Reid Spencer | 5f016e2 | 2007-07-11 17:01:13 +0000 | [diff] [blame] | 1 | //===--- PrintPreprocessedOutput.cpp - Implement the -E mode --------------===// |
| 2 | // |
| 3 | // The LLVM Compiler Infrastructure |
| 4 | // |
| 5 | // This file was developed by Chris Lattner and is distributed under |
| 6 | // the University of Illinois Open Source License. See LICENSE.TXT for details. |
| 7 | // |
| 8 | //===----------------------------------------------------------------------===// |
| 9 | // |
| 10 | // This code simply runs the preprocessor on the input file and prints out the |
| 11 | // result. This is the traditional behavior of the -E option. |
| 12 | // |
| 13 | //===----------------------------------------------------------------------===// |
| 14 | |
| 15 | #include "clang.h" |
| 16 | #include "clang/Lex/PPCallbacks.h" |
| 17 | #include "clang/Lex/Preprocessor.h" |
| 18 | #include "clang/Lex/Pragma.h" |
| 19 | #include "clang/Basic/SourceManager.h" |
| 20 | #include "llvm/Support/CommandLine.h" |
| 21 | #include "llvm/ADT/StringExtras.h" |
| 22 | #include "llvm/Config/config.h" |
| 23 | #include <cstdio> |
| 24 | using namespace clang; |
| 25 | |
| 26 | //===----------------------------------------------------------------------===// |
| 27 | // Simple buffered I/O |
| 28 | //===----------------------------------------------------------------------===// |
| 29 | // |
| 30 | // Empirically, iostream is over 30% slower than stdio for this workload, and |
| 31 | // stdio itself isn't very well suited. The problem with stdio is use of |
| 32 | // putchar_unlocked. We have many newline characters that need to be emitted, |
| 33 | // but stdio needs to do extra checks to handle line buffering mode. These |
| 34 | // extra checks make putchar_unlocked fall off its inlined code path, hitting |
| 35 | // slow system code. In practice, using 'write' directly makes 'clang -E -P' |
| 36 | // about 10% faster than using the stdio path on darwin. |
| 37 | |
| 38 | #ifdef HAVE_UNISTD_H |
| 39 | #include <unistd.h> |
| 40 | #else |
| 41 | #define USE_STDIO 1 |
| 42 | #endif |
| 43 | |
| 44 | static char *OutBufStart = 0, *OutBufEnd, *OutBufCur; |
| 45 | |
| 46 | /// InitOutputBuffer - Initialize our output buffer. |
| 47 | /// |
| 48 | static void InitOutputBuffer() { |
| 49 | #ifndef USE_STDIO |
| 50 | OutBufStart = new char[64*1024]; |
| 51 | OutBufEnd = OutBufStart+64*1024; |
| 52 | OutBufCur = OutBufStart; |
| 53 | #endif |
| 54 | } |
| 55 | |
| 56 | /// FlushBuffer - Write the accumulated bytes to the output stream. |
| 57 | /// |
| 58 | static void FlushBuffer() { |
| 59 | #ifndef USE_STDIO |
| 60 | write(STDOUT_FILENO, OutBufStart, OutBufCur-OutBufStart); |
| 61 | OutBufCur = OutBufStart; |
| 62 | #endif |
| 63 | } |
| 64 | |
| 65 | /// CleanupOutputBuffer - Finish up output. |
| 66 | /// |
| 67 | static void CleanupOutputBuffer() { |
| 68 | #ifndef USE_STDIO |
| 69 | FlushBuffer(); |
| 70 | delete [] OutBufStart; |
| 71 | #endif |
| 72 | } |
| 73 | |
| 74 | static void OutputChar(char c) { |
| 75 | #ifdef USE_STDIO |
| 76 | putchar_unlocked(c); |
| 77 | #else |
| 78 | if (OutBufCur >= OutBufEnd) |
| 79 | FlushBuffer(); |
| 80 | *OutBufCur++ = c; |
| 81 | #endif |
| 82 | } |
| 83 | |
| 84 | static void OutputString(const char *Ptr, unsigned Size) { |
| 85 | #ifdef USE_STDIO |
| 86 | fwrite(Ptr, Size, 1, stdout); |
| 87 | #else |
| 88 | if (OutBufCur+Size >= OutBufEnd) |
| 89 | FlushBuffer(); |
Chris Lattner | e225e37 | 2007-07-23 06:23:07 +0000 | [diff] [blame] | 90 | |
| 91 | switch (Size) { |
