Preprocessor.cpp revision 47246be8ac5b0ddde6c402b8fc6946b6135487b5
1//===--- Preprocess.cpp - C Language Family Preprocessor Implementation ---===// 2// 3// The LLVM Compiler Infrastructure 4// 5// This file is distributed under the University of Illinois Open Source 6// License. See LICENSE.TXT for details. 7// 8//===----------------------------------------------------------------------===// 9// 10// This file implements the Preprocessor interface. 11// 12//===----------------------------------------------------------------------===// 13// 14// Options to support: 15// -H - Print the name of each header file used. 16// -d[MDNI] - Dump various things. 17// -fworking-directory - #line's with preprocessor's working dir. 18// -fpreprocessed 19// -dependency-file,-M,-MM,-MF,-MG,-MP,-MT,-MQ,-MD,-MMD 20// -W* 21// -w 22// 23// Messages to emit: 24// "Multiple include guards may be useful for:\n" 25// 26//===----------------------------------------------------------------------===// 27 28#include "clang/Lex/Preprocessor.h" 29#include "clang/Lex/HeaderSearch.h" 30#include "clang/Lex/MacroInfo.h" 31#include "clang/Lex/Pragma.h" 32#include "clang/Lex/ScratchBuffer.h" 33#include "clang/Basic/Diagnostic.h" 34#include "clang/Basic/SourceManager.h" 35#include "clang/Basic/TargetInfo.h" 36#include "llvm/ADT/APFloat.h" 37#include "llvm/ADT/SmallVector.h" 38#include "llvm/Support/MemoryBuffer.h" 39#include "llvm/Support/Streams.h" 40using namespace clang; 41 42//===----------------------------------------------------------------------===// 43 44PreprocessorFactory::~PreprocessorFactory() {} 45 46Preprocessor::Preprocessor(Diagnostic &diags, const LangOptions &opts, 47 TargetInfo &target, SourceManager &SM, 48 HeaderSearch &Headers, 49 IdentifierInfoLookup* IILookup) 50 : Diags(diags), Features(opts), Target(target), FileMgr(Headers.getFileMgr()), 51 SourceMgr(SM), HeaderInfo(Headers), Identifiers(opts, IILookup), 52 CurPPLexer(0), CurDirLookup(0), Callbacks(0) { 53 ScratchBuf = new ScratchBuffer(SourceMgr); 54 55 // Clear stats. 56 NumDirectives = NumDefined = NumUndefined = NumPragma = 0; 57 NumIf = NumElse = NumEndif = 0; 58 NumEnteredSourceFiles = 0; 59 NumMacroExpanded = NumFnMacroExpanded = NumBuiltinMacroExpanded = 0; 60 NumFastMacroExpanded = NumTokenPaste = NumFastTokenPaste = 0; 61 MaxIncludeStackDepth = 0; 62 NumSkipped = 0; 63 64 // Default to discarding comments. 65 KeepComments = false; 66 KeepMacroComments = false; 67 68 // Macro expansion is enabled. 69 DisableMacroExpansion = false; 70 InMacroArgs = false; 71 NumCachedTokenLexers = 0; 72 73 CachedLexPos = 0; 74 75 // "Poison" __VA_ARGS__, which can only appear in the expansion of a macro. 76 // This gets unpoisoned where it is allowed. 77 (Ident__VA_ARGS__ = getIdentifierInfo("__VA_ARGS__"))->setIsPoisoned(); 78 79 // Initialize the pragma handlers. 80 PragmaHandlers = new PragmaNamespace(0); 81 RegisterBuiltinPragmas(); 82 83 // Initialize builtin macros like __LINE__ and friends. 84 RegisterBuiltinMacros(); 85} 86 87Preprocessor::~Preprocessor() { 88 assert(BacktrackPositions.empty() && "EnableBacktrack/Backtrack imbalance!"); 89 90 while (!IncludeMacroStack.empty()) { 91 delete IncludeMacroStack.back().TheLexer; 92 delete IncludeMacroStack.back().TheTokenLexer; 93 IncludeMacroStack.pop_back(); 94 } 95 96 // Free any macro definitions. 97 for (llvm::DenseMap<IdentifierInfo*, MacroInfo*>::iterator I = 98 Macros.begin(), E = Macros.end(); I != E; ++I) { 99 // We don't need to free the MacroInfo objects directly. These 100 // will be released when the BumpPtrAllocator 'BP' object gets 101 // destroyed. We still need to run the dstor, however, to free 102 // memory alocated by MacroInfo. 