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" |
Reid Spencer | 5f016e2 | 2007-07-11 17:01:13 +0000 | [diff] [blame] | 15 | #include "clang/Basic/SourceManager.h" |
Chris Lattner | 500d329 | 2009-01-29 05:15:15 +0000 | [diff] [blame] | 16 | #include "clang/Lex/LexDiagnostic.h" |
Stephen Hines | 651f13c | 2014-04-23 16:59:28 -0700 | [diff] [blame] | 17 | #include "clang/Lex/MacroArgs.h" |
Chandler Carruth | 55fc873 | 2012-12-04 09:13:33 +0000 | [diff] [blame] | 18 | #include "clang/Lex/MacroInfo.h" |
| 19 | #include "clang/Lex/Preprocessor.h" |
Benjamin Kramer | 8fe83e1 | 2012-02-04 13:45:25 +0000 | [diff] [blame] | 20 | #include "llvm/ADT/SmallString.h" |
Reid Spencer | 5f016e2 | 2007-07-11 17:01:13 +0000 | [diff] [blame] | 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. |
Richard Smith | c30981a | 2012-08-30 13:38:46 +0000 | [diff] [blame] | 26 | void TokenLexer::Init(Token &Tok, SourceLocation ELEnd, MacroInfo *MI, |
| 27 | MacroArgs *Actuals) { |
Chris Lattner | 1543e9c | 2008-03-09 02:18:51 +0000 | [diff] [blame] | 28 | // 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] | 29 | // associated with it. |
| 30 | destroy(); |
Mike Stump | 1eb4433 | 2009-09-09 15:08:12 +0000 | [diff] [blame] | 31 | |
Richard Smith | c30981a | 2012-08-30 13:38:46 +0000 | [diff] [blame] | 32 | Macro = MI; |
Chris Lattner | 9594acf | 2007-07-15 00:25:26 +0000 | [diff] [blame] | 33 | ActualArgs = Actuals; |
| 34 | CurToken = 0; |
Mike Stump | 1eb4433 | 2009-09-09 15:08:12 +0000 | [diff] [blame] | 35 | |
Chandler Carruth | d0f5e44 | 2011-07-14 08:20:34 +0000 | [diff] [blame] | 36 | ExpandLocStart = Tok.getLocation(); |
| 37 | ExpandLocEnd = ELEnd; |
Chris Lattner | 9594acf | 2007-07-15 00:25:26 +0000 | [diff] [blame] | 38 | AtStartOfLine = Tok.isAtStartOfLine(); |
| 39 | HasLeadingSpace = Tok.hasLeadingSpace(); |
Stephen Hines | 651f13c | 2014-04-23 16:59:28 -0700 | [diff] [blame] | 40 | NextTokGetsSpace = false; |
Chris Lattner | 8d89643 | 2008-03-09 02:07:49 +0000 | [diff] [blame] | 41 | Tokens = &*Macro->tokens_begin(); |
| 42 | OwnsTokens = false; |
Chris Lattner | 6b88450 | 2008-03-10 06:06:04 +0000 | [diff] [blame] | 43 | DisableMacroExpansion = false; |
Chris Lattner | 8d89643 | 2008-03-09 02:07:49 +0000 | [diff] [blame] | 44 | NumTokens = Macro->tokens_end()-Macro->tokens_begin(); |
Argyrios Kyrtzidis | b73377e | 2011-07-07 03:40:34 +0000 | [diff] [blame] | 45 | MacroExpansionStart = SourceLocation(); |
| 46 | |
| 47 | SourceManager &SM = PP.getSourceManager(); |
Douglas Gregor | f62d43d | 2011-07-19 16:10:42 +0000 | [diff] [blame] | 48 | MacroStartSLocOffset = SM.getNextLocalOffset(); |
Argyrios Kyrtzidis | b73377e | 2011-07-07 03:40:34 +0000 | [diff] [blame] | 49 | |
| 50 | if (NumTokens > 0) { |
| 51 | assert(Tokens[0].getLocation().isValid()); |
| 52 | assert((Tokens[0].getLocation().isFileID() || Tokens[0].is(tok::comment)) && |
| 53 | "Macro defined in macro?"); |
Chandler Carruth | d0f5e44 | 2011-07-14 08:20:34 +0000 | [diff] [blame] | 54 | assert(ExpandLocStart.isValid()); |
Argyrios Kyrtzidis | b73377e | 2011-07-07 03:40:34 +0000 | [diff] [blame] | 55 | |
| 56 | // Reserve a source location entry chunk for the length of the macro |
| 57 | // definition. Tokens that get lexed directly from the definition will |
| 58 | // have their locations pointing inside this chunk. This is to avoid |
| 59 | // creating separate source location entries for each token. |
Argyrios Kyrtzidis | 499ea55 | 2011-08-23 21:02:38 +0000 | [diff] [blame] | 60 | MacroDefStart = SM.getExpansionLoc(Tokens[0].getLocation()); |
| 61 | MacroDefLength = Macro->getDefinitionLength(SM); |
| 62 | MacroExpansionStart = SM.createExpansionLoc(MacroDefStart, |
Chandler Carruth | bf340e4 | 2011-07-26 03:03:05 +0000 | [diff] [blame] | 63 | ExpandLocStart, |
| 64 | ExpandLocEnd, |
Argyrios Kyrtzidis | 499ea55 | 2011-08-23 21:02:38 +0000 | [diff] [blame] | 65 | MacroDefLength); |
Argyrios Kyrtzidis | b73377e | 2011-07-07 03:40:34 +0000 | [diff] [blame] | 66 | } |
Reid Spencer | 5f016e2 | 2007-07-11 17:01:13 +0000 | [diff] [blame] | 67 | |
| 68 | // If this is a function-like macro, expand the arguments and change |
Chris Lattner | 8d89643 | 2008-03-09 02:07:49 +0000 | [diff] [blame] | 69 | // Tokens to point to the expanded tokens. |
Reid Spencer | 5f016e2 | 2007-07-11 17:01:13 +0000 | [diff] [blame] | 70 | if (Macro->isFunctionLike() && Macro->getNumArgs()) |
| 71 | ExpandFunctionArguments(); |
Mike Stump | 1eb4433 | 2009-09-09 15:08:12 +0000 | [diff] [blame] | 72 | |
Reid Spencer | 5f016e2 | 2007-07-11 17:01:13 +0000 | [diff] [blame] | 73 | // Mark the macro as currently disabled, so that it is not recursively |
| 74 | // expanded. The macro must be disabled only after argument pre-expansion of |
| 75 | // function-like macro arguments occurs. |
| 76 | Macro->DisableMacro(); |
| 77 | } |
| 78 | |
Chris Lattner | 9594acf | 2007-07-15 00:25:26 +0000 | [diff] [blame] | 79 | |
| 80 | |
Chris Lattner | 1543e9c | 2008-03-09 02:18:51 +0000 | [diff] [blame] | 81 | /// Create a TokenLexer for the specified token stream. This does not |
Reid Spencer | 5f016e2 | 2007-07-11 17:01:13 +0000 | [diff] [blame] | 82 | /// take ownership of the specified token vector. |
Chris Lattner | 6b88450 | 2008-03-10 06:06:04 +0000 | [diff] [blame] | 83 | void TokenLexer::Init(const Token *TokArray, unsigned NumToks, |
| 84 | bool disableMacroExpansion, bool ownsTokens) { |
Chris Lattner | 1543e9c | 2008-03-09 02:18:51 +0000 | [diff] [blame] | 85 | // 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] | 86 | // associated with it. |
| 87 | destroy(); |
Mike Stump | 1eb4433 | 2009-09-09 15:08:12 +0000 | [diff] [blame] | 88 | |
Stephen Hines | 6bcf27b | 2014-05-29 04:14:42 -0700 | [diff] [blame] | 89 | Macro = nullptr; |
| 90 | ActualArgs = nullptr; |
Chris Lattner | 8d89643 | 2008-03-09 02:07:49 +0000 | [diff] [blame] | 91 | Tokens = TokArray; |
Chris Lattner | 6b88450 | 2008-03-10 06:06:04 +0000 | [diff] [blame] | 92 | OwnsTokens = ownsTokens; |
| 93 | DisableMacroExpansion = disableMacroExpansion; |
Chris Lattner | 8d89643 | 2008-03-09 02:07:49 +0000 | [diff] [blame] | 94 | NumTokens = NumToks; |
Chris Lattner | 9594acf | 2007-07-15 00:25:26 +0000 | [diff] [blame] | 95 | CurToken = 0; |
Chandler Carruth | d0f5e44 | 2011-07-14 08:20:34 +0000 | [diff] [blame] | 96 | ExpandLocStart = ExpandLocEnd = SourceLocation(); |
Chris Lattner | 9594acf | 2007-07-15 00:25:26 +0000 | [diff] [blame] | 97 | AtStartOfLine = false; |
| 98 | HasLeadingSpace = false; |
Stephen Hines | 651f13c | 2014-04-23 16:59:28 -0700 | [diff] [blame] | 99 | NextTokGetsSpace = false; |
Argyrios Kyrtzidis | b73377e | 2011-07-07 03:40:34 +0000 | [diff] [blame] | 100 | MacroExpansionStart = SourceLocation(); |
Mike Stump | 1eb4433 | 2009-09-09 15:08:12 +0000 | [diff] [blame] | 101 | |
Reid Spencer | 5f016e2 | 2007-07-11 17:01:13 +0000 | [diff] [blame] | 102 | // Set HasLeadingSpace/AtStartOfLine so that the first token will be |
| 103 | // returned unmodified. |
| 104 | if (NumToks != 0) { |
| 105 | AtStartOfLine = TokArray[0].isAtStartOfLine(); |
| 106 | HasLeadingSpace = TokArray[0].hasLeadingSpace(); |
| 107 | } |
| 108 | } |
| 109 | |
| 110 | |
Chris Lattner | 1543e9c | 2008-03-09 02:18:51 +0000 | [diff] [blame] | 111 | void TokenLexer::destroy() { |
Reid Spencer | 5f016e2 | 2007-07-11 17:01:13 +0000 | [diff] [blame] | 112 | // If this was a function-like macro that actually uses its arguments, delete |
| 113 | // the expanded tokens. |
Chris Lattner | 8d89643 | 2008-03-09 02:07:49 +0000 | [diff] [blame] | 114 | if (OwnsTokens) { |
| 115 | delete [] Tokens; |
Stephen Hines | 6bcf27b | 2014-05-29 04:14:42 -0700 | [diff] [blame] | 116 | Tokens = nullptr; |
Chris Lattner | 32fca72 | 2009-03-04 06:50:57 +0000 | [diff] [blame] | 117 | OwnsTokens = false; |
Chris Lattner | 9c68306 | 2007-07-22 01:16:55 +0000 | [diff] [blame] | 118 | } |
Mike Stump | 1eb4433 | 2009-09-09 15:08:12 +0000 | [diff] [blame] | 119 | |
Chris Lattner | 1543e9c | 2008-03-09 02:18:51 +0000 | [diff] [blame] | 120 | // TokenLexer owns its formal arguments. |