| 92 | default: |
| 93 | memcpy(OutBufCur, Ptr, Size); |
| 94 | break; |
| 95 | case 3: |
| 96 | OutBufCur[2] = Ptr[2]; |
| 97 | case 2: |
| 98 | OutBufCur[1] = Ptr[1]; |
| 99 | case 1: |
| 100 | OutBufCur[0] = Ptr[0]; |
| 101 | case 0: |
| 102 | break; |
| 103 | } |
Reid Spencer | 5f016e2 | 2007-07-11 17:01:13 +0000 | [diff] [blame] | 104 | OutBufCur += Size; |
| 105 | #endif |
| 106 | } |
| 107 | |
| 108 | |
| 109 | //===----------------------------------------------------------------------===// |
| 110 | // Preprocessed token printer |
| 111 | //===----------------------------------------------------------------------===// |
| 112 | |
| 113 | static llvm::cl::opt<bool> |
| 114 | DisableLineMarkers("P", llvm::cl::desc("Disable linemarker output in -E mode")); |
| 115 | static llvm::cl::opt<bool> |
| 116 | EnableCommentOutput("C", llvm::cl::desc("Enable comment output in -E mode")); |
| 117 | static llvm::cl::opt<bool> |
| 118 | EnableMacroCommentOutput("CC", |
| 119 | llvm::cl::desc("Enable comment output in -E mode, " |
| 120 | "even from macro expansions")); |
| 121 | |
| 122 | namespace { |
| 123 | class PrintPPOutputPPCallbacks : public PPCallbacks { |
| 124 | Preprocessor &PP; |
| 125 | unsigned CurLine; |
| 126 | std::string CurFilename; |
| 127 | bool EmittedTokensOnThisLine; |
| 128 | DirectoryLookup::DirType FileType; |
| 129 | public: |
| 130 | PrintPPOutputPPCallbacks(Preprocessor &pp) : PP(pp) { |
| 131 | CurLine = 0; |
Chris Lattner | 0cbc4b5 | 2007-07-22 06:38:50 +0000 | [diff] [blame] | 132 | CurFilename = "<uninit>"; |
Reid Spencer | 5f016e2 | 2007-07-11 17:01:13 +0000 | [diff] [blame] | 133 | EmittedTokensOnThisLine = false; |
| 134 | FileType = DirectoryLookup::NormalHeaderDir; |
| 135 | } |
| 136 | |
| 137 | void SetEmittedTokensOnThisLine() { EmittedTokensOnThisLine = true; } |
Chris Lattner | f0f2b29 | 2007-07-23 06:09:34 +0000 | [diff] [blame] | 138 | bool hasEmittedTokensOnThisLine() const { return EmittedTokensOnThisLine; } |
Reid Spencer | 5f016e2 | 2007-07-11 17:01:13 +0000 | [diff] [blame] | 139 | |
| 140 | virtual void FileChanged(SourceLocation Loc, FileChangeReason Reason, |
| 141 | DirectoryLookup::DirType FileType); |
| 142 | virtual void Ident(SourceLocation Loc, const std::string &str); |
| 143 | |
| 144 | |
Chris Lattner | d217773 | 2007-07-20 16:59:19 +0000 | [diff] [blame] | 145 | void HandleFirstTokOnLine(Token &Tok); |
Reid Spencer | 5f016e2 | 2007-07-11 17:01:13 +0000 | [diff] [blame] | 146 | void MoveToLine(SourceLocation Loc); |
Chris Lattner | d217773 | 2007-07-20 16:59:19 +0000 | [diff] [blame] | 147 | bool AvoidConcat(const Token &PrevTok, const Token &Tok); |
Reid Spencer | 5f016e2 | 2007-07-11 17:01:13 +0000 | [diff] [blame] | 148 | }; |
| 149 | } |
| 150 | |
Chris Lattner | f063721 | 2007-07-23 06:31:11 +0000 | [diff] [blame] | 151 | /// UToStr - Do itoa on the specified number, in-place in the specified buffer. |
| 152 | /// endptr points to the end of the buffer. |
| 153 | static char *UToStr(unsigned N, char *EndPtr) { |
| 154 | // Null terminate the buffer. |
| 155 | *--EndPtr = '\0'; |
| 156 | if (N == 0) // Zero is a special case. |
| 157 | *--EndPtr = '0'; |
| 158 | while (N) { |
| 159 | *--EndPtr = '0' + char(N % 10); |
| 160 | N /= 10; |
| 161 | } |
| 162 | return EndPtr; |
| 163 | } |
| 164 | |
| 165 | |
Reid Spencer | 5f016e2 | 2007-07-11 17:01:13 +0000 | [diff] [blame] | 166 | /// MoveToLine - Move the output to the source line specified by the location |
| 167 | /// object. We can do this by emitting some number of \n's, or be emitting a |