103 I->second->~MacroInfo(); 104 I->first->setHasMacroDefinition(false); 105 } 106 107 // Free any cached macro expanders. 108 for (unsigned i = 0, e = NumCachedTokenLexers; i != e; ++i) 109 delete TokenLexerCache[i]; 110 111 // Release pragma information. 112 delete PragmaHandlers; 113 114 // Delete the scratch buffer info. 115 delete ScratchBuf; 116 117 delete Callbacks; 118} 119 120void Preprocessor::DumpToken(const Token &Tok, bool DumpFlags) const { 121 llvm::cerr << tok::getTokenName(Tok.getKind()) << " '" 122 << getSpelling(Tok) << "'"; 123 124 if (!DumpFlags) return; 125 126 llvm::cerr << "\t"; 127 if (Tok.isAtStartOfLine()) 128 llvm::cerr << " [StartOfLine]"; 129 if (Tok.hasLeadingSpace()) 130 llvm::cerr << " [LeadingSpace]"; 131 if (Tok.isExpandDisabled()) 132 llvm::cerr << " [ExpandDisabled]"; 133 if (Tok.needsCleaning()) { 134 const char *Start = SourceMgr.getCharacterData(Tok.getLocation()); 135 llvm::cerr << " [UnClean='" << std::string(Start, Start+Tok.getLength()) 136 << "']"; 137 } 138 139 llvm::cerr << "\tLoc=<"; 140 DumpLocation(Tok.getLocation()); 141 llvm::cerr << ">"; 142} 143 144void Preprocessor::DumpLocation(SourceLocation Loc) const { 145 SourceLocation LogLoc = SourceMgr.getInstantiationLoc(Loc); 146 llvm::cerr << SourceMgr.getSourceName(LogLoc) << ':' 147 << SourceMgr.getLineNumber(LogLoc) << ':' 148 << SourceMgr.getColumnNumber(LogLoc); 149 150 SourceLocation SpellingLoc = SourceMgr.getSpellingLoc(Loc); 151 if (SpellingLoc != LogLoc) { 152 llvm::cerr << " <SpellingLoc="; 153 DumpLocation(SpellingLoc); 154 llvm::cerr << ">"; 155 } 156} 157 158void Preprocessor::DumpMacro(const MacroInfo &MI) const { 159 llvm::cerr << "MACRO: "; 160 for (unsigned i = 0, e = MI.getNumTokens(); i != e; ++i) { 161 DumpToken(MI.getReplacementToken(i)); 162 llvm::cerr << " "; 163 } 164 llvm::cerr << "\n"; 165} 166 167void Preprocessor::PrintStats() { 168 llvm::cerr << "\n*** Preprocessor Stats:\n"; 169 llvm::cerr << NumDirectives << " directives found:\n"; 170 llvm::cerr << " " << NumDefined << " #define.\n"; 171 llvm::cerr << " " << NumUndefined << " #undef.\n"; 172 llvm::cerr << " #include/#include_next/#import:\n"; 173 llvm::cerr << " " << NumEnteredSourceFiles << " source files entered.\n"; 174 llvm::cerr << " " << MaxIncludeStackDepth << " max include stack depth\n"; 175 llvm::cerr << " " << NumIf << " #if/#ifndef/#ifdef.\n"; 176 llvm::cerr << " " << NumElse << " #else/#elif.\n"; 177 llvm::cerr << " " << NumEndif << " #endif.\n"; 178 llvm::cerr << " " << NumPragma << " #pragma.\n"; 179 llvm::cerr << NumSkipped << " #if/#ifndef#ifdef regions skipped\n"; 180 181 llvm::cerr << NumMacroExpanded << "/" << NumFnMacroExpanded << "/" 182 << NumBuiltinMacroExpanded << " obj/fn/builtin macros expanded, " 183 << NumFastMacroExpanded << " on the fast path.\n"; 184 llvm::cerr << (NumFastTokenPaste+NumTokenPaste) 185 << " token paste (##) operations performed, " 186 << NumFastTokenPaste << " on the fast path.\n"; 187} 188 189//===----------------------------------------------------------------------===// 190// Token Spelling 191//===----------------------------------------------------------------------===// 192 193 194/// getSpelling() - Return the 'spelling' of this token. The spelling of a 195/// token are the characters used to represent the token in the source file 196/// after trigraph expansion and escaped-newline folding. In particular, this 197/// wants to get the true, uncanonicalized, spelling of things like digraphs 198/// UCNs, etc. 199std::string Preprocessor::getSpelling(const Token &Tok) const { 200 assert((int)Tok.getLength() >= 0 && "Token character range is bogus!"); 201 const char* TokStart; 202 203 if (PTH) { 204 if (unsigned Len = PTH->getSpelling(Tok.getLocation(), TokStart)) { 205 assert(!Tok.needsCleaning()); 206 return std::string(TokStart, TokStart+Len); 207 } 208 } 209 210 // If this token contains nothing interesting, return it directly. 