Chris Lattner | 561395b | 2009-12-14 22:12:52 +0000 | [diff] [blame] | 121 | if (ActualArgs) ActualArgs->destroy(PP); |
Reid Spencer | 5f016e2 | 2007-07-11 17:01:13 +0000 | [diff] [blame] | 122 | } |
| 123 | |
Stephen Hines | 6bcf27b | 2014-05-29 04:14:42 -0700 | [diff] [blame] | 124 | bool TokenLexer::MaybeRemoveCommaBeforeVaArgs( |
| 125 | SmallVectorImpl<Token> &ResultToks, bool HasPasteOperator, MacroInfo *Macro, |
| 126 | unsigned MacroArgNo, Preprocessor &PP) { |
Andy Gibbs | 02f0022 | 2012-11-09 13:24:30 +0000 | [diff] [blame] | 127 | // Is the macro argument __VA_ARGS__? |
| 128 | if (!Macro->isVariadic() || MacroArgNo != Macro->getNumArgs()-1) |
| 129 | return false; |
| 130 | |
| 131 | // In Microsoft-compatibility mode, a comma is removed in the expansion |
| 132 | // of " ... , __VA_ARGS__ " if __VA_ARGS__ is empty. This extension is |
| 133 | // not supported by gcc. |
Stephen Hines | 651f13c | 2014-04-23 16:59:28 -0700 | [diff] [blame] | 134 | if (!HasPasteOperator && !PP.getLangOpts().MSVCCompat) |
Andy Gibbs | 02f0022 | 2012-11-09 13:24:30 +0000 | [diff] [blame] | 135 | return false; |
| 136 | |
| 137 | // GCC removes the comma in the expansion of " ... , ## __VA_ARGS__ " if |
| 138 | // __VA_ARGS__ is empty, but not in strict C99 mode where there are no |
| 139 | // named arguments, where it remains. In all other modes, including C99 |
| 140 | // with GNU extensions, it is removed regardless of named arguments. |
| 141 | // Microsoft also appears to support this extension, unofficially. |
| 142 | if (PP.getLangOpts().C99 && !PP.getLangOpts().GNUMode |
| 143 | && Macro->getNumArgs() < 2) |
| 144 | return false; |
| 145 | |
| 146 | // Is a comma available to be removed? |
| 147 | if (ResultToks.empty() || !ResultToks.back().is(tok::comma)) |
| 148 | return false; |
| 149 | |
| 150 | // Issue an extension diagnostic for the paste operator. |
| 151 | if (HasPasteOperator) |
| 152 | PP.Diag(ResultToks.back().getLocation(), diag::ext_paste_comma); |
| 153 | |
| 154 | // Remove the comma. |
| 155 | ResultToks.pop_back(); |
| 156 | |
| 157 | // If the comma was right after another paste (e.g. "X##,##__VA_ARGS__"), |
| 158 | // then removal of the comma should produce a placemarker token (in C99 |
| 159 | // terms) which we model by popping off the previous ##, giving us a plain |
| 160 | // "X" when __VA_ARGS__ is empty. |
| 161 | if (!ResultToks.empty() && ResultToks.back().is(tok::hashhash)) |
| 162 | ResultToks.pop_back(); |
| 163 | |
| 164 | // Never add a space, even if the comma, ##, or arg had a space. |
| 165 | NextTokGetsSpace = false; |
| 166 | return true; |
| 167 | } |
| 168 | |
Reid Spencer | 5f016e2 | 2007-07-11 17:01:13 +0000 | [diff] [blame] | 169 | /// 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] | 170 | /// return preexpanded tokens from Tokens. |
Chris Lattner | 1543e9c | 2008-03-09 02:18:51 +0000 | [diff] [blame] | 171 | void TokenLexer::ExpandFunctionArguments() { |
Argyrios Kyrtzidis | 0fd9c48 | 2011-07-07 03:40:37 +0000 | [diff] [blame] | 172 | |
Chris Lattner | 5f9e272 | 2011-07-23 10:55:15 +0000 | [diff] [blame] | 173 | SmallVector<Token, 128> ResultToks; |
Mike Stump | 1eb4433 | 2009-09-09 15:08:12 +0000 | [diff] [blame] | 174 | |
Chris Lattner | 8d89643 | 2008-03-09 02:07:49 +0000 | [diff] [blame] | 175 | // Loop through 'Tokens', expanding them into ResultToks. Keep |
Reid Spencer | 5f016e2 | 2007-07-11 17:01:13 +0000 | [diff] [blame] | 176 | // 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] | 177 | // we install the newly expanded sequence as the new 'Tokens' list. |
Reid Spencer | 5f016e2 | 2007-07-11 17:01:13 +0000 | [diff] [blame] | 178 | bool MadeChange = false; |
Mike Stump | 1eb4433 | 2009-09-09 15:08:12 +0000 | [diff] [blame] | 179 | |
Chris Lattner | 8d89643 | 2008-03-09 02:07:49 +0000 | [diff] [blame] | 180 | for (unsigned i = 0, e = NumTokens; i != e; ++i) { |
Reid Spencer | 5f016e2 | 2007-07-11 17:01:13 +0000 | [diff] [blame] | 181 | // If we found the stringify operator, get the argument stringified. The |
| 182 | // preprocessor already verified that the following token is a macro name |
| 183 | // when the #define was parsed. |
Chris Lattner | 8d89643 | 2008-03-09 02:07:49 +0000 | [diff] [blame] | 184 | const Token &CurTok = Tokens[i]; |
Stephen Hines | 651f13c | 2014-04-23 16:59:28 -0700 | [diff] [blame] | 185 | if (i != 0 && !Tokens[i-1].is(tok::hashhash) && CurTok.hasLeadingSpace()) |
| 186 | NextTokGetsSpace = true; |
| 187 | |
Chris Lattner | 22f6bbc | 2007-10-09 18:02:16 +0000 | [diff] [blame] | 188 | if (CurTok.is(tok::hash) || CurTok.is(tok::hashat)) { |
Chris Lattner | 8d89643 | 2008-03-09 02:07:49 +0000 | [diff] [blame] | 189 | int ArgNo = Macro->getArgumentNum(Tokens[i+1].getIdentifierInfo()); |
Reid Spencer | 5f016e2 | 2007-07-11 17:01:13 +0000 | [diff] [blame] | 190 | assert(ArgNo != -1 && "Token following # is not an argument?"); |
Mike Stump | 1eb4433 | 2009-09-09 15:08:12 +0000 | [diff] [blame] | 191 | |
Abramo Bagnara | a08529c | 2011-10-03 18:39:03 +0000 | [diff] [blame] | 192 | SourceLocation ExpansionLocStart = |
Argyrios Kyrtzidis | c0069d8 | 2011-08-23 21:02:32 +0000 | [diff] [blame] | 193 | getExpansionLocForMacroDefLoc(CurTok.getLocation()); |
Abramo Bagnara | a08529c | 2011-10-03 18:39:03 +0000 | [diff] [blame] | 194 | SourceLocation ExpansionLocEnd = |
| 195 | getExpansionLocForMacroDefLoc(Tokens[i+1].getLocation()); |
Argyrios Kyrtzidis | b73377e | 2011-07-07 03:40:34 +0000 | [diff] [blame] | 196 | |
Chris Lattner | d217773 | 2007-07-20 16:59:19 +0000 | [diff] [blame] | 197 | Token Res; |
Chris Lattner | 22f6bbc | 2007-10-09 18:02:16 +0000 | [diff] [blame] | 198 | if (CurTok.is(tok::hash)) // Stringify |
Abramo Bagnara | a08529c | 2011-10-03 18:39:03 +0000 | [diff] [blame] | 199 | Res = ActualArgs->getStringifiedArgument(ArgNo, PP, |
| 200 | ExpansionLocStart, |
| 201 | ExpansionLocEnd); |
Reid Spencer | 5f016e2 | 2007-07-11 17:01:13 +0000 | [diff] [blame] | 202 | else { |
| 203 | // 'charify': don't bother caching these. |
Chris Lattner | e5c8ffe | 2008-03-09 02:55:12 +0000 | [diff] [blame] | 204 | Res = MacroArgs::StringifyArgument(ActualArgs->getUnexpArgument(ArgNo), |
Abramo Bagnara | a08529c | 2011-10-03 18:39:03 +0000 | [diff] [blame] | 205 | PP, true, |
| 206 | ExpansionLocStart, |
| 207 | ExpansionLocEnd); |
Reid Spencer | 5f016e2 | 2007-07-11 17:01:13 +0000 | [diff] [blame] | 208 | } |
Stephen Hines | 0e2c34f | 2015-03-23 12:09:02 -0700 | [diff] [blame] | 209 | Res.setFlag(Token::StringifiedInMacro); |
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 | // The stringified/charified string leading space flag gets set to match |
| 212 | // the #/#@ operator. |
Stephen Hines | 651f13c | 2014-04-23 16:59:28 -0700 | [diff] [blame] | 213 | if (NextTokGetsSpace) |
Chris Lattner | d217773 | 2007-07-20 16:59:19 +0000 | [diff] [blame] | 214 | Res.setFlag(Token::LeadingSpace); |
Mike Stump | 1eb4433 | 2009-09-09 15:08:12 +0000 | [diff] [blame] | 215 | |
Reid Spencer | 5f016e2 | 2007-07-11 17:01:13 +0000 | [diff] [blame] | 216 | ResultToks.push_back(Res); |
| 217 | MadeChange = true; |
| 218 | ++i; // Skip arg name. |
| 219 | NextTokGetsSpace = false; |
| 220 | continue; |
| 221 | } |
Mike Stump | 1eb4433 | 2009-09-09 15:08:12 +0000 | [diff] [blame] | 222 | |
Stephen Hines | 651f13c | 2014-04-23 16:59:28 -0700 | [diff] [blame] | 223 | // Find out if there is a paste (##) operator before or after the token. |
| 224 | bool NonEmptyPasteBefore = |
| 225 | !ResultToks.empty() && ResultToks.back().is(tok::hashhash); |
| 226 | bool PasteBefore = i != 0 && Tokens[i-1].is(tok::hashhash); |
| 227 | bool PasteAfter = i+1 != e && Tokens[i+1].is(tok::hashhash); |
| 228 | assert(!NonEmptyPasteBefore || PasteBefore); |
| 229 | |
Reid Spencer | 5f016e2 | 2007-07-11 17:01:13 +0000 | [diff] [blame] | 230 | // Otherwise, if this is not an argument token, just add the token to the |
| 231 | // output buffer. |
| 232 | IdentifierInfo *II = CurTok.getIdentifierInfo(); |
| 233 | int ArgNo = II ? Macro->getArgumentNum(II) : -1; |
| 234 | if (ArgNo == -1) { |
| 235 | // This isn't an argument, just add it. |