| 168 | /// #line directive. |
| 169 | void PrintPPOutputPPCallbacks::MoveToLine(SourceLocation Loc) { |
| 170 | if (DisableLineMarkers) { |
| 171 | if (EmittedTokensOnThisLine) { |
| 172 | OutputChar('\n'); |
| 173 | EmittedTokensOnThisLine = false; |
| 174 | } |
| 175 | return; |
| 176 | } |
| 177 | |
Chris Lattner | 9dc1f53 | 2007-07-20 16:37:10 +0000 | [diff] [blame] | 178 | unsigned LineNo = PP.getSourceManager().getLogicalLineNumber(Loc); |
Reid Spencer | 5f016e2 | 2007-07-11 17:01:13 +0000 | [diff] [blame] | 179 | |
| 180 | // If this line is "close enough" to the original line, just print newlines, |
| 181 | // otherwise print a #line directive. |
| 182 | if (LineNo-CurLine < 8) { |
Chris Lattner | 822f940 | 2007-07-23 05:14:05 +0000 | [diff] [blame] | 183 | if (LineNo-CurLine == 1) |
Reid Spencer | 5f016e2 | 2007-07-11 17:01:13 +0000 | [diff] [blame] | 184 | OutputChar('\n'); |
Chris Lattner | 822f940 | 2007-07-23 05:14:05 +0000 | [diff] [blame] | 185 | else { |
| 186 | const char *NewLines = "\n\n\n\n\n\n\n\n"; |
| 187 | OutputString(NewLines, LineNo-CurLine); |
| 188 | CurLine = LineNo; |
| 189 | } |
Reid Spencer | 5f016e2 | 2007-07-11 17:01:13 +0000 | [diff] [blame] | 190 | } else { |
| 191 | if (EmittedTokensOnThisLine) { |
| 192 | OutputChar('\n'); |
| 193 | EmittedTokensOnThisLine = false; |
| 194 | } |
| 195 | |
| 196 | CurLine = LineNo; |
| 197 | |
| 198 | OutputChar('#'); |
| 199 | OutputChar(' '); |
Chris Lattner | f063721 | 2007-07-23 06:31:11 +0000 | [diff] [blame] | 200 | char NumberBuffer[20]; |
| 201 | const char *NumStr = UToStr(LineNo, NumberBuffer+20); |
| 202 | OutputString(NumStr, (NumberBuffer+20)-NumStr-1); |
Reid Spencer | 5f016e2 | 2007-07-11 17:01:13 +0000 | [diff] [blame] | 203 | OutputChar(' '); |
Chris Lattner | 0cbc4b5 | 2007-07-22 06:38:50 +0000 | [diff] [blame] | 204 | OutputChar('"'); |
Reid Spencer | 5f016e2 | 2007-07-11 17:01:13 +0000 | [diff] [blame] | 205 | OutputString(&CurFilename[0], CurFilename.size()); |
Chris Lattner | 0cbc4b5 | 2007-07-22 06:38:50 +0000 | [diff] [blame] | 206 | OutputChar('"'); |
Reid Spencer | 5f016e2 | 2007-07-11 17:01:13 +0000 | [diff] [blame] | 207 | |
| 208 | if (FileType == DirectoryLookup::SystemHeaderDir) |
| 209 | OutputString(" 3", 2); |
| 210 | else if (FileType == DirectoryLookup::ExternCSystemHeaderDir) |
| 211 | OutputString(" 3 4", 4); |
| 212 | OutputChar('\n'); |
| 213 | } |
| 214 | } |
| 215 | |
| 216 | |
| 217 | /// FileChanged - Whenever the preprocessor enters or exits a #include file |
| 218 | /// it invokes this handler. Update our conception of the current source |
| 219 | /// position. |
| 220 | void PrintPPOutputPPCallbacks::FileChanged(SourceLocation Loc, |
| 221 | FileChangeReason Reason, |
| 222 | DirectoryLookup::DirType FileType) { |
| 223 | if (DisableLineMarkers) return; |
| 224 | |
| 225 | // Unless we are exiting a #include, make sure to skip ahead to the line the |
| 226 | // #include directive was at. |
| 227 | SourceManager &SourceMgr = PP.getSourceManager(); |
| 228 | if (Reason == PPCallbacks::EnterFile) { |
Chris Lattner | 9dc1f53 | 2007-07-20 16:37:10 +0000 | [diff] [blame] | 229 | MoveToLine(SourceMgr.getIncludeLoc(Loc)); |
Reid Spencer | 5f016e2 | 2007-07-11 17:01:13 +0000 | [diff] [blame] | 230 | } else if (Reason == PPCallbacks::SystemHeaderPragma) { |
| 231 | MoveToLine(Loc); |
| 232 | |
| 233 | // TODO GCC emits the # directive for this directive on the line AFTER the |
| 234 | // directive and emits a bunch of spaces that aren't needed. Emulate this |
| 235 | // strange behavior. |