211 TokStart = SourceMgr.getCharacterData(Tok.getLocation()); 212 if (!Tok.needsCleaning()) 213 return std::string(TokStart, TokStart+Tok.getLength()); 214 215 std::string Result; 216 Result.reserve(Tok.getLength()); 217 218 // Otherwise, hard case, relex the characters into the string. 219 for (const char *Ptr = TokStart, *End = TokStart+Tok.getLength(); 220 Ptr != End; ) { 221 unsigned CharSize; 222 Result.push_back(Lexer::getCharAndSizeNoWarn(Ptr, CharSize, Features)); 223 Ptr += CharSize; 224 } 225 assert(Result.size() != unsigned(Tok.getLength()) && 226 "NeedsCleaning flag set on something that didn't need cleaning!"); 227 return Result; 228} 229 230/// getSpelling - This method is used to get the spelling of a token into a 231/// preallocated buffer, instead of as an std::string. The caller is required 232/// to allocate enough space for the token, which is guaranteed to be at least 233/// Tok.getLength() bytes long. The actual length of the token is returned. 234/// 235/// Note that this method may do two possible things: it may either fill in 236/// the buffer specified with characters, or it may *change the input pointer* 237/// to point to a constant buffer with the data already in it (avoiding a 238/// copy). The caller is not allowed to modify the returned buffer pointer 239/// if an internal buffer is returned. 240unsigned Preprocessor::getSpelling(const Token &Tok, 241 const char *&Buffer) const { 242 assert((int)Tok.getLength() >= 0 && "Token character range is bogus!"); 243 244 // If this token is an identifier, just return the string from the identifier 245 // table, which is very quick. 246 if (const IdentifierInfo *II = Tok.getIdentifierInfo()) { 247 Buffer = II->getName(); 248 return II->getLength(); 249 } 250 251 // If using PTH, try and get the spelling from the PTH file. 252 if (PTH) { 253 unsigned Len; 254 255 if (CurPTHLexer) { 256 Len = CurPTHLexer.get()->getSpelling(Tok.getLocation(), Buffer); 257 } else { 258 Len = PTH->getSpelling(Tok.getLocation(), Buffer); 259 } 260 261 // Did we find a spelling? If so return its length. Otherwise fall 262 // back to the default behavior for getting the spelling by looking at 263 // at the source code. 264 if (Len) 265 return Len; 266 } 267 268 // Otherwise, compute the start of the token in the input lexer buffer. 269 const char *TokStart = 0; 270 271 if (Tok.isLiteral()) 272 TokStart = Tok.getLiteralData(); 273 274 if (TokStart == 0) 275 TokStart = SourceMgr.getCharacterData(Tok.getLocation()); 276 277 // If this token contains nothing interesting, return it directly. 278 if (!Tok.needsCleaning()) { 279 Buffer = TokStart; 280 return Tok.getLength(); 281 } 282 283 // Otherwise, hard case, relex the characters into the string. 284 char *OutBuf = const_cast<char*>(Buffer); 285 for (const char *Ptr = TokStart, *End = TokStart+Tok.getLength(); 286 Ptr != End; ) { 287 unsigned CharSize; 288 *OutBuf++ = Lexer::getCharAndSizeNoWarn(Ptr, CharSize, Features); 289 Ptr += CharSize; 290 } 291 assert(unsigned(OutBuf-Buffer) != Tok.getLength() && 292 "NeedsCleaning flag set on something that didn't need cleaning!"); 293 294 return OutBuf-Buffer; 295} 296 297 298/// CreateString - Plop the specified string into a scratch buffer and return a 299/// location for it. If specified, the source location provides a source 300/// location for the token. 