID).first; |
Chris Lattner | bcc2a67 | 2009-01-19 06:46:35 +0000 | [diff] [blame] | 435 | |
| 436 | // 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] | 437 | // 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] | 438 | Lexer TL(SourceMgr.getLocForStartOfFile(LocFileID), |
| 439 | PP.getLangOptions(), ScratchBufStart, |
Chris Lattner | 1fa4953 | 2009-03-08 08:08:45 +0000 | [diff] [blame] | 440 | ResultTokStrPtr, ResultTokStrPtr+LHSLen+RHSLen); |
Reid Spencer | 5f016e2 | 2007-07-11 17:01:13 +0000 | [diff] [blame] | 441 | |
| 442 | // Lex a token in raw mode. This way it won't look up identifiers |
| 443 | // automatically, lexing off the end will return an eof token, and |
| 444 | // warnings are disabled. This returns true if the result token is the |
| 445 | // entire buffer. |
Chris Lattner | 1fa4953 | 2009-03-08 08:08:45 +0000 | [diff] [blame] | 446 | bool isInvalid = !TL.LexFromRawLexer(Result); |
Reid Spencer | 5f016e2 | 2007-07-11 17:01:13 +0000 | [diff] [blame] | 447 | |
| 448 | // If we got an EOF token, we didn't form even ONE token. For example, we |
| 449 | // did "/ ## /" to get "//". |
Chris Lattner | 1fa4953 | 2009-03-08 08:08:45 +0000 | [diff] [blame] | 450 | isInvalid |= Result.is(tok::eof); |
Reid Spencer | 5f016e2 | 2007-07-11 17:01:13 +0000 | [diff] [blame] | 451 | |
Chris Lattner | 1fa4953 | 2009-03-08 08:08:45 +0000 | [diff] [blame] | 452 | // If pasting the two tokens didn't form a full new token, this is an |
| 453 | // error. This occurs with "x ## +" and other stuff. Return with Tok |
| 454 | // unmodified and with RHS as the next token to lex. |
| 455 | if (isInvalid) { |
| 456 | // Test for the Microsoft extension of /##/ turning into // here on the |
| 457 | // error path. |
| 458 | if (PP.getLangOptions().Microsoft && Tok.is(tok::slash) && |
| 459 | RHS.is(tok::slash)) { |
| 460 | HandleMicrosoftCommentPaste(Tok); |
| 461 | return true; |
| 462 | } |
Chris Lattner | 47246be | 2009-01-26 19:29:26 +0000 | [diff] [blame] | 463 | |
Chris Lattner | 48be380 | 2009-03-18 20:31:57 +0000 | [diff] [blame] | 464 | // Do not emit the warning when preprocessing assembler code. |
| 465 | if (!PP.getLangOptions().AsmPreprocessor) |
| 466 | PP.Diag(PasteOpLoc, diag::err_pp_bad_paste) |
| 467 | << std::string(Buffer.begin(), Buffer.end()-1); |
Chris Lattner | f1538de | 2009-03-23 17:32:45 +0000 | [diff] [blame] | 468 | |
| 469 | // Do not consume the RHS. |
| 470 | --CurToken; |
Chris Lattner | 1fa4953 | 2009-03-08 08:08:45 +0000 | [diff] [blame] | 471 | } |
| 472 | |
| 473 | // Turn ## into 'unknown' to avoid # ## # from looking like a paste |
| 474 | // operator. |
| 475 | if (Result.is(tok::hashhash)) |
| 476 | Result.setKind(tok::unknown); |
Reid Spencer | 5f016e2 | 2007-07-11 17:01:13 +0000 | [diff] [blame] | 477 | } |
Chris Lattner | 1fa4953 | 2009-03-08 08:08:45 +0000 | [diff] [blame] | 478 | |
Reid Spencer | 5f016e2 | 2007-07-11 17:01:13 +0000 | [diff] [blame] | 479 | // Transfer properties of the LHS over the the Result. |
Chris Lattner | d217773 | 2007-07-20 16:59:19 +0000 | [diff] [blame] | 480 | Result.setFlagValue(Token::StartOfLine , Tok.isAtStartOfLine()); |
| 481 | Result.setFlagValue(Token::LeadingSpace, Tok.hasLeadingSpace()); |