| 236 | } |
| 237 | |
Chris Lattner | 9dc1f53 | 2007-07-20 16:37:10 +0000 | [diff] [blame] | 238 | Loc = SourceMgr.getLogicalLoc(Loc); |
Reid Spencer | 5f016e2 | 2007-07-11 17:01:13 +0000 | [diff] [blame] | 239 | CurLine = SourceMgr.getLineNumber(Loc); |
Chris Lattner | 0cbc4b5 | 2007-07-22 06:38:50 +0000 | [diff] [blame] | 240 | CurFilename = Lexer::Stringify(SourceMgr.getSourceName(Loc)); |
Reid Spencer | 5f016e2 | 2007-07-11 17:01:13 +0000 | [diff] [blame] | 241 | FileType = FileType; |
| 242 | |
| 243 | if (EmittedTokensOnThisLine) { |
| 244 | OutputChar('\n'); |
| 245 | EmittedTokensOnThisLine = false; |
| 246 | } |
| 247 | |
Reid Spencer | 5f016e2 | 2007-07-11 17:01:13 +0000 | [diff] [blame] | 248 | OutputChar('#'); |
| 249 | OutputChar(' '); |
| 250 | std::string Num = llvm::utostr_32(CurLine); |
| 251 | OutputString(&Num[0], Num.size()); |
| 252 | OutputChar(' '); |
Chris Lattner | 0cbc4b5 | 2007-07-22 06:38:50 +0000 | [diff] [blame] | 253 | OutputChar('"'); |
Reid Spencer | 5f016e2 | 2007-07-11 17:01:13 +0000 | [diff] [blame] | 254 | OutputString(&CurFilename[0], CurFilename.size()); |
Chris Lattner | 0cbc4b5 | 2007-07-22 06:38:50 +0000 | [diff] [blame] | 255 | OutputChar('"'); |
Reid Spencer | 5f016e2 | 2007-07-11 17:01:13 +0000 | [diff] [blame] | 256 | |
| 257 | switch (Reason) { |
| 258 | case PPCallbacks::EnterFile: |
| 259 | OutputString(" 1", 2); |
| 260 | break; |
| 261 | case PPCallbacks::ExitFile: |
| 262 | OutputString(" 2", 2); |
| 263 | break; |
| 264 | case PPCallbacks::SystemHeaderPragma: break; |
| 265 | case PPCallbacks::RenameFile: break; |
| 266 | } |
| 267 | |
| 268 | if (FileType == DirectoryLookup::SystemHeaderDir) |
| 269 | OutputString(" 3", 2); |
| 270 | else if (FileType == DirectoryLookup::ExternCSystemHeaderDir) |
| 271 | OutputString(" 3 4", 4); |
| 272 | |
| 273 | OutputChar('\n'); |
| 274 | } |
| 275 | |
| 276 | /// HandleIdent - Handle #ident directives when read by the preprocessor. |
| 277 | /// |
| 278 | void PrintPPOutputPPCallbacks::Ident(SourceLocation Loc, const std::string &S) { |
| 279 | MoveToLine(Loc); |
| 280 | |
| 281 | OutputString("#ident ", strlen("#ident ")); |
| 282 | OutputString(&S[0], S.size()); |
| 283 | EmittedTokensOnThisLine = true; |
| 284 | } |
| 285 | |
| 286 | /// HandleFirstTokOnLine - When emitting a preprocessed file in -E mode, this |
| 287 | /// is called for the first token on each new line. |
Chris Lattner | d217773 | 2007-07-20 16:59:19 +0000 | [diff] [blame] | 288 | void PrintPPOutputPPCallbacks::HandleFirstTokOnLine(Token &Tok) { |
Reid Spencer | 5f016e2 | 2007-07-11 17:01:13 +0000 | [diff] [blame] | 289 | // Figure out what line we went to and insert the appropriate number of |
| 290 | // newline characters. |
| 291 | MoveToLine(Tok.getLocation()); |
| 292 | |
| 293 | // Print out space characters so that the first token on a line is |
| 294 | // indented for easy reading. |
Chris Lattner | 9dc1f53 | 2007-07-20 16:37:10 +0000 | [diff] [blame] | 295 | const SourceManager &SourceMgr = PP.getSourceManager(); |
| 296 | unsigned ColNo = SourceMgr.getLogicalColumnNumber(Tok.getLocation()); |
Reid Spencer | 5f016e2 | 2007-07-11 17:01:13 +0000 | [diff] [blame] | 297 | |
| 298 | // This hack prevents stuff like: |
| 299 | // #define HASH # |
| 300 | // HASH define foo bar |
| 301 | // From having the # character end up at column 1, which makes it so it |
| 302 | // is not handled as a #define next time through the preprocessor if in |
| 303 | // -fpreprocessed mode. |
| 304 | if (ColNo <= 1 && Tok.getKind() == tok::hash) |