301void Preprocessor::CreateString(const char *Buf, unsigned Len, Token &Tok, 302 SourceLocation InstantiationLoc) { 303 Tok.setLength(Len); 304 305 const char *DestPtr; 306 SourceLocation Loc = ScratchBuf->getToken(Buf, Len, DestPtr); 307 308 if (InstantiationLoc.isValid()) 309 Loc = SourceMgr.createInstantiationLoc(Loc, InstantiationLoc, Len); 310 Tok.setLocation(Loc); 311 312 // If this is a literal token, set the pointer data. 313 if (Tok.isLiteral()) 314 Tok.setLiteralData(DestPtr); 315} 316 317 318/// AdvanceToTokenCharacter - Given a location that specifies the start of a 319/// token, return a new location that specifies a character within the token. 320SourceLocation Preprocessor::AdvanceToTokenCharacter(SourceLocation TokStart, 321 unsigned CharNo) { 322 // If they request the first char of the token, we're trivially done. If this 323 // is a macro expansion, it doesn't make sense to point to a character within 324 // the instantiation point (the name). We could point to the source 325 // character, but without also pointing to instantiation info, this is 326 // confusing. 327 if (CharNo == 0 || TokStart.isMacroID()) return TokStart; 328 329 // Figure out how many physical characters away the specified instantiation 330 // character is. This needs to take into consideration newlines and 331 // trigraphs. 332 const char *TokPtr = SourceMgr.getCharacterData(TokStart); 333 unsigned PhysOffset = 0; 334 335 // The usual case is that tokens don't contain anything interesting. Skip 336 // over the uninteresting characters. If a token only consists of simple 337 // chars, this method is extremely fast. 338 while (CharNo && Lexer::isObviouslySimpleCharacter(*TokPtr)) 339 ++TokPtr, --CharNo, ++PhysOffset; 340 341 // If we have a character that may be a trigraph or escaped newline, use a 342 // lexer to parse it correctly. 343 if (CharNo != 0) { 344 // Skip over characters the remaining characters. 345 for (; CharNo; --CharNo) { 346 unsigned Size; 347 Lexer::getCharAndSizeNoWarn(TokPtr, Size, Features); 348 TokPtr += Size; 349 PhysOffset += Size; 350 } 351 } 352 353 return TokStart.getFileLocWithOffset(PhysOffset); 354} 355 356 357//===----------------------------------------------------------------------===// 358// Preprocessor Initialization Methods 359//===----------------------------------------------------------------------===// 360 361// Append a #define line to Buf for Macro. Macro should be of the form XXX, 362// in which case we emit "#define XXX 1" or "XXX=Y z W" in which case we emit 363// "#define XXX Y z W". To get a #define with no value, use "XXX=". 364static void DefineBuiltinMacro(std::vector<char> &Buf, const char *Macro, 365 const char *Command = "#define ") { 366 Buf.insert(Buf.end(), Command, Command+strlen(Command)); 367 if (const char *Equal = strchr(Macro, '=')) { 368 // Turn the = into ' '. 369 Buf.insert(Buf.end(), Macro, Equal); 370 Buf.push_back(' '); 371 Buf.insert(Buf.end(), Equal+1, Equal+strlen(Equal)); 372 } else { 373 // Push "macroname 1". 374 Buf.insert(Buf.end(), Macro, Macro+strlen(Macro)); 375 Buf.push_back(' '); 376 Buf.push_back('1'); 377 } 378 Buf.push_back('\n'); 379} 380 381/// PickFP - This is used to pick a value based on the FP semantics of the 382/// specified FP model. 