Reid Spencer | 5f016e2 | 2007-07-11 17:01:13 +0000 | [diff] [blame] | 482 | |
| 483 | // Finally, replace LHS with the result, consume the RHS, and iterate. |
| 484 | ++CurToken; |
| 485 | Tok = Result; |
Chris Lattner | 8d89643 | 2008-03-09 02:07:49 +0000 | [diff] [blame] | 486 | } while (!isAtEnd() && Tokens[CurToken].is(tok::hashhash)); |
Reid Spencer | 5f016e2 | 2007-07-11 17:01:13 +0000 | [diff] [blame] | 487 | |
| 488 | // Now that we got the result token, it will be subject to expansion. Since |
| 489 | // token pasting re-lexes the result token in raw mode, identifier information |
| 490 | // isn't looked up. As such, if the result is an identifier, look up id info. |
Chris Lattner | 22f6bbc | 2007-10-09 18:02:16 +0000 | [diff] [blame] | 491 | if (Tok.is(tok::identifier)) { |
Reid Spencer | 5f016e2 | 2007-07-11 17:01:13 +0000 | [diff] [blame] | 492 | // Look up the identifier info for the token. We disabled identifier lookup |
| 493 | // by saying we're skipping contents, so we need to do this manually. |
Chris Lattner | 88e43a5 | 2009-04-19 20:06:32 +0000 | [diff] [blame] | 494 | IdentifierInfo *II = PP.LookUpIdentifierInfo(Tok, ResultTokStrPtr); |
| 495 | Tok.setIdentifierInfo(II); |
Reid Spencer | 5f016e2 | 2007-07-11 17:01:13 +0000 | [diff] [blame] | 496 | } |
Chris Lattner | 3f1cc83 | 2008-02-07 06:03:59 +0000 | [diff] [blame] | 497 | return false; |
Reid Spencer | 5f016e2 | 2007-07-11 17:01:13 +0000 | [diff] [blame] | 498 | } |
| 499 | |
| 500 | /// isNextTokenLParen - If the next token lexed will pop this macro off the |
| 501 | /// expansion stack, return 2. If the next unexpanded token is a '(', return |
| 502 | /// 1, otherwise return 0. |
Chris Lattner | 1543e9c | 2008-03-09 02:18:51 +0000 | [diff] [blame] | 503 | unsigned TokenLexer::isNextTokenLParen() const { |
Reid Spencer | 5f016e2 | 2007-07-11 17:01:13 +0000 | [diff] [blame] | 504 | // Out of tokens? |
| 505 | if (isAtEnd()) |
| 506 | return 2; |
Chris Lattner | 8d89643 | 2008-03-09 02:07:49 +0000 | [diff] [blame] | 507 | return Tokens[CurToken].is(tok::l_paren); |
Reid Spencer | 5f016e2 | 2007-07-11 17:01:13 +0000 | [diff] [blame] | 508 | } |
Chris Lattner | 3f1cc83 | 2008-02-07 06:03:59 +0000 | [diff] [blame] | 509 | |
| 510 | |
| 511 | /// HandleMicrosoftCommentPaste - In microsoft compatibility mode, /##/ pastes |
| 512 | /// together to form a comment that comments out everything in the current |
| 513 | /// macro, other active macros, and anything left on the current physical |
| 514 | /// source line of the instantiated buffer. Handle this by returning the |
| 515 | /// first token on the next line. |
Chris Lattner | 1543e9c | 2008-03-09 02:18:51 +0000 | [diff] [blame] | 516 | void TokenLexer::HandleMicrosoftCommentPaste(Token &Tok) { |
Chris Lattner | 3f1cc83 | 2008-02-07 06:03:59 +0000 | [diff] [blame] | 517 | // We 'comment out' the rest of this macro by just ignoring the rest of the |
| 518 | // tokens that have not been lexed yet, if any. |
| 519 | |
| 520 | // Since this must be a macro, mark the macro enabled now that it is no longer |
| 521 | // being expanded. |
| 522 | assert(Macro && "Token streams can't paste comments"); |
| 523 | Macro->EnableMacro(); |
| 524 | |
| 525 | PP.HandleMicrosoftCommentPaste(Tok); |
| 526 | } |