| 305 | OutputChar(' '); |
| 306 | |
| 307 | // Otherwise, indent the appropriate number of spaces. |
| 308 | for (; ColNo > 1; --ColNo) |
| 309 | OutputChar(' '); |
| 310 | } |
| 311 | |
| 312 | namespace { |
| 313 | struct UnknownPragmaHandler : public PragmaHandler { |
| 314 | const char *Prefix; |
| 315 | PrintPPOutputPPCallbacks *Callbacks; |
| 316 | |
| 317 | UnknownPragmaHandler(const char *prefix, PrintPPOutputPPCallbacks *callbacks) |
| 318 | : PragmaHandler(0), Prefix(prefix), Callbacks(callbacks) {} |
Chris Lattner | d217773 | 2007-07-20 16:59:19 +0000 | [diff] [blame] | 319 | virtual void HandlePragma(Preprocessor &PP, Token &PragmaTok) { |
Reid Spencer | 5f016e2 | 2007-07-11 17:01:13 +0000 | [diff] [blame] | 320 | // Figure out what line we went to and insert the appropriate number of |
| 321 | // newline characters. |
| 322 | Callbacks->MoveToLine(PragmaTok.getLocation()); |
| 323 | OutputString(Prefix, strlen(Prefix)); |
| 324 | |
| 325 | // Read and print all of the pragma tokens. |
| 326 | while (PragmaTok.getKind() != tok::eom) { |
| 327 | if (PragmaTok.hasLeadingSpace()) |
| 328 | OutputChar(' '); |
| 329 | std::string TokSpell = PP.getSpelling(PragmaTok); |
| 330 | OutputString(&TokSpell[0], TokSpell.size()); |
| 331 | PP.LexUnexpandedToken(PragmaTok); |
| 332 | } |
| 333 | OutputChar('\n'); |
| 334 | } |
| 335 | }; |
| 336 | } // end anonymous namespace |
| 337 | |
Chris Lattner | f0f2b29 | 2007-07-23 06:09:34 +0000 | [diff] [blame] | 338 | |
| 339 | enum AvoidConcatInfo { |
| 340 | /// By default, a token never needs to avoid concatenation. Most tokens (e.g. |
| 341 | /// ',', ')', etc) don't cause a problem when concatenated. |
| 342 | aci_never_avoid_concat = 0, |
| 343 | |
| 344 | /// aci_custom_firstchar - AvoidConcat contains custom code to handle this |
| 345 | /// token's requirements, and it needs to know the first character of the |
| 346 | /// token. |
| 347 | aci_custom_firstchar = 1, |
| 348 | |
| 349 | /// aci_custom - AvoidConcat contains custom code to handle this token's |
| 350 | /// requirements, but it doesn't need to know the first character of the |
| 351 | /// token. |
| 352 | aci_custom = 2, |
| 353 | |
| 354 | /// aci_avoid_equal - Many tokens cannot be safely followed by an '=' |
| 355 | /// character. For example, "<<" turns into "<<=" when followed by an =. |
| 356 | aci_avoid_equal = 4 |
| 357 | }; |
| 358 | |
| 359 | /// This array contains information for each token on what action to take when |
| 360 | /// avoiding concatenation of tokens in the AvoidConcat method. |
| 361 | static char TokenInfo[tok::NUM_TOKENS]; |
| 362 | |
| 363 | /// InitAvoidConcatTokenInfo - Tokens that must avoid concatenation should be |
| 364 | /// marked by this function. |
| 365 | static void InitAvoidConcatTokenInfo() { |
| 366 | // These tokens have custom code in AvoidConcat. |
| 367 | TokenInfo[tok::identifier ] |= aci_custom; |
| 368 | TokenInfo[tok::numeric_constant] |= aci_custom_firstchar; |
| 369 | TokenInfo[tok::period ] |= aci_custom_firstchar; |
| 370 | TokenInfo[tok::amp ] |= aci_custom_firstchar; |
| 371 | TokenInfo[tok::plus ] |= aci_custom_firstchar; |
| 372 | TokenInfo[tok::minus ] |= aci_custom_firstchar; |
| 373 | TokenInfo[tok::slash ] |= aci_custom_firstchar; |
| 374 | TokenInfo[tok::less ] |= aci_custom_firstchar; |
| 375 | TokenInfo[tok::greater ] |= aci_custom_firstchar; |
| 376 | TokenInfo[tok::pipe ] |= aci_custom_firstchar; |
| 377 | TokenInfo[tok::percent ] |= aci_custom_firstchar; |
| 378 | TokenInfo[tok::colon ] |= aci_custom_firstchar; |