| 236 | ResultToks.push_back(CurTok); |
| 237 | |
| 238 | if (NextTokGetsSpace) { |
Chris Lattner | d217773 | 2007-07-20 16:59:19 +0000 | [diff] [blame] | 239 | ResultToks.back().setFlag(Token::LeadingSpace); |
Reid Spencer | 5f016e2 | 2007-07-11 17:01:13 +0000 | [diff] [blame] | 240 | NextTokGetsSpace = false; |
Stephen Hines | 651f13c | 2014-04-23 16:59:28 -0700 | [diff] [blame] | 241 | } else if (PasteBefore && !NonEmptyPasteBefore) |
| 242 | ResultToks.back().clearFlag(Token::LeadingSpace); |
| 243 | |
Reid Spencer | 5f016e2 | 2007-07-11 17:01:13 +0000 | [diff] [blame] | 244 | continue; |
| 245 | } |
Mike Stump | 1eb4433 | 2009-09-09 15:08:12 +0000 | [diff] [blame] | 246 | |
Reid Spencer | 5f016e2 | 2007-07-11 17:01:13 +0000 | [diff] [blame] | 247 | // An argument is expanded somehow, the result is different than the |
| 248 | // input. |
| 249 | MadeChange = true; |
| 250 | |
Stephen Hines | 651f13c | 2014-04-23 16:59:28 -0700 | [diff] [blame] | 251 | // Otherwise, this is a use of the argument. |
Mike Stump | 1eb4433 | 2009-09-09 15:08:12 +0000 | [diff] [blame] | 252 | |
Andy Gibbs | 02f0022 | 2012-11-09 13:24:30 +0000 | [diff] [blame] | 253 | // In Microsoft mode, remove the comma before __VA_ARGS__ to ensure there |
| 254 | // are no trailing commas if __VA_ARGS__ is empty. |
| 255 | if (!PasteBefore && ActualArgs->isVarargsElidedUse() && |
Stephen Hines | 651f13c | 2014-04-23 16:59:28 -0700 | [diff] [blame] | 256 | MaybeRemoveCommaBeforeVaArgs(ResultToks, |
Andy Gibbs | 02f0022 | 2012-11-09 13:24:30 +0000 | [diff] [blame] | 257 | /*HasPasteOperator=*/false, |
| 258 | Macro, ArgNo, PP)) |
| 259 | continue; |
| 260 | |
Reid Spencer | 5f016e2 | 2007-07-11 17:01:13 +0000 | [diff] [blame] | 261 | // If it is not the LHS/RHS of a ## operator, we must pre-expand the |
| 262 | // argument and substitute the expanded tokens into the result. This is |
| 263 | // C99 6.10.3.1p1. |
| 264 | if (!PasteBefore && !PasteAfter) { |
Chris Lattner | d217773 | 2007-07-20 16:59:19 +0000 | [diff] [blame] | 265 | const Token *ResultArgToks; |
Reid Spencer | 5f016e2 | 2007-07-11 17:01:13 +0000 | [diff] [blame] | 266 | |
| 267 | // Only preexpand the argument if it could possibly need it. This |
| 268 | // avoids some work in common cases. |
Chris Lattner | d217773 | 2007-07-20 16:59:19 +0000 | [diff] [blame] | 269 | const Token *ArgTok = ActualArgs->getUnexpArgument(ArgNo); |
Chris Lattner | cc1a875 | 2007-10-07 08:44:20 +0000 | [diff] [blame] | 270 | if (ActualArgs->ArgNeedsPreexpansion(ArgTok, PP)) |
Chris Lattner | f5809a7 | 2009-12-28 06:17:16 +0000 | [diff] [blame] | 271 | ResultArgToks = &ActualArgs->getPreExpArgument(ArgNo, Macro, PP)[0]; |
Reid Spencer | 5f016e2 | 2007-07-11 17:01:13 +0000 | [diff] [blame] | 272 | else |
| 273 | ResultArgToks = ArgTok; // Use non-preexpanded tokens. |
Mike Stump | 1eb4433 | 2009-09-09 15:08:12 +0000 | [diff] [blame] | 274 | |
Reid Spencer | 5f016e2 | 2007-07-11 17:01:13 +0000 | [diff] [blame] | 275 | // If the arg token expanded into anything, append it. |
Chris Lattner | 22f6bbc | 2007-10-09 18:02:16 +0000 | [diff] [blame] | 276 | if (ResultArgToks->isNot(tok::eof)) { |
Reid Spencer | 5f016e2 | 2007-07-11 17:01:13 +0000 | [diff] [blame] | 277 | unsigned FirstResult = ResultToks.size(); |
| 278 | unsigned NumToks = MacroArgs::getArgLength(ResultArgToks); |
| 279 | ResultToks.append(ResultArgToks, ResultArgToks+NumToks); |
Mike Stump | 1eb4433 | 2009-09-09 15:08:12 +0000 | [diff] [blame] | 280 | |
Reid Kleckner | 11be064 | 2013-06-26 17:16:08 +0000 | [diff] [blame] | 281 | // In Microsoft-compatibility mode, we follow MSVC's preprocessing |
| 282 | // behavior by not considering single commas from nested macro |
| 283 | // expansions as argument separators. Set a flag on the token so we can |
| 284 | // test for this later when the macro expansion is processed. |
Stephen Hines | 651f13c | 2014-04-23 16:59:28 -0700 | [diff] [blame] | 285 | if (PP.getLangOpts().MSVCCompat && NumToks == 1 && |
Reid Kleckner | 11be064 | 2013-06-26 17:16:08 +0000 | [diff] [blame] | 286 | ResultToks.back().is(tok::comma)) |
| 287 | ResultToks.back().setFlag(Token::IgnoredComma); |
| 288 | |
Argyrios Kyrtzidis | 3582a6c | 2011-07-07 18:04:47 +0000 | [diff] [blame] | 289 | // If the '##' came from expanding an argument, turn it into 'unknown' |
| 290 | // to avoid pasting. |
| 291 | for (unsigned i = FirstResult, e = ResultToks.size(); i != e; ++i) { |
| 292 | Token &Tok = ResultToks[i]; |
| 293 | if (Tok.is(tok::hashhash)) |
| 294 | Tok.setKind(tok::unknown); |
| 295 | } |
| 296 | |
Chandler Carruth | d0f5e44 | 2011-07-14 08:20:34 +0000 | [diff] [blame] | 297 | if(ExpandLocStart.isValid()) { |
Argyrios Kyrtzidis | d60a34a | 2011-08-19 22:34:17 +0000 | [diff] [blame] | 298 | updateLocForMacroArgTokens(CurTok.getLocation(), |
| 299 | ResultToks.begin()+FirstResult, |
| 300 | ResultToks.end()); |
Argyrios Kyrtzidis | b73377e | 2011-07-07 03:40:34 +0000 | [diff] [blame] | 301 | } |
| 302 | |
Reid Spencer | 5f016e2 | 2007-07-11 17:01:13 +0000 | [diff] [blame] | 303 | // If any tokens were substituted from the argument, the whitespace |
| 304 | // before the first token should match the whitespace of the arg |
| 305 | // identifier. |
Chris Lattner | d217773 | 2007-07-20 16:59:19 +0000 | [diff] [blame] | 306 | ResultToks[FirstResult].setFlagValue(Token::LeadingSpace, |
Reid Spencer | 5f016e2 | 2007-07-11 17:01:13 +0000 | [diff] [blame] | 307 | NextTokGetsSpace); |
| 308 | NextTokGetsSpace = false; |
Reid Spencer | 5f016e2 | 2007-07-11 17:01:13 +0000 | [diff] [blame] | 309 | } |
| 310 | continue; |
| 311 | } |
Mike Stump | 1eb4433 | 2009-09-09 15:08:12 +0000 | [diff] [blame] | 312 | |
Reid Spencer | 5f016e2 | 2007-07-11 17:01:13 +0000 | [diff] [blame] | 313 | // Okay, we have a token that is either the LHS or RHS of a paste (##) |
| 314 | // argument. It gets substituted as its non-pre-expanded tokens. |
Chris Lattner | d217773 | 2007-07-20 16:59:19 +0000 | [diff] [blame] | 315 | const Token *ArgToks = ActualArgs->getUnexpArgument(ArgNo); |
Reid Spencer | 5f016e2 | 2007-07-11 17:01:13 +0000 | [diff] [blame] | 316 | unsigned NumToks = MacroArgs::getArgLength(ArgToks); |
| 317 | if (NumToks) { // Not an empty argument? |
Richard Smith | 9f728cd | 2012-06-22 23:59:08 +0000 | [diff] [blame] | 318 | // If this is the GNU ", ## __VA_ARGS__" extension, and we just learned |
| 319 | // that __VA_ARGS__ expands to multiple tokens, avoid a pasting error when |
| 320 | // the expander trys to paste ',' with the first token of the __VA_ARGS__ |
Chris Lattner | 71a3a8d | 2008-01-29 07:54:23 +0000 | [diff] [blame] | 321 | // expansion. |
Argyrios Kyrtzidis | 461f2d8 | 2013-05-25 01:35:18 +0000 | [diff] [blame] | 322 | if (NonEmptyPasteBefore && ResultToks.size() >= 2 && |
Chris Lattner | 71a3a8d | 2008-01-29 07:54:23 +0000 | [diff] [blame] | 323 | ResultToks[ResultToks.size()-2].is(tok::comma) && |
| 324 | (unsigned)ArgNo == Macro->getNumArgs()-1 && |
| 325 | Macro->isVariadic()) { |
| 326 | // Remove the paste operator, report use of the extension. |
Robert Wilhelm | 344472e | 2013-08-23 16:11:15 +0000 | [diff] [blame] | 327 | PP.Diag(ResultToks.pop_back_val().getLocation(), diag::ext_paste_comma); |
Chris Lattner | 71a3a8d | 2008-01-29 07:54:23 +0000 | [diff] [blame] | 328 | } |
Mike Stump | 1eb4433 | 2009-09-09 15:08:12 +0000 | [diff] [blame] | 329 | |
Reid Spencer | 5f016e2 | 2007-07-11 17:01:13 +0000 | [diff] [blame] | 330 | ResultToks.append(ArgToks, ArgToks+NumToks); |
Mike Stump | 1eb4433 | 2009-09-09 15:08:12 +0000 | [diff] [blame] | 331 | |
Argyrios Kyrtzidis | 3582a6c | 2011-07-07 18:04:47 +0000 | [diff] [blame] | 332 | // If the '##' came from expanding an argument, turn it into 'unknown' |
| 333 | // to avoid pasting. |
| 334 | for (unsigned i = ResultToks.size() - NumToks, e = ResultToks.size(); |
| 335 | i != e; ++i) { |
| 336 | Token &Tok = ResultToks[i]; |
| 337 | if (Tok.is(tok::hashhash)) |
| 338 | Tok.setKind(tok::unknown); |
| 339 | } |
| 340 | |
Chandler Carruth | d0f5e44 | 2011-07-14 08:20:34 +0000 | [diff] [blame] | 341 | if (ExpandLocStart.isValid()) { |