383template <typename T> 384static T PickFP(const llvm::fltSemantics *Sem, T IEEESingleVal, 385 T IEEEDoubleVal, T X87DoubleExtendedVal, T PPCDoubleDoubleVal) { 386 if (Sem == &llvm::APFloat::IEEEsingle) 387 return IEEESingleVal; 388 if (Sem == &llvm::APFloat::IEEEdouble) 389 return IEEEDoubleVal; 390 if (Sem == &llvm::APFloat::x87DoubleExtended) 391 return X87DoubleExtendedVal; 392 assert(Sem == &llvm::APFloat::PPCDoubleDouble); 393 return PPCDoubleDoubleVal; 394} 395 396static void DefineFloatMacros(std::vector<char> &Buf, const char *Prefix, 397 const llvm::fltSemantics *Sem) { 398 const char *DenormMin, *Epsilon, *Max, *Min; 399 DenormMin = PickFP(Sem, "1.40129846e-45F", "4.9406564584124654e-324", 400 "3.64519953188247460253e-4951L", 401 "4.94065645841246544176568792868221e-324L"); 402 int Digits = PickFP(Sem, 6, 15, 18, 31); 403 Epsilon = PickFP(Sem, "1.19209290e-7F", "2.2204460492503131e-16", 404 "1.08420217248550443401e-19L", 405 "4.94065645841246544176568792868221e-324L"); 406 int HasInifinity = 1, HasQuietNaN = 1; 407 int MantissaDigits = PickFP(Sem, 24, 53, 64, 106); 408 int Min10Exp = PickFP(Sem, -37, -307, -4931, -291); 409 int Max10Exp = PickFP(Sem, 38, 308, 4932, 308); 410 int MinExp = PickFP(Sem, -125, -1021, -16381, -968); 411 int MaxExp = PickFP(Sem, 128, 1024, 16384, 1024); 412 Min = PickFP(Sem, "1.17549435e-38F", "2.2250738585072014e-308", 413 "3.36210314311209350626e-4932L", 414 "2.00416836000897277799610805135016e-292L"); 415 Max = PickFP(Sem, "3.40282347e+38F", "1.7976931348623157e+308", 416 "1.18973149535723176502e+4932L", 417 "1.79769313486231580793728971405301e+308L"); 418 419 char MacroBuf[60]; 420 sprintf(MacroBuf, "__%s_DENORM_MIN__=%s", Prefix, DenormMin); 421 DefineBuiltinMacro(Buf, MacroBuf); 422 sprintf(MacroBuf, "__%s_DIG__=%d", Prefix, Digits); 423 DefineBuiltinMacro(Buf, MacroBuf); 424 sprintf(MacroBuf, "__%s_EPSILON__=%s", Prefix, Epsilon); 425 DefineBuiltinMacro(Buf, MacroBuf); 426 sprintf(MacroBuf, "__%s_HAS_INFINITY__=%d", Prefix, HasInifinity); 427 DefineBuiltinMacro(Buf, MacroBuf); 428 sprintf(MacroBuf, "__%s_HAS_QUIET_NAN__=%d", Prefix, HasQuietNaN); 429 DefineBuiltinMacro(Buf, MacroBuf); 430 sprintf(MacroBuf, "__%s_MANT_DIG__=%d", Prefix, MantissaDigits); 431 DefineBuiltinMacro(Buf, MacroBuf); 432 sprintf(MacroBuf, "__%s_MAX_10_EXP__=%d", Prefix, Max10Exp); 433 DefineBuiltinMacro(Buf, MacroBuf); 434 sprintf(MacroBuf, "__%s_MAX_EXP__=%d", Prefix, MaxExp); 435 DefineBuiltinMacro(Buf, MacroBuf); 436 sprintf(MacroBuf, "__%s_MAX__=%s", Prefix, Max); 437 DefineBuiltinMacro(Buf, MacroBuf); 438 sprintf(MacroBuf, "__%s_MIN_10_EXP__=(%d)", Prefix, Min10Exp); 439 DefineBuiltinMacro(Buf, MacroBuf); 440 sprintf(MacroBuf, "__%s_MIN_EXP__=(%d)", Prefix, MinExp); 441 DefineBuiltinMacro(Buf, MacroBuf); 442 sprintf(MacroBuf, "__%s_MIN__=%s", Prefix, Min); 443 DefineBuiltinMacro(Buf, MacroBuf); 444} 445 446 447static void InitializePredefinedMacros(Preprocessor &PP, 448 std::vector<char> &Buf) { 449 // Compiler version introspection macros. 450 DefineBuiltinMacro(Buf, "__llvm__=1"); // LLVM Backend 451 DefineBuiltinMacro(Buf, "__clang__=1"); // Clang Frontend 452 453 // Currently claim to be compatible with GCC 4.2.1-5621. 454 DefineBuiltinMacro(Buf, "__APPLE_CC__=5621"); 455 DefineBuiltinMacro(Buf, "__GNUC_MINOR__=2"); 456 DefineBuiltinMacro(Buf, "__GNUC_PATCHLEVEL__=1"); 457 DefineBuiltinMacro(Buf, "__GNUC__=4"); 458 DefineBuiltinMacro(Buf, "__GXX_ABI_VERSION=1002"); 459 DefineBuiltinMacro(Buf, "__VERSION__=\"4.2.1 (Apple Computer, Inc. " 460 "build 5621) (dot 3)\""); 461 462 463 // Initialize language-specific preprocessor defines. 464 465 // FIXME: Implement magic like cpp_init_builtins for things like __STDC__ 466 // and __DATE__ etc. 467 // These should all be defined in the preprocessor according to the 468 // current language configuration. 