| 379 | TokenInfo[tok::hash ] |= aci_custom_firstchar; |
| 380 | TokenInfo[tok::arrow ] |= aci_custom_firstchar; |
| 381 | |
| 382 | // These tokens change behavior if followed by an '='. |
| 383 | TokenInfo[tok::amp ] |= aci_avoid_equal; // &= |
| 384 | TokenInfo[tok::plus ] |= aci_avoid_equal; // += |
| 385 | TokenInfo[tok::minus ] |= aci_avoid_equal; // -= |
| 386 | TokenInfo[tok::slash ] |= aci_avoid_equal; // /= |
| 387 | TokenInfo[tok::less ] |= aci_avoid_equal; // <= |
| 388 | TokenInfo[tok::greater ] |= aci_avoid_equal; // >= |
| 389 | TokenInfo[tok::pipe ] |= aci_avoid_equal; // |= |
| 390 | TokenInfo[tok::percent ] |= aci_avoid_equal; // %= |
| 391 | TokenInfo[tok::star ] |= aci_avoid_equal; // *= |
| 392 | TokenInfo[tok::exclaim ] |= aci_avoid_equal; // != |
| 393 | TokenInfo[tok::lessless ] |= aci_avoid_equal; // <<= |
| 394 | TokenInfo[tok::greaterequal] |= aci_avoid_equal; // >>= |
| 395 | TokenInfo[tok::caret ] |= aci_avoid_equal; // ^= |
| 396 | TokenInfo[tok::equal ] |= aci_avoid_equal; // == |
| 397 | } |
| 398 | |
Reid Spencer | 5f016e2 | 2007-07-11 17:01:13 +0000 | [diff] [blame] | 399 | /// AvoidConcat - If printing PrevTok immediately followed by Tok would cause |
| 400 | /// the two individual tokens to be lexed as a single token, return true (which |
| 401 | /// causes a space to be printed between them). This allows the output of -E |
| 402 | /// mode to be lexed to the same token stream as lexing the input directly |
| 403 | /// would. |
| 404 | /// |
| 405 | /// This code must conservatively return true if it doesn't want to be 100% |
| 406 | /// accurate. This will cause the output to include extra space characters, but |
| 407 | /// the resulting output won't have incorrect concatenations going on. Examples |
| 408 | /// include "..", which we print with a space between, because we don't want to |
| 409 | /// track enough to tell "x.." from "...". |
Chris Lattner | d217773 | 2007-07-20 16:59:19 +0000 | [diff] [blame] | 410 | bool PrintPPOutputPPCallbacks::AvoidConcat(const Token &PrevTok, |
| 411 | const Token &Tok) { |
Reid Spencer | 5f016e2 | 2007-07-11 17:01:13 +0000 | [diff] [blame] | 412 | char Buffer[256]; |
| 413 | |
Chris Lattner | f0f2b29 | 2007-07-23 06:09:34 +0000 | [diff] [blame] | 414 | tok::TokenKind PrevKind = PrevTok.getKind(); |
| 415 | if (PrevTok.getIdentifierInfo()) // Language keyword or named operator. |
| 416 | PrevKind = tok::identifier; |
| 417 | |
| 418 | // Look up information on when we should avoid concatenation with prevtok. |
| 419 | unsigned ConcatInfo = TokenInfo[PrevKind]; |
| 420 | |
| 421 | // If prevtok never causes a problem for anything after it, return quickly. |
| 422 | if (ConcatInfo == 0) return false; |
Reid Spencer | 5f016e2 | 2007-07-11 17:01:13 +0000 | [diff] [blame] | 423 | |
Chris Lattner | f0f2b29 | 2007-07-23 06:09:34 +0000 | [diff] [blame] | 424 | if (ConcatInfo & aci_avoid_equal) { |
| 425 | // If the next token is '=' or '==', avoid concatenation. |
| 426 | if (Tok.getKind() == tok::equal || |
| 427 | Tok.getKind() == tok::equalequal) |
| 428 | return true; |
Chris Lattner | b638a30 | 2007-07-23 23:21:34 +0000 | [diff] [blame^] | 429 | ConcatInfo &= ~aci_avoid_equal; |
Chris Lattner | f0f2b29 | 2007-07-23 06:09:34 +0000 | [diff] [blame] | 430 | } |
| 431 | |
| 432 | if (ConcatInfo == 0) return false; |
| 433 | |
| 434 | |
| 435 | |
Reid Spencer | 5f016e2 | 2007-07-11 17:01:13 +0000 | [diff] [blame] | 436 | // Basic algorithm: we look at the first character of the second token, and |
| 437 | // determine whether it, if appended to the first token, would form (or would |