Argyrios Kyrtzidis | d60a34a | 2011-08-19 22:34:17 +0000 | [diff] [blame] | 342 | updateLocForMacroArgTokens(CurTok.getLocation(), |
| 343 | ResultToks.end()-NumToks, ResultToks.end()); |
Argyrios Kyrtzidis | b73377e | 2011-07-07 03:40:34 +0000 | [diff] [blame] | 344 | } |
| 345 | |
Chris Lattner | 442a661 | 2009-05-25 16:23:08 +0000 | [diff] [blame] | 346 | // If this token (the macro argument) was supposed to get leading |
| 347 | // whitespace, transfer this information onto the first token of the |
| 348 | // expansion. |
| 349 | // |
| 350 | // Do not do this if the paste operator occurs before the macro argument, |
| 351 | // as in "A ## MACROARG". In valid code, the first token will get |
| 352 | // smooshed onto the preceding one anyway (forming AMACROARG). In |
| 353 | // assembler-with-cpp mode, invalid pastes are allowed through: in this |
| 354 | // case, we do not want the extra whitespace to be added. For example, |
| 355 | // we want ". ## foo" -> ".foo" not ". foo". |
Stephen Hines | 651f13c | 2014-04-23 16:59:28 -0700 | [diff] [blame] | 356 | if (NextTokGetsSpace) |
Chris Lattner | 442a661 | 2009-05-25 16:23:08 +0000 | [diff] [blame] | 357 | ResultToks[ResultToks.size()-NumToks].setFlag(Token::LeadingSpace); |
Mike Stump | 1eb4433 | 2009-09-09 15:08:12 +0000 | [diff] [blame] | 358 | |
Chris Lattner | 442a661 | 2009-05-25 16:23:08 +0000 | [diff] [blame] | 359 | NextTokGetsSpace = false; |
Reid Spencer | 5f016e2 | 2007-07-11 17:01:13 +0000 | [diff] [blame] | 360 | continue; |
| 361 | } |
Mike Stump | 1eb4433 | 2009-09-09 15:08:12 +0000 | [diff] [blame] | 362 | |
Reid Spencer | 5f016e2 | 2007-07-11 17:01:13 +0000 | [diff] [blame] | 363 | // If an empty argument is on the LHS or RHS of a paste, the standard (C99 |
| 364 | // 6.10.3.3p2,3) calls for a bunch of placemarker stuff to occur. We |
| 365 | // implement this by eating ## operators when a LHS or RHS expands to |
| 366 | // empty. |
Reid Spencer | 5f016e2 | 2007-07-11 17:01:13 +0000 | [diff] [blame] | 367 | if (PasteAfter) { |
| 368 | // Discard the argument token and skip (don't copy to the expansion |
| 369 | // buffer) the paste operator after it. |
Reid Spencer | 5f016e2 | 2007-07-11 17:01:13 +0000 | [diff] [blame] | 370 | ++i; |
| 371 | continue; |
| 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 on the RHS of a paste operator, we've already copied the |
Argyrios Kyrtzidis | 461f2d8 | 2013-05-25 01:35:18 +0000 | [diff] [blame] | 375 | // paste operator to the ResultToks list, unless the LHS was empty too. |
| 376 | // Remove it. |
| 377 | assert(PasteBefore); |
| 378 | if (NonEmptyPasteBefore) { |
| 379 | assert(ResultToks.back().is(tok::hashhash)); |
Stephen Hines | 651f13c | 2014-04-23 16:59:28 -0700 | [diff] [blame] | 380 | ResultToks.pop_back(); |
Argyrios Kyrtzidis | 461f2d8 | 2013-05-25 01:35:18 +0000 | [diff] [blame] | 381 | } |
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 is the __VA_ARGS__ token, and if the argument wasn't provided, |
| 384 | // and if the macro had at least one real argument, and if the token before |
Andy Gibbs | 02f0022 | 2012-11-09 13:24:30 +0000 | [diff] [blame] | 385 | // the ## was a comma, remove the comma. This is a GCC extension which is |
| 386 | // disabled when using -std=c99. |
| 387 | if (ActualArgs->isVarargsElidedUse()) |
Stephen Hines | 651f13c | 2014-04-23 16:59:28 -0700 | [diff] [blame] | 388 | MaybeRemoveCommaBeforeVaArgs(ResultToks, |
Andy Gibbs | 02f0022 | 2012-11-09 13:24:30 +0000 | [diff] [blame] | 389 | /*HasPasteOperator=*/true, |
| 390 | Macro, ArgNo, PP); |
| 391 | |
Reid Spencer | 5f016e2 | 2007-07-11 17:01:13 +0000 | [diff] [blame] | 392 | continue; |
| 393 | } |
Mike Stump | 1eb4433 | 2009-09-09 15:08:12 +0000 | [diff] [blame] | 394 | |
Chris Lattner | 8d89643 | 2008-03-09 02:07:49 +0000 | [diff] [blame] | 395 | // If anything changed, install this as the new Tokens list. |
Reid Spencer | 5f016e2 | 2007-07-11 17:01:13 +0000 | [diff] [blame] | 396 | if (MadeChange) { |
Chris Lattner | 32fca72 | 2009-03-04 06:50:57 +0000 | [diff] [blame] | 397 | assert(!OwnsTokens && "This would leak if we already own the token list"); |
Reid Spencer | 5f016e2 | 2007-07-11 17:01:13 +0000 | [diff] [blame] | 398 | // This is deleted in the dtor. |
Chris Lattner | 8d89643 | 2008-03-09 02:07:49 +0000 | [diff] [blame] | 399 | NumTokens = ResultToks.size(); |
Argyrios Kyrtzidis | 5b3284a | 2011-06-29 22:20:11 +0000 | [diff] [blame] | 400 | // The tokens will be added to Preprocessor's cache and will be removed |
| 401 | // when this TokenLexer finishes lexing them. |
| 402 | Tokens = PP.cacheMacroExpandedTokens(this, ResultToks); |
Mike Stump | 1eb4433 | 2009-09-09 15:08:12 +0000 | [diff] [blame] | 403 | |
Argyrios Kyrtzidis | 5b3284a | 2011-06-29 22:20:11 +0000 | [diff] [blame] | 404 | // The preprocessor cache of macro expanded tokens owns these tokens,not us. |
Chris Lattner | 32fca72 | 2009-03-04 06:50:57 +0000 | [diff] [blame] | 405 | OwnsTokens = false; |
Reid Spencer | 5f016e2 | 2007-07-11 17:01:13 +0000 | [diff] [blame] | 406 | } |
| 407 | } |
| 408 | |
Stephen Hines | 0e2c34f | 2015-03-23 12:09:02 -0700 | [diff] [blame] | 409 | /// \brief Checks if two tokens form wide string literal. |
| 410 | static bool isWideStringLiteralFromMacro(const Token &FirstTok, |
| 411 | const Token &SecondTok) { |
| 412 | return FirstTok.is(tok::identifier) && |
| 413 | FirstTok.getIdentifierInfo()->isStr("L") && SecondTok.isLiteral() && |
| 414 | SecondTok.stringifiedInMacro(); |
| 415 | } |
| 416 | |
Reid Spencer | 5f016e2 | 2007-07-11 17:01:13 +0000 | [diff] [blame] | 417 | /// Lex - Lex and return a token from this macro stream. |
| 418 | /// |
Eli Friedman | d2f9308 | 2013-09-19 00:41:32 +0000 | [diff] [blame] | 419 | bool TokenLexer::Lex(Token &Tok) { |
Reid Spencer | 5f016e2 | 2007-07-11 17:01:13 +0000 | [diff] [blame] | 420 | // Lexing off the end of the macro, pop this macro off the expansion stack. |
| 421 | if (isAtEnd()) { |
| 422 | // If this is a macro (not a token stream), mark the macro enabled now |
| 423 | // that it is no longer being expanded. |
| 424 | if (Macro) Macro->EnableMacro(); |
| 425 | |
Eli Friedman | d2f9308 | 2013-09-19 00:41:32 +0000 | [diff] [blame] | 426 | Tok.startToken(); |
| 427 | Tok.setFlagValue(Token::StartOfLine , AtStartOfLine); |
Stephen Hines | 651f13c | 2014-04-23 16:59:28 -0700 | [diff] [blame] | 428 | Tok.setFlagValue(Token::LeadingSpace, HasLeadingSpace || NextTokGetsSpace); |
Eli Friedman | d2f9308 | 2013-09-19 00:41:32 +0000 | [diff] [blame] | 429 | if (CurToken == 0) |
| 430 | Tok.setFlag(Token::LeadingEmptyMacro); |
| 431 | return PP.HandleEndOfTokenLexer(Tok); |
Reid Spencer | 5f016e2 | 2007-07-11 17:01:13 +0000 | [diff] [blame] | 432 | } |
Mike Stump | 1eb4433 | 2009-09-09 15:08:12 +0000 | [diff] [blame] | 433 | |
Argyrios Kyrtzidis | 0fd9c48 | 2011-07-07 03:40:37 +0000 | [diff] [blame] | 434 | SourceManager &SM = PP.getSourceManager(); |
| 435 | |
Reid Spencer | 5f016e2 | 2007-07-11 17:01:13 +0000 | [diff] [blame] | 436 | // If this is the first token of the expanded result, we inherit spacing |
| 437 | // properties later. |
| 438 | bool isFirstToken = CurToken == 0; |
Mike Stump | 1eb4433 | 2009-09-09 15:08:12 +0000 | [diff] [blame] | 439 | |
Reid Spencer | 5f016e2 | 2007-07-11 17:01:13 +0000 | [diff] [blame] | 440 | // Get the next token to return. |
Chris Lattner | 8d89643 | 2008-03-09 02:07:49 +0000 | [diff] [blame] | 441 | Tok = Tokens[CurToken++]; |
Mike Stump | 1eb4433 | 2009-09-09 15:08:12 +0000 | [diff] [blame] | 442 | |
Chris Lattner | 332ee08 | 2009-04-19 20:29:42 +0000 | [diff] [blame] | 443 | bool TokenIsFromPaste = false; |
Mike Stump | 1eb4433 | 2009-09-09 15:08:12 +0000 | [diff] [blame] | 444 | |
Reid Spencer | 5f016e2 | 2007-07-11 17:01:13 +0000 | [diff] [blame] | 445 | // 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] | 446 | // Note that ## is a normal token when not expanding a macro. |
Stephen Hines | 0e2c34f | 2015-03-23 12:09:02 -0700 | [diff] [blame] | 447 | if (!isAtEnd() && Macro && |
| 448 | (Tokens[CurToken].is(tok::hashhash) || |
| 449 | // Special processing of L#x macros in -fms-compatibility mode. |