469 if (!PP.getLangOptions().Microsoft) 470 DefineBuiltinMacro(Buf, "__STDC__=1"); 471 if (PP.getLangOptions().AsmPreprocessor) 472 DefineBuiltinMacro(Buf, "__ASSEMBLER__=1"); 473 if (PP.getLangOptions().C99 && !PP.getLangOptions().CPlusPlus) 474 DefineBuiltinMacro(Buf, "__STDC_VERSION__=199901L"); 475 else if (0) // STDC94 ? 476 DefineBuiltinMacro(Buf, "__STDC_VERSION__=199409L"); 477 478 DefineBuiltinMacro(Buf, "__STDC_HOSTED__=1"); 479 if (PP.getLangOptions().ObjC1) { 480 DefineBuiltinMacro(Buf, "__OBJC__=1"); 481 482 if (PP.getLangOptions().getGCMode() == LangOptions::NonGC) { 483 DefineBuiltinMacro(Buf, "__weak="); 484 DefineBuiltinMacro(Buf, "__strong="); 485 } else { 486 DefineBuiltinMacro(Buf, "__weak=__attribute__((objc_gc(weak)))"); 487 DefineBuiltinMacro(Buf, "__strong=__attribute__((objc_gc(strong)))"); 488 DefineBuiltinMacro(Buf, "__OBJC_GC__=1"); 489 } 490 491 if (PP.getLangOptions().NeXTRuntime) 492 DefineBuiltinMacro(Buf, "__NEXT_RUNTIME__=1"); 493 } 494 495 // darwin_constant_cfstrings controls this. This is also dependent 496 // on other things like the runtime I believe. This is set even for C code. 497 DefineBuiltinMacro(Buf, "__CONSTANT_CFSTRINGS__=1"); 498 499 if (PP.getLangOptions().ObjC2) 500 DefineBuiltinMacro(Buf, "OBJC_NEW_PROPERTIES"); 501 502 if (PP.getLangOptions().PascalStrings) 503 DefineBuiltinMacro(Buf, "__PASCAL_STRINGS__"); 504 505 if (PP.getLangOptions().Blocks) { 506 DefineBuiltinMacro(Buf, "__block=__attribute__((__blocks__(byref)))"); 507 DefineBuiltinMacro(Buf, "__BLOCKS__=1"); 508 } 509 510 if (PP.getLangOptions().CPlusPlus) { 511 DefineBuiltinMacro(Buf, "__DEPRECATED=1"); 512 DefineBuiltinMacro(Buf, "__EXCEPTIONS=1"); 513 DefineBuiltinMacro(Buf, "__GNUG__=4"); 514 DefineBuiltinMacro(Buf, "__GXX_WEAK__=1"); 515 DefineBuiltinMacro(Buf, "__cplusplus=1"); 516 DefineBuiltinMacro(Buf, "__private_extern__=extern"); 517 } 518 519 // Filter out some microsoft extensions when trying to parse in ms-compat 520 // mode. 521 if (PP.getLangOptions().Microsoft) { 522 DefineBuiltinMacro(Buf, "_cdecl=__cdecl"); 523 DefineBuiltinMacro(Buf, "__int8=char"); 524 DefineBuiltinMacro(Buf, "__int16=short"); 525 DefineBuiltinMacro(Buf, "__int32=int"); 526 DefineBuiltinMacro(Buf, "__int64=long long"); 527 } 528 529 530 // Initialize target-specific preprocessor defines. 531 const TargetInfo &TI = PP.getTargetInfo(); 532 533 // Define type sizing macros based on the target properties. 534 assert(TI.getCharWidth() == 8 && "Only support 8-bit char so far"); 535 DefineBuiltinMacro(Buf, "__CHAR_BIT__=8"); 536 DefineBuiltinMacro(Buf, "__SCHAR_MAX__=127"); 537 538 assert(TI.getWCharWidth() == 32 && "Only support 32-bit wchar so far"); 539 DefineBuiltinMacro(Buf, "__WCHAR_MAX__=2147483647"); 540 DefineBuiltinMacro(Buf, "__WCHAR_TYPE__=int"); 541 DefineBuiltinMacro(Buf, "__WINT_TYPE__=int"); 542 543 assert(TI.getShortWidth() == 16 && "Only support 16-bit short so far"); 544 DefineBuiltinMacro(Buf, "__SHRT_MAX__=32767"); 545 546 if (TI.getIntWidth() == 32) 547 DefineBuiltinMacro(Buf, "__INT_MAX__=2147483647"); 548 else if (TI.getIntWidth() == 16) 549 DefineBuiltinMacro(Buf, "__INT_MAX__=32767"); 550 else 551 assert(0 && "Unknown integer size"); 552 553 if (TI.getLongLongWidth() == 64) 554 DefineBuiltinMacro(Buf, "__LONG_LONG_MAX__=9223372036854775807LL"); 555 else if (TI.getLongLongWidth() == 32) 556 DefineBuiltinMacro(Buf, "__LONG_LONG_MAX__=2147483647L"); 557 558 if (TI.getLongWidth() == 32) 559 DefineBuiltinMacro(Buf, "__LONG_MAX__=2147483647L"); 560 else if (TI.getLongWidth() == 64) 561 DefineBuiltinMacro(Buf, "__LONG_MAX__=9223372036854775807L"); 562 else if (TI.getLongWidth() == 16) 563 DefineBuiltinMacro(Buf, "__LONG_MAX__=32767L"); 564 else 565 assert(0 && "Unknown long size"); 566 char MacroBuf[60]; 567 sprintf(MacroBuf, "__INTMAX_MAX__=%lld", 568 (TI.getIntMaxType() == TargetInfo::UnsignedLongLong? 