| 438 | // contribute) to a larger token if concatenated. |
Chris Lattner | f0f2b29 | 2007-07-23 06:09:34 +0000 | [diff] [blame] | 439 | char FirstChar = 0; |
| 440 | if (ConcatInfo & aci_custom) { |
| 441 | // If the token does not need to know the first character, don't get it. |
| 442 | } else if (IdentifierInfo *II = Tok.getIdentifierInfo()) { |
Reid Spencer | 5f016e2 | 2007-07-11 17:01:13 +0000 | [diff] [blame] | 443 | // Avoid spelling identifiers, the most common form of token. |
| 444 | FirstChar = II->getName()[0]; |
Chris Lattner | b19f5e8 | 2007-07-23 05:18:42 +0000 | [diff] [blame] | 445 | } else if (!Tok.needsCleaning()) { |
| 446 | SourceManager &SrcMgr = PP.getSourceManager(); |
| 447 | FirstChar = |
| 448 | *SrcMgr.getCharacterData(SrcMgr.getPhysicalLoc(Tok.getLocation())); |
Reid Spencer | 5f016e2 | 2007-07-11 17:01:13 +0000 | [diff] [blame] | 449 | } else if (Tok.getLength() < 256) { |
| 450 | const char *TokPtr = Buffer; |
| 451 | PP.getSpelling(Tok, TokPtr); |
| 452 | FirstChar = TokPtr[0]; |
| 453 | } else { |
| 454 | FirstChar = PP.getSpelling(Tok)[0]; |
| 455 | } |
Chris Lattner | f0f2b29 | 2007-07-23 06:09:34 +0000 | [diff] [blame] | 456 | |
Reid Spencer | 5f016e2 | 2007-07-11 17:01:13 +0000 | [diff] [blame] | 457 | switch (PrevKind) { |
Chris Lattner | f0f2b29 | 2007-07-23 06:09:34 +0000 | [diff] [blame] | 458 | default: assert(0 && "InitAvoidConcatTokenInfo built wrong"); |
Reid Spencer | 5f016e2 | 2007-07-11 17:01:13 +0000 | [diff] [blame] | 459 | case tok::identifier: // id+id or id+number or id+L"foo". |
Chris Lattner | f0f2b29 | 2007-07-23 06:09:34 +0000 | [diff] [blame] | 460 | if (Tok.getKind() == tok::numeric_constant || Tok.getIdentifierInfo() || |
| 461 | Tok.getKind() == tok::wide_string_literal /* || |
| 462 | Tok.getKind() == tok::wide_char_literal*/) |
| 463 | return true; |
| 464 | if (Tok.getKind() != tok::char_constant) |
| 465 | return false; |
| 466 | |
| 467 | // FIXME: need a wide_char_constant! |
| 468 | if (!Tok.needsCleaning()) { |
| 469 | SourceManager &SrcMgr = PP.getSourceManager(); |
| 470 | return *SrcMgr.getCharacterData(SrcMgr.getPhysicalLoc(Tok.getLocation())) |
| 471 | == 'L'; |
| 472 | } else if (Tok.getLength() < 256) { |
| 473 | const char *TokPtr = Buffer; |
| 474 | PP.getSpelling(Tok, TokPtr); |
| 475 | return TokPtr[0] == 'L'; |
| 476 | } else { |
| 477 | return PP.getSpelling(Tok)[0] == 'L'; |
| 478 | } |
Reid Spencer | 5f016e2 | 2007-07-11 17:01:13 +0000 | [diff] [blame] | 479 | case tok::numeric_constant: |
| 480 | return isalnum(FirstChar) || Tok.getKind() == tok::numeric_constant || |
| 481 | FirstChar == '+' || FirstChar == '-' || FirstChar == '.'; |
| 482 | case tok::period: // ..., .*, .1234 |
| 483 | return FirstChar == '.' || FirstChar == '*' || isdigit(FirstChar); |
Chris Lattner | f0f2b29 | 2007-07-23 06:09:34 +0000 | [diff] [blame] | 484 | case tok::amp: // && |
| 485 | return FirstChar == '&'; |
| 486 | case tok::plus: // ++ |
| 487 | return FirstChar == '+'; |
| 488 | case tok::minus: // --, ->, ->* |
| 489 | return FirstChar == '-' || FirstChar == '>'; |
| 490 | case tok::slash: //, /*, // |
| 491 | return FirstChar == '*' || FirstChar == '/'; |
| 492 | case tok::less: // <<, <<=, <:, <% |
| 493 | return FirstChar == '<' || FirstChar == ':' || FirstChar == '%'; |
| 494 | case tok::greater: // >>, >>= |
| 495 | return FirstChar == '>'; |
| 496 | case tok::pipe: // || |
| 497 | return FirstChar == '|'; |
| 498 | case tok::percent: // %>, %: |
| 499 | return FirstChar == '>' || FirstChar == ':'; |