| 450 | // Microsoft compiler is able to form a wide string literal from |
| 451 | // 'L#macro_arg' construct in a function-like macro. |
| 452 | (PP.getLangOpts().MSVCCompat && |
| 453 | isWideStringLiteralFromMacro(Tok, Tokens[CurToken])))) { |
Chris Lattner | 04f2bd8 | 2009-12-04 06:14:03 +0000 | [diff] [blame] | 454 | // When handling the microsoft /##/ extension, the final token is |
| 455 | // returned by PasteTokens, not the pasted token. |
| 456 | if (PasteTokens(Tok)) |
Eli Friedman | d2f9308 | 2013-09-19 00:41:32 +0000 | [diff] [blame] | 457 | return true; |
Kovarththanan Rajaratnam | 65cc1e8 | 2010-03-13 08:53:33 +0000 | [diff] [blame] | 458 | |
Chris Lattner | 04f2bd8 | 2009-12-04 06:14:03 +0000 | [diff] [blame] | 459 | TokenIsFromPaste = true; |
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 | |
| 462 | // The token's current location indicate where the token was lexed from. We |
| 463 | // need this information to compute the spelling of the token, but any |
| 464 | // diagnostics for the expanded token should appear as if they came from |
Chandler Carruth | d0f5e44 | 2011-07-14 08:20:34 +0000 | [diff] [blame] | 465 | // ExpansionLoc. Pull this information together into a new SourceLocation |
Reid Spencer | 5f016e2 | 2007-07-11 17:01:13 +0000 | [diff] [blame] | 466 | // that captures all of this. |
Chandler Carruth | d0f5e44 | 2011-07-14 08:20:34 +0000 | [diff] [blame] | 467 | if (ExpandLocStart.isValid() && // Don't do this for token streams. |
Argyrios Kyrtzidis | b73377e | 2011-07-07 03:40:34 +0000 | [diff] [blame] | 468 | // Check that the token's location was not already set properly. |
Argyrios Kyrtzidis | 5d579e7 | 2011-08-23 21:02:35 +0000 | [diff] [blame] | 469 | SM.isBeforeInSLocAddrSpace(Tok.getLocation(), MacroStartSLocOffset)) { |
Argyrios Kyrtzidis | b73377e | 2011-07-07 03:40:34 +0000 | [diff] [blame] | 470 | SourceLocation instLoc; |
| 471 | if (Tok.is(tok::comment)) { |
Chandler Carruth | bf340e4 | 2011-07-26 03:03:05 +0000 | [diff] [blame] | 472 | instLoc = SM.createExpansionLoc(Tok.getLocation(), |
| 473 | ExpandLocStart, |
| 474 | ExpandLocEnd, |
| 475 | Tok.getLength()); |
Argyrios Kyrtzidis | b73377e | 2011-07-07 03:40:34 +0000 | [diff] [blame] | 476 | } else { |
Argyrios Kyrtzidis | ec3b538 | 2011-08-19 22:34:14 +0000 | [diff] [blame] | 477 | instLoc = getExpansionLocForMacroDefLoc(Tok.getLocation()); |
Argyrios Kyrtzidis | b73377e | 2011-07-07 03:40:34 +0000 | [diff] [blame] | 478 | } |
| 479 | |
| 480 | Tok.setLocation(instLoc); |
Reid Spencer | 5f016e2 | 2007-07-11 17:01:13 +0000 | [diff] [blame] | 481 | } |
Mike Stump | 1eb4433 | 2009-09-09 15:08:12 +0000 | [diff] [blame] | 482 | |
Reid Spencer | 5f016e2 | 2007-07-11 17:01:13 +0000 | [diff] [blame] | 483 | // If this is the first token, set the lexical properties of the token to |
| 484 | // match the lexical properties of the macro identifier. |
| 485 | if (isFirstToken) { |
Chris Lattner | d217773 | 2007-07-20 16:59:19 +0000 | [diff] [blame] | 486 | Tok.setFlagValue(Token::StartOfLine , AtStartOfLine); |
| 487 | Tok.setFlagValue(Token::LeadingSpace, HasLeadingSpace); |
Stephen Hines | 651f13c | 2014-04-23 16:59:28 -0700 | [diff] [blame] | 488 | } else { |
| 489 | // If this is not the first token, we may still need to pass through |
| 490 | // leading whitespace if we've expanded a macro. |
| 491 | if (AtStartOfLine) Tok.setFlag(Token::StartOfLine); |
| 492 | if (HasLeadingSpace) Tok.setFlag(Token::LeadingSpace); |
Reid Spencer | 5f016e2 | 2007-07-11 17:01:13 +0000 | [diff] [blame] | 493 | } |
Stephen Hines | 651f13c | 2014-04-23 16:59:28 -0700 | [diff] [blame] | 494 | AtStartOfLine = false; |
| 495 | HasLeadingSpace = false; |
Mike Stump | 1eb4433 | 2009-09-09 15:08:12 +0000 | [diff] [blame] | 496 | |
Reid Spencer | 5f016e2 | 2007-07-11 17:01:13 +0000 | [diff] [blame] | 497 | // Handle recursive expansion! |
Stephen Hines | 6bcf27b | 2014-05-29 04:14:42 -0700 | [diff] [blame] | 498 | if (!Tok.isAnnotation() && Tok.getIdentifierInfo() != nullptr) { |
Chris Lattner | 863c486 | 2009-01-23 18:35:48 +0000 | [diff] [blame] | 499 | // Change the kind of this identifier to the appropriate token kind, e.g. |
| 500 | // turning "for" into a keyword. |
Argyrios Kyrtzidis | 6f421f9 | 2009-05-22 21:09:31 +0000 | [diff] [blame] | 501 | IdentifierInfo *II = Tok.getIdentifierInfo(); |
Chris Lattner | 863c486 | 2009-01-23 18:35:48 +0000 | [diff] [blame] | 502 | Tok.setKind(II->getTokenID()); |
Mike Stump | 1eb4433 | 2009-09-09 15:08:12 +0000 | [diff] [blame] | 503 | |
Chris Lattner | 332ee08 | 2009-04-19 20:29:42 +0000 | [diff] [blame] | 504 | // If this identifier was poisoned and from a paste, emit an error. This |
| 505 | // won't be handled by Preprocessor::HandleIdentifier because this is coming |
| 506 | // from a macro expansion. |
| 507 | if (II->isPoisoned() && TokenIsFromPaste) { |
John Wiegley | 28bbe4b | 2011-04-28 01:08:34 +0000 | [diff] [blame] | 508 | PP.HandlePoisonedIdentifier(Tok); |
Chris Lattner | 332ee08 | 2009-04-19 20:29:42 +0000 | [diff] [blame] | 509 | } |
Mike Stump | 1eb4433 | 2009-09-09 15:08:12 +0000 | [diff] [blame] | 510 | |
Chris Lattner | 863c486 | 2009-01-23 18:35:48 +0000 | [diff] [blame] | 511 | if (!DisableMacroExpansion && II->isHandleIdentifierCase()) |
Eli Friedman | d2f9308 | 2013-09-19 00:41:32 +0000 | [diff] [blame] | 512 | return PP.HandleIdentifier(Tok); |
Chris Lattner | 863c486 | 2009-01-23 18:35:48 +0000 | [diff] [blame] | 513 | } |
Reid Spencer | 5f016e2 | 2007-07-11 17:01:13 +0000 | [diff] [blame] | 514 | |
| 515 | // Otherwise, return a normal token. |
Eli Friedman | d2f9308 | 2013-09-19 00:41:32 +0000 | [diff] [blame] | 516 | return true; |
Reid Spencer | 5f016e2 | 2007-07-11 17:01:13 +0000 | [diff] [blame] | 517 | } |
| 518 | |
| 519 | /// PasteTokens - Tok is the LHS of a ## operator, and CurToken is the ## |
| 520 | /// 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] | 521 | /// are more ## after it, chomp them iteratively. Return the result as Tok. |
Chris Lattner | 3f1cc83 | 2008-02-07 06:03:59 +0000 | [diff] [blame] | 522 | /// If this returns true, the caller should immediately return the token. |
Chris Lattner | 1543e9c | 2008-03-09 02:18:51 +0000 | [diff] [blame] | 523 | bool TokenLexer::PasteTokens(Token &Tok) { |
Dylan Noblesmith | f7ccbad | 2012-02-05 02:13:05 +0000 | [diff] [blame] | 524 | SmallString<128> Buffer; |
Stephen Hines | 6bcf27b | 2014-05-29 04:14:42 -0700 | [diff] [blame] | 525 | const char *ResultTokStrPtr = nullptr; |
Abramo Bagnara | a08529c | 2011-10-03 18:39:03 +0000 | [diff] [blame] | 526 | SourceLocation StartLoc = Tok.getLocation(); |
Argyrios Kyrtzidis | b73377e | 2011-07-07 03:40:34 +0000 | [diff] [blame] | 527 | SourceLocation PasteOpLoc; |
Reid Spencer | 5f016e2 | 2007-07-11 17:01:13 +0000 | [diff] [blame] | 528 | do { |
Stephen Hines | 0e2c34f | 2015-03-23 12:09:02 -0700 | [diff] [blame] | 529 | // Consume the ## operator if any. |
Argyrios Kyrtzidis | b73377e | 2011-07-07 03:40:34 +0000 | [diff] [blame] | 530 | PasteOpLoc = Tokens[CurToken].getLocation(); |
Stephen Hines | 0e2c34f | 2015-03-23 12:09:02 -0700 | [diff] [blame] | 531 | if (Tokens[CurToken].is(tok::hashhash)) |
| 532 | ++CurToken; |
Reid Spencer | 5f016e2 | 2007-07-11 17:01:13 +0000 | [diff] [blame] | 533 | assert(!isAtEnd() && "No token on the RHS of a paste operator!"); |
Mike Stump | 1eb4433 | 2009-09-09 15:08:12 +0000 | [diff] [blame] | 534 | |
Reid Spencer | 5f016e2 | 2007-07-11 17:01:13 +0000 | [diff] [blame] | 535 | // Get the RHS token. |
Chris Lattner | 8d89643 | 2008-03-09 02:07:49 +0000 | [diff] [blame] | 536 | const Token &RHS = Tokens[CurToken]; |
Mike Stump | 1eb4433 | 2009-09-09 15:08:12 +0000 | [diff] [blame] | 537 | |
Reid Spencer | 5f016e2 | 2007-07-11 17:01:13 +0000 | [diff] [blame] | 538 | // 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] | 539 | // the two tokens. |
| 540 | Buffer.resize(Tok.getLength() + RHS.getLength()); |
Mike Stump | 1eb4433 | 2009-09-09 15:08:12 +0000 | [diff] [blame] | 541 | |
Reid Spencer | 5f016e2 | 2007-07-11 17:01:13 +0000 | [diff] [blame] | 542 | // Get the spelling of the LHS token in Buffer. |