569 (1LL << (TI.getLongLongWidth() - 1)) : 570 ((1LL << (TI.getLongLongWidth() - 2)) - 1))); 571 DefineBuiltinMacro(Buf, MacroBuf); 572 573 if (TI.getIntMaxType() == TargetInfo::UnsignedLongLong) 574 DefineBuiltinMacro(Buf, "__INTMAX_TYPE__=unsigned long long int"); 575 else if (TI.getIntMaxType() == TargetInfo::SignedLongLong) 576 DefineBuiltinMacro(Buf, "__INTMAX_TYPE__=long long int"); 577 else if (TI.getIntMaxType() == TargetInfo::UnsignedLong) 578 DefineBuiltinMacro(Buf, "__INTMAX_TYPE__=unsigned long int"); 579 else if (TI.getIntMaxType() == TargetInfo::SignedLong) 580 DefineBuiltinMacro(Buf, "__INTMAX_TYPE__=long int"); 581 else if (TI.getIntMaxType() == TargetInfo::UnsignedInt) 582 DefineBuiltinMacro(Buf, "__INTMAX_TYPE__=unsigned int"); 583 else 584 DefineBuiltinMacro(Buf, "__INTMAX_TYPE__=int"); 585 586 if (TI.getUIntMaxType() == TargetInfo::UnsignedLongLong) 587 DefineBuiltinMacro(Buf, "__UINTMAX_TYPE__=unsigned long long int"); 588 else if (TI.getUIntMaxType() == TargetInfo::SignedLongLong) 589 DefineBuiltinMacro(Buf, "__UINTMAX_TYPE__=long long int"); 590 else if (TI.getUIntMaxType() == TargetInfo::UnsignedLong) 591 DefineBuiltinMacro(Buf, "__UINTMAX_TYPE__=unsigned long int"); 592 else if (TI.getUIntMaxType() == TargetInfo::SignedLong) 593 DefineBuiltinMacro(Buf, "__UINTMAX_TYPE__=long int"); 594 else if (TI.getUIntMaxType() == TargetInfo::UnsignedInt) 595 DefineBuiltinMacro(Buf, "__UINTMAX_TYPE__=unsigned int"); 596 else 597 DefineBuiltinMacro(Buf, "__UINTMAX_TYPE__=int"); 598 599 if (TI.getPtrDiffType(0) == TargetInfo::UnsignedLongLong) 600 DefineBuiltinMacro(Buf, "__PTRDIFF_TYPE__=unsigned long long int"); 601 else if (TI.getPtrDiffType(0) == TargetInfo::SignedLongLong) 602 DefineBuiltinMacro(Buf, "__PTRDIFF_TYPE__=long long int"); 603 else if (TI.getPtrDiffType(0) == TargetInfo::UnsignedLong) 604 DefineBuiltinMacro(Buf, "__PTRDIFF_TYPE__=unsigned long int"); 605 else if (TI.getPtrDiffType(0) == TargetInfo::SignedLong) 606 DefineBuiltinMacro(Buf, "__PTRDIFF_TYPE__=long int"); 607 else if (TI.getPtrDiffType(0) == TargetInfo::UnsignedInt) 608 DefineBuiltinMacro(Buf, "__PTRDIFF_TYPE__=unsigned int"); 609 else 610 DefineBuiltinMacro(Buf, "__PTRDIFF_TYPE__=int"); 611 612 if (TI.getSizeType() == TargetInfo::UnsignedLongLong) 613 DefineBuiltinMacro(Buf, "__SIZE_TYPE__=unsigned long long int"); 614 else if (TI.getSizeType() == TargetInfo::SignedLongLong) 615 DefineBuiltinMacro(Buf, "__SIZE_TYPE__=long long int"); 616 else if (TI.getSizeType() == TargetInfo::UnsignedLong) 617 DefineBuiltinMacro(Buf, "__SIZE_TYPE__=unsigned long int"); 618 else if (TI.getSizeType() == TargetInfo::SignedLong) 619 DefineBuiltinMacro(Buf, "__SIZE_TYPE__=long int"); 620 else if (TI.getSizeType() == TargetInfo::UnsignedInt) 621 DefineBuiltinMacro(Buf, "__SIZE_TYPE__=unsigned int"); 622 else if (TI.getSizeType() == TargetInfo::SignedInt) 623 DefineBuiltinMacro(Buf, "__SIZE_TYPE__=int"); 624 else 625 DefineBuiltinMacro(Buf, "__SIZE_TYPE__=unsigned short"); 626 627 DefineFloatMacros(Buf, "FLT", &TI.getFloatFormat()); 628 DefineFloatMacros(Buf, "DBL", &TI.getDoubleFormat()); 629 DefineFloatMacros(Buf, "LDBL", &TI.getLongDoubleFormat()); 630 631 632 // Add __builtin_va_list typedef. 