Reid Spencer | 5f016e2 | 2007-07-11 17:01:13 +0000 | [diff] [blame] | 500 | case tok::colon: // ::, :> |
| 501 | return FirstChar == ':' || FirstChar == '>'; |
| 502 | case tok::hash: // ##, #@, %:%: |
| 503 | return FirstChar == '#' || FirstChar == '@' || FirstChar == '%'; |
| 504 | case tok::arrow: // ->* |
| 505 | return FirstChar == '*'; |
Reid Spencer | 5f016e2 | 2007-07-11 17:01:13 +0000 | [diff] [blame] | 506 | } |
| 507 | } |
| 508 | |
| 509 | /// DoPrintPreprocessedInput - This implements -E mode. |
| 510 | /// |
| 511 | void clang::DoPrintPreprocessedInput(unsigned MainFileID, Preprocessor &PP, |
| 512 | const LangOptions &Options) { |
| 513 | // Inform the preprocessor whether we want it to retain comments or not, due |
| 514 | // to -C or -CC. |
| 515 | PP.SetCommentRetentionState(EnableCommentOutput, EnableMacroCommentOutput); |
| 516 | |
| 517 | InitOutputBuffer(); |
Chris Lattner | f0f2b29 | 2007-07-23 06:09:34 +0000 | [diff] [blame] | 518 | InitAvoidConcatTokenInfo(); |
Reid Spencer | 5f016e2 | 2007-07-11 17:01:13 +0000 | [diff] [blame] | 519 | |
Chris Lattner | d217773 | 2007-07-20 16:59:19 +0000 | [diff] [blame] | 520 | Token Tok, PrevTok; |
Reid Spencer | 5f016e2 | 2007-07-11 17:01:13 +0000 | [diff] [blame] | 521 | char Buffer[256]; |
| 522 | PrintPPOutputPPCallbacks *Callbacks = new PrintPPOutputPPCallbacks(PP); |
| 523 | PP.setPPCallbacks(Callbacks); |
| 524 | |
| 525 | PP.AddPragmaHandler(0, new UnknownPragmaHandler("#pragma", Callbacks)); |
| 526 | PP.AddPragmaHandler("GCC", new UnknownPragmaHandler("#pragma GCC",Callbacks)); |
| 527 | |
| 528 | // After we have configured the preprocessor, enter the main file. |
| 529 | |
| 530 | // Start parsing the specified input file. |
| 531 | PP.EnterSourceFile(MainFileID, 0, true); |
| 532 | |
| 533 | do { |
| 534 | PrevTok = Tok; |
| 535 | PP.Lex(Tok); |
| 536 | |
| 537 | // If this token is at the start of a line, emit newlines if needed. |
| 538 | if (Tok.isAtStartOfLine()) { |
| 539 | Callbacks->HandleFirstTokOnLine(Tok); |
| 540 | } else if (Tok.hasLeadingSpace() || |
Chris Lattner | f0f2b29 | 2007-07-23 06:09:34 +0000 | [diff] [blame] | 541 | // If we haven't emitted a token on this line yet, PrevTok isn't |
| 542 | // useful to look at and no concatenation could happen anyway. |
Chris Lattner | b638a30 | 2007-07-23 23:21:34 +0000 | [diff] [blame^] | 543 | (Callbacks->hasEmittedTokensOnThisLine() && |
Chris Lattner | f0f2b29 | 2007-07-23 06:09:34 +0000 | [diff] [blame] | 544 | // Don't print "-" next to "-", it would form "--". |
| 545 | Callbacks->AvoidConcat(PrevTok, Tok))) { |
Reid Spencer | 5f016e2 | 2007-07-11 17:01:13 +0000 | [diff] [blame] | 546 | OutputChar(' '); |
| 547 | } |
| 548 | |
Chris Lattner | 2933f41 | 2007-07-23 06:14:36 +0000 | [diff] [blame] | 549 | if (IdentifierInfo *II = Tok.getIdentifierInfo()) { |
| 550 | const char *Str = II->getName(); |
| 551 | unsigned Len = Tok.needsCleaning() ? strlen(Str) : Tok.getLength(); |
| 552 | OutputString(Str, Len); |
| 553 | } else if (Tok.getLength() < 256) { |
Reid Spencer | 5f016e2 | 2007-07-11 17:01:13 +0000 | [diff] [blame] | 554 | const char *TokPtr = Buffer; |
| 555 | unsigned Len = PP.getSpelling(Tok, TokPtr); |
| 556 | OutputString(TokPtr, Len); |
| 557 | } else { |
| 558 | std::string S = PP.getSpelling(Tok); |
| 559 | OutputString(&S[0], S.size()); |
| 560 | } |
| 561 | Callbacks->SetEmittedTokensOnThisLine(); |
| 562 | } while (Tok.getKind() != tok::eof); |
| 563 | OutputChar('\n'); |
| 564 | |
| 565 | CleanupOutputBuffer(); |
| 566 | } |
| 567 | |