| 543 | const char *BufPtr = &Buffer[0]; |
Douglas Gregor | 453091c | 2010-03-16 22:30:13 +0000 | [diff] [blame] | 544 | bool Invalid = false; |
| 545 | unsigned LHSLen = PP.getSpelling(Tok, BufPtr, &Invalid); |
Reid Spencer | 5f016e2 | 2007-07-11 17:01:13 +0000 | [diff] [blame] | 546 | if (BufPtr != &Buffer[0]) // Really, we want the chars in Buffer! |
| 547 | memcpy(&Buffer[0], BufPtr, LHSLen); |
Douglas Gregor | 453091c | 2010-03-16 22:30:13 +0000 | [diff] [blame] | 548 | if (Invalid) |
| 549 | return true; |
Stephen Hines | 176edba | 2014-12-01 14:53:08 -0800 | [diff] [blame] | 550 | |
| 551 | BufPtr = Buffer.data() + LHSLen; |
Douglas Gregor | 453091c | 2010-03-16 22:30:13 +0000 | [diff] [blame] | 552 | unsigned RHSLen = PP.getSpelling(RHS, BufPtr, &Invalid); |
| 553 | if (Invalid) |
| 554 | return true; |
Stephen Hines | 176edba | 2014-12-01 14:53:08 -0800 | [diff] [blame] | 555 | if (RHSLen && BufPtr != &Buffer[LHSLen]) |
| 556 | // Really, we want the chars in Buffer! |
Reid Spencer | 5f016e2 | 2007-07-11 17:01:13 +0000 | [diff] [blame] | 557 | memcpy(&Buffer[LHSLen], BufPtr, RHSLen); |
Mike Stump | 1eb4433 | 2009-09-09 15:08:12 +0000 | [diff] [blame] | 558 | |
Reid Spencer | 5f016e2 | 2007-07-11 17:01:13 +0000 | [diff] [blame] | 559 | // Trim excess space. |
Chris Lattner | 1fa4953 | 2009-03-08 08:08:45 +0000 | [diff] [blame] | 560 | Buffer.resize(LHSLen+RHSLen); |
Mike Stump | 1eb4433 | 2009-09-09 15:08:12 +0000 | [diff] [blame] | 561 | |
Reid Spencer | 5f016e2 | 2007-07-11 17:01:13 +0000 | [diff] [blame] | 562 | // Plop the pasted result (including the trailing newline and null) into a |
| 563 | // scratch buffer where we can lex it. |
Chris Lattner | 47246be | 2009-01-26 19:29:26 +0000 | [diff] [blame] | 564 | Token ResultTokTmp; |
| 565 | ResultTokTmp.startToken(); |
Mike Stump | 1eb4433 | 2009-09-09 15:08:12 +0000 | [diff] [blame] | 566 | |
Chris Lattner | 47246be | 2009-01-26 19:29:26 +0000 | [diff] [blame] | 567 | // 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] | 568 | // character pointer back from CreateString in getLiteralData(). |
Chris Lattner | 47246be | 2009-01-26 19:29:26 +0000 | [diff] [blame] | 569 | ResultTokTmp.setKind(tok::string_literal); |
Dmitri Gribenko | 374b383 | 2012-09-24 21:07:17 +0000 | [diff] [blame] | 570 | PP.CreateString(Buffer, ResultTokTmp); |
Chris Lattner | 47246be | 2009-01-26 19:29:26 +0000 | [diff] [blame] | 571 | SourceLocation ResultTokLoc = ResultTokTmp.getLocation(); |
| 572 | ResultTokStrPtr = ResultTokTmp.getLiteralData(); |
| 573 | |
Reid Spencer | 5f016e2 | 2007-07-11 17:01:13 +0000 | [diff] [blame] | 574 | // Lex the resultant pasted token into Result. |
Chris Lattner | d217773 | 2007-07-20 16:59:19 +0000 | [diff] [blame] | 575 | Token Result; |
Mike Stump | 1eb4433 | 2009-09-09 15:08:12 +0000 | [diff] [blame] | 576 | |
Abramo Bagnara | c4bf2b9 | 2010-12-22 08:23:18 +0000 | [diff] [blame] | 577 | if (Tok.isAnyIdentifier() && RHS.isAnyIdentifier()) { |
Reid Spencer | 5f016e2 | 2007-07-11 17:01:13 +0000 | [diff] [blame] | 578 | // Common paste case: identifier+identifier = identifier. Avoid creating |
| 579 | // a lexer and other overhead. |
| 580 | PP.IncrementPasteCounter(true); |
| 581 | Result.startToken(); |
Abramo Bagnara | c4bf2b9 | 2010-12-22 08:23:18 +0000 | [diff] [blame] | 582 | Result.setKind(tok::raw_identifier); |
| 583 | Result.setRawIdentifierData(ResultTokStrPtr); |
Reid Spencer | 5f016e2 | 2007-07-11 17:01:13 +0000 | [diff] [blame] | 584 | Result.setLocation(ResultTokLoc); |
| 585 | Result.setLength(LHSLen+RHSLen); |
| 586 | } else { |
| 587 | PP.IncrementPasteCounter(false); |
Mike Stump | 1eb4433 | 2009-09-09 15:08:12 +0000 | [diff] [blame] | 588 | |
Chris Lattner | bcc2a67 | 2009-01-19 06:46:35 +0000 | [diff] [blame] | 589 | assert(ResultTokLoc.isFileID() && |
| 590 | "Should be a raw location into scratch buffer"); |
Reid Spencer | 5f016e2 | 2007-07-11 17:01:13 +0000 | [diff] [blame] | 591 | SourceManager &SourceMgr = PP.getSourceManager(); |
Chris Lattner | 47246be | 2009-01-26 19:29:26 +0000 | [diff] [blame] | 592 | FileID LocFileID = SourceMgr.getFileID(ResultTokLoc); |
Mike Stump | 1eb4433 | 2009-09-09 15:08:12 +0000 | [diff] [blame] | 593 | |
Douglas Gregor | f715ca1 | 2010-03-16 00:06:06 +0000 | [diff] [blame] | 594 | bool Invalid = false; |
Douglas Gregor | aea67db | 2010-03-15 22:54:52 +0000 | [diff] [blame] | 595 | const char *ScratchBufStart |
Benjamin Kramer | f6ac97b | 2010-03-16 14:14:31 +0000 | [diff] [blame] | 596 | = SourceMgr.getBufferData(LocFileID, &Invalid).data(); |
Douglas Gregor | f715ca1 | 2010-03-16 00:06:06 +0000 | [diff] [blame] | 597 | if (Invalid) |
Douglas Gregor | aea67db | 2010-03-15 22:54:52 +0000 | [diff] [blame] | 598 | return false; |
Mike Stump | 1eb4433 | 2009-09-09 15:08:12 +0000 | [diff] [blame] | 599 | |
Chris Lattner | bcc2a67 | 2009-01-19 06:46:35 +0000 | [diff] [blame] | 600 | // 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] | 601 | // 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] | 602 | Lexer TL(SourceMgr.getLocForStartOfFile(LocFileID), |
David Blaikie | 4e4d084 | 2012-03-11 07:00:24 +0000 | [diff] [blame] | 603 | PP.getLangOpts(), ScratchBufStart, |
Chris Lattner | 1fa4953 | 2009-03-08 08:08:45 +0000 | [diff] [blame] | 604 | ResultTokStrPtr, ResultTokStrPtr+LHSLen+RHSLen); |
Mike Stump | 1eb4433 | 2009-09-09 15:08:12 +0000 | [diff] [blame] | 605 | |
Reid Spencer | 5f016e2 | 2007-07-11 17:01:13 +0000 | [diff] [blame] | 606 | // Lex a token in raw mode. This way it won't look up identifiers |
| 607 | // automatically, lexing off the end will return an eof token, and |
| 608 | // warnings are disabled. This returns true if the result token is the |
| 609 | // entire buffer. |
Chris Lattner | 1fa4953 | 2009-03-08 08:08:45 +0000 | [diff] [blame] | 610 | bool isInvalid = !TL.LexFromRawLexer(Result); |
Mike Stump | 1eb4433 | 2009-09-09 15:08:12 +0000 | [diff] [blame] | 611 | |
Reid Spencer | 5f016e2 | 2007-07-11 17:01:13 +0000 | [diff] [blame] | 612 | // If we got an EOF token, we didn't form even ONE token. For example, we |
| 613 | // did "/ ## /" to get "//". |
Chris Lattner | 1fa4953 | 2009-03-08 08:08:45 +0000 | [diff] [blame] | 614 | isInvalid |= Result.is(tok::eof); |
Mike Stump | 1eb4433 | 2009-09-09 15:08:12 +0000 | [diff] [blame] | 615 | |
Chris Lattner | 1fa4953 | 2009-03-08 08:08:45 +0000 | [diff] [blame] | 616 | // If pasting the two tokens didn't form a full new token, this is an |
| 617 | // error. This occurs with "x ## +" and other stuff. Return with Tok |
| 618 | // unmodified and with RHS as the next token to lex. |
| 619 | if (isInvalid) { |
| 620 | // Test for the Microsoft extension of /##/ turning into // here on the |
| 621 | // error path. |
David Blaikie | 4e4d084 | 2012-03-11 07:00:24 +0000 | [diff] [blame] | 622 | if (PP.getLangOpts().MicrosoftExt && Tok.is(tok::slash) && |
Chris Lattner | 1fa4953 | 2009-03-08 08:08:45 +0000 | [diff] [blame] | 623 | RHS.is(tok::slash)) { |
| 624 | HandleMicrosoftCommentPaste(Tok); |
| 625 | return true; |
| 626 | } |
Mike Stump | 1eb4433 | 2009-09-09 15:08:12 +0000 | [diff] [blame] | 627 | |
Chris Lattner | 920bf45 | 2010-07-17 16:24:30 +0000 | [diff] [blame] | 628 | // Do not emit the error when preprocessing assembler code. |
David Blaikie | 4e4d084 | 2012-03-11 07:00:24 +0000 | [diff] [blame] | 629 | if (!PP.getLangOpts().AsmPreprocessor) { |
Chandler Carruth | d0f5e44 | 2011-07-14 08:20:34 +0000 | [diff] [blame] | 630 | // Explicitly convert the token location to have proper expansion |
Chris Lattner | 3c17b8a | 2009-05-28 05:39:39 +0000 | [diff] [blame] | 631 | // information so that the user knows where it came from. |
| 632 | SourceManager &SM = PP.getSourceManager(); |
Chris Lattner | b11e43c | 2011-06-14 18:19:37 +0000 | [diff] [blame] | 633 | SourceLocation Loc = |
Chandler Carruth | bf340e4 | 2011-07-26 03:03:05 +0000 | [diff] [blame] | 634 | SM.createExpansionLoc(PasteOpLoc, ExpandLocStart, ExpandLocEnd, 2); |
Chris Lattner | 920bf45 | 2010-07-17 16:24:30 +0000 | [diff] [blame] | 635 | // If we're in microsoft extensions mode, downgrade this from a hard |
Stephen Hines | 651f13c | 2014-04-23 16:59:28 -0700 | [diff] [blame] | 636 | // error to an extension that defaults to an error. This allows |
Chris Lattner | 920bf45 | 2010-07-17 16:24:30 +0000 | [diff] [blame] | 637 | // disabling it. |
Stephen Hines | 651f13c | 2014-04-23 16:59:28 -0700 | [diff] [blame] | 638 | PP.Diag(Loc, PP.getLangOpts().MicrosoftExt ? diag::ext_pp_bad_paste_ms |
| 639 | : diag::err_pp_bad_paste) |
Pirama Arumuga Nainar | 3ea9e33 | 2015-04-08 08:57:32 -0700 | [diff] [blame] | 640 | << Buffer; |
Chris Lattner | 3c17b8a | 2009-05-28 05:39:39 +0000 | [diff] [blame] | 641 | } |
Mike Stump | 1eb4433 | 2009-09-09 15:08:12 +0000 | [diff] [blame] | 642 | |
Richard Smith | c5f7459 | 2012-06-13 19:02:56 +0000 | [diff] [blame] | 643 | // An error has occurred so exit loop. |
| 644 | break; |
Chris Lattner | 1fa4953 | 2009-03-08 08:08:45 +0000 | [diff] [blame] | 645 | } |
Mike Stump | 1eb4433 | 2009-09-09 15:08:12 +0000 | [diff] [blame] | 646 | |
Chris Lattner | 1fa4953 | 2009-03-08 08:08:45 +0000 | [diff] [blame] | 647 | // Turn ## into 'unknown' to avoid # ## # from looking like a paste |
| 648 | // operator. |
| 649 | if (Result.is(tok::hashhash)) |
| 650 | Result.setKind(tok::unknown); |
Reid Spencer | 5f016e2 | 2007-07-11 17:01:13 +0000 | [diff] [blame] | 651 | } |
Mike Stump | 1eb4433 | 2009-09-09 15:08:12 +0000 | [diff] [blame] | 652 | |
Sylvestre Ledru | bed28ac | 2012-07-23 08:59:39 +0000 | [diff] [blame] | 653 | // Transfer properties of the LHS over the Result. |
Chris Lattner | d217773 | 2007-07-20 16:59:19 +0000 | [diff] [blame] | 654 | Result.setFlagValue(Token::StartOfLine , Tok.isAtStartOfLine()); |
| 655 | Result.setFlagValue(Token::LeadingSpace, Tok.hasLeadingSpace()); |
Argyrios Kyrtzidis | b73377e | 2011-07-07 03:40:34 +0000 | [diff] [blame] | 656 | |
Reid Spencer | 5f016e2 | 2007-07-11 17:01:13 +0000 | [diff] [blame] | 657 | // Finally, replace LHS with the result, consume the RHS, and iterate. |
| 658 | ++CurToken; |
| 659 | Tok = Result; |
Chris Lattner | 8d89643 | 2008-03-09 02:07:49 +0000 | [diff] [blame] | 660 | } while (!isAtEnd() && Tokens[CurToken].is(tok::hashhash)); |
Mike Stump | 1eb4433 | 2009-09-09 15:08:12 +0000 | [diff] [blame] | 661 | |
Abramo Bagnara | a08529c | 2011-10-03 18:39:03 +0000 | [diff] [blame] | 662 | SourceLocation EndLoc = Tokens[CurToken - 1].getLocation(); |
| 663 | |
Argyrios Kyrtzidis | b73377e | 2011-07-07 03:40:34 +0000 | [diff] [blame] | 664 | // The token's current location indicate where the token was lexed from. We |
| 665 | // need this information to compute the spelling of the token, but any |
| 666 | // diagnostics for the expanded token should appear as if the token was |
Abramo Bagnara | a08529c | 2011-10-03 18:39:03 +0000 | [diff] [blame] | 667 | // expanded from the full ## expression. Pull this information together into |
Argyrios Kyrtzidis | b73377e | 2011-07-07 03:40:34 +0000 | [diff] [blame] | 668 | // a new SourceLocation that captures all of this. |
Argyrios Kyrtzidis | c0069d8 | 2011-08-23 21:02:32 +0000 | [diff] [blame] | 669 | SourceManager &SM = PP.getSourceManager(); |
Abramo Bagnara | a08529c | 2011-10-03 18:39:03 +0000 | [diff] [blame] | 670 | if (StartLoc.isFileID()) |
| 671 | StartLoc = getExpansionLocForMacroDefLoc(StartLoc); |
| 672 | if (EndLoc.isFileID()) |
| 673 | EndLoc = getExpansionLocForMacroDefLoc(EndLoc); |
Eli Friedman | 19c8ce0 | 2012-12-01 01:15:54 +0000 | [diff] [blame] | 674 | FileID MacroFID = SM.getFileID(MacroExpansionStart); |
| 675 | while (SM.getFileID(StartLoc) != MacroFID) |
| 676 | StartLoc = SM.getImmediateExpansionRange(StartLoc).first; |
| 677 | while (SM.getFileID(EndLoc) != MacroFID) |
| 678 | EndLoc = SM.getImmediateExpansionRange(EndLoc).second; |
| 679 | |
Abramo Bagnara | a08529c | 2011-10-03 18:39:03 +0000 | [diff] [blame] | 680 | Tok.setLocation(SM.createExpansionLoc(Tok.getLocation(), StartLoc, EndLoc, |
Argyrios Kyrtzidis | c0069d8 | 2011-08-23 21:02:32 +0000 | [diff] [blame] | 681 | Tok.getLength())); |
Argyrios Kyrtzidis | b73377e | 2011-07-07 03:40:34 +0000 | [diff] [blame] | 682 | |
Reid Spencer | 5f016e2 | 2007-07-11 17:01:13 +0000 | [diff] [blame] | 683 | // Now that we got the result token, it will be subject to expansion. Since |
| 684 | // token pasting re-lexes the result token in raw mode, identifier information |
| 685 | // 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] | 686 | if (Tok.is(tok::raw_identifier)) { |
Reid Spencer | 5f016e2 | 2007-07-11 17:01:13 +0000 | [diff] [blame] | 687 | // Look up the identifier info for the token. We disabled identifier lookup |
| 688 | // 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] | 689 | PP.LookUpIdentifierInfo(Tok); |
Reid Spencer | 5f016e2 | 2007-07-11 17:01:13 +0000 | [diff] [blame] | 690 | } |
Chris Lattner | 3f1cc83 | 2008-02-07 06:03:59 +0000 | [diff] [blame] | 691 | return false; |
Reid Spencer | 5f016e2 | 2007-07-11 17:01:13 +0000 | [diff] [blame] | 692 | } |
| 693 | |
| 694 | /// isNextTokenLParen - If the next token lexed will pop this macro off the |
| 695 | /// expansion stack, return 2. If the next unexpanded token is a '(', return |
| 696 | /// 1, otherwise return 0. |
Chris Lattner | 1543e9c | 2008-03-09 02:18:51 +0000 | [diff] [blame] | 697 | unsigned TokenLexer::isNextTokenLParen() const { |
Reid Spencer | 5f016e2 | 2007-07-11 17:01:13 +0000 | [diff] [blame] | 698 | // Out of tokens? |
| 699 | if (isAtEnd()) |
| 700 | return 2; |
Chris Lattner | 8d89643 | 2008-03-09 02:07:49 +0000 | [diff] [blame] | 701 | return Tokens[CurToken].is(tok::l_paren); |
Reid Spencer | 5f016e2 | 2007-07-11 17:01:13 +0000 | [diff] [blame] | 702 | } |
Chris Lattner | 3f1cc83 | 2008-02-07 06:03:59 +0000 | [diff] [blame] | 703 | |
Peter Collingbourne | b2eb53d | 2011-02-22 13:49:00 +0000 | [diff] [blame] | 704 | /// isParsingPreprocessorDirective - Return true if we are in the middle of a |
| 705 | /// preprocessor directive. |
| 706 | bool TokenLexer::isParsingPreprocessorDirective() const { |
Peter Collingbourne | 8402155 | 2011-02-28 02:37:51 +0000 | [diff] [blame] | 707 | return Tokens[NumTokens-1].is(tok::eod) && !isAtEnd(); |
Peter Collingbourne | b2eb53d | 2011-02-22 13:49:00 +0000 | [diff] [blame] | 708 | } |
Chris Lattner | 3f1cc83 | 2008-02-07 06:03:59 +0000 | [diff] [blame] | 709 | |
| 710 | /// HandleMicrosoftCommentPaste - In microsoft compatibility mode, /##/ pastes |
| 711 | /// together to form a comment that comments out everything in the current |
| 712 | /// macro, other active macros, and anything left on the current physical |
Chandler Carruth | d0f5e44 | 2011-07-14 08:20:34 +0000 | [diff] [blame] | 713 | /// source line of the expanded buffer. Handle this by returning the |
Chris Lattner | 3f1cc83 | 2008-02-07 06:03:59 +0000 | [diff] [blame] | 714 | /// first token on the next line. |
Chris Lattner | 1543e9c | 2008-03-09 02:18:51 +0000 | [diff] [blame] | 715 | void TokenLexer::HandleMicrosoftCommentPaste(Token &Tok) { |
Chris Lattner | 3f1cc83 | 2008-02-07 06:03:59 +0000 | [diff] [blame] | 716 | // We 'comment out' the rest of this macro by just ignoring the rest of the |
| 717 | // tokens that have not been lexed yet, if any. |
Mike Stump | 1eb4433 | 2009-09-09 15:08:12 +0000 | [diff] [blame] | 718 | |
Chris Lattner | 3f1cc83 | 2008-02-07 06:03:59 +0000 | [diff] [blame] | 719 | // Since this must be a macro, mark the macro enabled now that it is no longer |
| 720 | // being expanded. |
| 721 | assert(Macro && "Token streams can't paste comments"); |
| 722 | Macro->EnableMacro(); |
Mike Stump | 1eb4433 | 2009-09-09 15:08:12 +0000 | [diff] [blame] | 723 | |
Chris Lattner | 3f1cc83 | 2008-02-07 06:03:59 +0000 | [diff] [blame] | 724 | PP.HandleMicrosoftCommentPaste(Tok); |
| 725 | } |
Argyrios Kyrtzidis | b73377e | 2011-07-07 03:40:34 +0000 | [diff] [blame] | 726 | |
Argyrios Kyrtzidis | ec3b538 | 2011-08-19 22:34:14 +0000 | [diff] [blame] | 727 | /// \brief If \arg loc is a file ID and points inside the current macro |
Argyrios Kyrtzidis | b73377e | 2011-07-07 03:40:34 +0000 | [diff] [blame] | 728 | /// definition, returns the appropriate source location pointing at the |
Argyrios Kyrtzidis | ec3b538 | 2011-08-19 22:34:14 +0000 | [diff] [blame] | 729 | /// macro expansion source location entry, otherwise it returns an invalid |
| 730 | /// SourceLocation. |
| 731 | SourceLocation |
| 732 | TokenLexer::getExpansionLocForMacroDefLoc(SourceLocation loc) const { |
Chandler Carruth | d0f5e44 | 2011-07-14 08:20:34 +0000 | [diff] [blame] | 733 | assert(ExpandLocStart.isValid() && MacroExpansionStart.isValid() && |
Argyrios Kyrtzidis | b73377e | 2011-07-07 03:40:34 +0000 | [diff] [blame] | 734 | "Not appropriate for token streams"); |
Argyrios Kyrtzidis | c0069d8 | 2011-08-23 21:02:32 +0000 | [diff] [blame] | 735 | assert(loc.isValid() && loc.isFileID()); |
Argyrios Kyrtzidis | b73377e | 2011-07-07 03:40:34 +0000 | [diff] [blame] | 736 | |
| 737 | SourceManager &SM = PP.getSourceManager(); |
Argyrios Kyrtzidis | 499ea55 | 2011-08-23 21:02:38 +0000 | [diff] [blame] | 738 | assert(SM.isInSLocAddrSpace(loc, MacroDefStart, MacroDefLength) && |
| 739 | "Expected loc to come from the macro definition"); |
Argyrios Kyrtzidis | b73377e | 2011-07-07 03:40:34 +0000 | [diff] [blame] | 740 | |
Argyrios Kyrtzidis | 499ea55 | 2011-08-23 21:02:38 +0000 | [diff] [blame] | 741 | unsigned relativeOffset = 0; |
| 742 | SM.isInSLocAddrSpace(loc, MacroDefStart, MacroDefLength, &relativeOffset); |
Argyrios Kyrtzidis | a64ccef | 2011-09-19 20:40:19 +0000 | [diff] [blame] | 743 | return MacroExpansionStart.getLocWithOffset(relativeOffset); |
Argyrios Kyrtzidis | b73377e | 2011-07-07 03:40:34 +0000 | [diff] [blame] | 744 | } |
Argyrios Kyrtzidis | d60a34a | 2011-08-19 22:34:17 +0000 | [diff] [blame] | 745 | |
| 746 | /// \brief Finds the tokens that are consecutive (from the same FileID) |
| 747 | /// creates a single SLocEntry, and assigns SourceLocations to each token that |
| 748 | /// point to that SLocEntry. e.g for |
| 749 | /// assert(foo == bar); |
| 750 | /// There will be a single SLocEntry for the "foo == bar" chunk and locations |
| 751 | /// for the 'foo', '==', 'bar' tokens will point inside that chunk. |
| 752 | /// |
| 753 | /// \arg begin_tokens will be updated to a position past all the found |
| 754 | /// consecutive tokens. |
| 755 | static void updateConsecutiveMacroArgTokens(SourceManager &SM, |
| 756 | SourceLocation InstLoc, |
| 757 | Token *&begin_tokens, |
| 758 | Token * end_tokens) { |
| 759 | assert(begin_tokens < end_tokens); |
Argyrios Kyrtzidis | d60a34a | 2011-08-19 22:34:17 +0000 | [diff] [blame] | 760 | |
Argyrios Kyrtzidis | b6c465e | 2011-08-23 21:02:41 +0000 | [diff] [blame] | 761 | SourceLocation FirstLoc = begin_tokens->getLocation(); |
| 762 | SourceLocation CurLoc = FirstLoc; |
| 763 | |
| 764 | // Compare the source location offset of tokens and group together tokens that |
| 765 | // are close, even if their locations point to different FileIDs. e.g. |
| 766 | // |
| 767 | // |bar | foo | cake | (3 tokens from 3 consecutive FileIDs) |
| 768 | // ^ ^ |
| 769 | // |bar foo cake| (one SLocEntry chunk for all tokens) |
| 770 | // |
| 771 | // we can perform this "merge" since the token's spelling location depends |
| 772 | // on the relative offset. |
| 773 | |
| 774 | Token *NextTok = begin_tokens + 1; |
| 775 | for (; NextTok < end_tokens; ++NextTok) { |
Argyrios Kyrtzidis | b29740a | 2012-12-19 23:55:44 +0000 | [diff] [blame] | 776 | SourceLocation NextLoc = NextTok->getLocation(); |
| 777 | if (CurLoc.isFileID() != NextLoc.isFileID()) |
| 778 | break; // Token from different kind of FileID. |
| 779 | |
Argyrios Kyrtzidis | b6c465e | 2011-08-23 21:02:41 +0000 | [diff] [blame] | 780 | int RelOffs; |
Argyrios Kyrtzidis | b29740a | 2012-12-19 23:55:44 +0000 | [diff] [blame] | 781 | if (!SM.isInSameSLocAddrSpace(CurLoc, NextLoc, &RelOffs)) |
Argyrios Kyrtzidis | b6c465e | 2011-08-23 21:02:41 +0000 | [diff] [blame] | 782 | break; // Token from different local/loaded location. |
| 783 | // Check that token is not before the previous token or more than 50 |
| 784 | // "characters" away. |
| 785 | if (RelOffs < 0 || RelOffs > 50) |
Argyrios Kyrtzidis | d60a34a | 2011-08-19 22:34:17 +0000 | [diff] [blame] | 786 | break; |
Argyrios Kyrtzidis | b29740a | 2012-12-19 23:55:44 +0000 | [diff] [blame] | 787 | CurLoc = NextLoc; |
Argyrios Kyrtzidis | b6c465e | 2011-08-23 21:02:41 +0000 | [diff] [blame] | 788 | } |
Argyrios Kyrtzidis | d60a34a | 2011-08-19 22:34:17 +0000 | [diff] [blame] | 789 | |
| 790 | // For the consecutive tokens, find the length of the SLocEntry to contain |
| 791 | // all of them. |
Argyrios Kyrtzidis | b6c465e | 2011-08-23 21:02:41 +0000 | [diff] [blame] | 792 | Token &LastConsecutiveTok = *(NextTok-1); |
Argyrios Kyrtzidis | c4b5582 | 2011-08-24 20:33:05 +0000 | [diff] [blame] | 793 | int LastRelOffs = 0; |
Argyrios Kyrtzidis | b6c465e | 2011-08-23 21:02:41 +0000 | [diff] [blame] | 794 | SM.isInSameSLocAddrSpace(FirstLoc, LastConsecutiveTok.getLocation(), |
| 795 | &LastRelOffs); |
| 796 | unsigned FullLength = LastRelOffs + LastConsecutiveTok.getLength(); |
Argyrios Kyrtzidis | d60a34a | 2011-08-19 22:34:17 +0000 | [diff] [blame] | 797 | |
| 798 | // Create a macro expansion SLocEntry that will "contain" all of the tokens. |
| 799 | SourceLocation Expansion = |
Argyrios Kyrtzidis | b6c465e | 2011-08-23 21:02:41 +0000 | [diff] [blame] | 800 | SM.createMacroArgExpansionLoc(FirstLoc, InstLoc,FullLength); |
Argyrios Kyrtzidis | d60a34a | 2011-08-19 22:34:17 +0000 | [diff] [blame] | 801 | |
| 802 | // Change the location of the tokens from the spelling location to the new |
| 803 | // expanded location. |
Argyrios Kyrtzidis | b6c465e | 2011-08-23 21:02:41 +0000 | [diff] [blame] | 804 | for (; begin_tokens < NextTok; ++begin_tokens) { |
Argyrios Kyrtzidis | d60a34a | 2011-08-19 22:34:17 +0000 | [diff] [blame] | 805 | Token &Tok = *begin_tokens; |
Argyrios Kyrtzidis | c4b5582 | 2011-08-24 20:33:05 +0000 | [diff] [blame] | 806 | int RelOffs = 0; |
Argyrios Kyrtzidis | b6c465e | 2011-08-23 21:02:41 +0000 | [diff] [blame] | 807 | SM.isInSameSLocAddrSpace(FirstLoc, Tok.getLocation(), &RelOffs); |
Argyrios Kyrtzidis | a64ccef | 2011-09-19 20:40:19 +0000 | [diff] [blame] | 808 | Tok.setLocation(Expansion.getLocWithOffset(RelOffs)); |
Argyrios Kyrtzidis | d60a34a | 2011-08-19 22:34:17 +0000 | [diff] [blame] | 809 | } |
| 810 | } |
| 811 | |
| 812 | /// \brief Creates SLocEntries and updates the locations of macro argument |
| 813 | /// tokens to their new expanded locations. |
| 814 | /// |
| 815 | /// \param ArgIdDefLoc the location of the macro argument id inside the macro |
| 816 | /// definition. |
| 817 | /// \param Tokens the macro argument tokens to update. |
| 818 | void TokenLexer::updateLocForMacroArgTokens(SourceLocation ArgIdSpellLoc, |
| 819 | Token *begin_tokens, |
| 820 | Token *end_tokens) { |
| 821 | SourceManager &SM = PP.getSourceManager(); |
| 822 | |
Argyrios Kyrtzidis | b6c465e | 2011-08-23 21:02:41 +0000 | [diff] [blame] | 823 | SourceLocation InstLoc = |
Argyrios Kyrtzidis | d60a34a | 2011-08-19 22:34:17 +0000 | [diff] [blame] | 824 | getExpansionLocForMacroDefLoc(ArgIdSpellLoc); |
Argyrios Kyrtzidis | d60a34a | 2011-08-19 22:34:17 +0000 | [diff] [blame] | 825 | |
Argyrios Kyrtzidis | b6c465e | 2011-08-23 21:02:41 +0000 | [diff] [blame] | 826 | while (begin_tokens < end_tokens) { |
| 827 | // If there's only one token just create a SLocEntry for it. |
| 828 | if (end_tokens - begin_tokens == 1) { |
| 829 | Token &Tok = *begin_tokens; |
| 830 | Tok.setLocation(SM.createMacroArgExpansionLoc(Tok.getLocation(), |
| 831 | InstLoc, |
| 832 | Tok.getLength())); |
| 833 | return; |
| 834 | } |
| 835 | |
| 836 | updateConsecutiveMacroArgTokens(SM, InstLoc, begin_tokens, end_tokens); |
| 837 | } |
Argyrios Kyrtzidis | d60a34a | 2011-08-19 22:34:17 +0000 | [diff] [blame] | 838 | } |
Eli Friedman | d2f9308 | 2013-09-19 00:41:32 +0000 | [diff] [blame] | 839 | |
| 840 | void TokenLexer::PropagateLineStartLeadingSpaceInfo(Token &Result) { |
| 841 | AtStartOfLine = Result.isAtStartOfLine(); |
| 842 | HasLeadingSpace = Result.hasLeadingSpace(); |
| 843 | } |