633 { 634 const char *VAList = TI.getVAListDeclaration(); 635 Buf.insert(Buf.end(), VAList, VAList+strlen(VAList)); 636 Buf.push_back('\n'); 637 } 638 639 if (const char *Prefix = TI.getUserLabelPrefix()) { 640 sprintf(MacroBuf, "__USER_LABEL_PREFIX__=%s", Prefix); 641 DefineBuiltinMacro(Buf, MacroBuf); 642 } 643 644 // Build configuration options. FIXME: these should be controlled by 645 // command line options or something. 646 DefineBuiltinMacro(Buf, "__DYNAMIC__=1"); 647 DefineBuiltinMacro(Buf, "__FINITE_MATH_ONLY__=0"); 648 DefineBuiltinMacro(Buf, "__NO_INLINE__=1"); 649 DefineBuiltinMacro(Buf, "__PIC__=1"); 650 651 // Macros to control C99 numerics and <float.h> 652 DefineBuiltinMacro(Buf, "__FLT_EVAL_METHOD__=0"); 653 DefineBuiltinMacro(Buf, "__FLT_RADIX__=2"); 654 sprintf(MacroBuf, "__DECIMAL_DIG__=%d", 655 PickFP(&TI.getLongDoubleFormat(), -1/*FIXME*/, 17, 21, 33)); 656 DefineBuiltinMacro(Buf, MacroBuf); 657 658 // Get other target #defines. 659 TI.getTargetDefines(Buf); 660 661 // FIXME: Should emit a #line directive here. 662} 663 664 665/// EnterMainSourceFile - Enter the specified FileID as the main source file, 666/// which implicitly adds the builtin defines etc. 667void Preprocessor::EnterMainSourceFile() { 668 669 FileID MainFileID = SourceMgr.getMainFileID(); 670 671 // Enter the main file source buffer. 672 EnterSourceFile(MainFileID, 0); 673 674 // Tell the header info that the main file was entered. If the file is later 675 // #imported, it won't be re-entered. 676 if (const FileEntry *FE = SourceMgr.getFileEntryForID(MainFileID)) 677 HeaderInfo.IncrementIncludeCount(FE); 678 679 std::vector<char> PrologFile; 680 PrologFile.reserve(4080); 681 682 // Install things like __POWERPC__, __GNUC__, etc into the macro table. 683 InitializePredefinedMacros(*this, PrologFile); 684 685 // Add on the predefines from the driver. 686 PrologFile.insert(PrologFile.end(), Predefines.begin(), Predefines.end()); 687 688 // Memory buffer must end with a null byte! 689 PrologFile.push_back(0); 690 691 // Now that we have emitted the predefined macros, #includes, etc into 692 // PrologFile, preprocess it to populate the initial preprocessor state. 693 llvm::MemoryBuffer *SB = 694 llvm::MemoryBuffer::getMemBufferCopy(&PrologFile.front(),&PrologFile.back(), 695 "<predefines>"); 696 assert(SB && "Cannot fail to create predefined source buffer"); 697 FileID FID = SourceMgr.createFileIDForMemBuffer(SB); 698 assert(!FID.isInvalid() && "Could not create FileID for predefines?"); 699 700 // Start parsing the predefines. 701 EnterSourceFile(FID, 0); 702} 703 704 705//===----------------------------------------------------------------------===// 706// Lexer Event Handling. 707//===----------------------------------------------------------------------===// 708 709/// LookUpIdentifierInfo - Given a tok::identifier token, look up the 710/// identifier information for the token and install it into the token. 711IdentifierInfo *Preprocessor::LookUpIdentifierInfo(Token &Identifier, 712 const char *BufPtr) { 713 assert(Identifier.is(tok::identifier) && "Not an identifier!"); 714 assert(Identifier.getIdentifierInfo() == 0 && "Identinfo already exists!"); 715 716 // Look up this token, see if it is a macro, or if it is a language keyword. 717 IdentifierInfo *II; 718 if (BufPtr && !Identifier.needsCleaning()) { 719 // No cleaning needed, just use the characters from the lexed buffer. 720 II = getIdentifierInfo(BufPtr, BufPtr+Identifier.getLength()); 721 } else { 722 // Cleaning needed, alloca a buffer, clean into it, then use the buffer. 723 llvm::SmallVector<char, 64> IdentifierBuffer; 724 IdentifierBuffer.resize(Identifier.getLength()); 725 const char *TmpBuf = &IdentifierBuffer[0]; 726 unsigned Size = getSpelling(Identifier, TmpBuf); 727 II = getIdentifierInfo(TmpBuf, TmpBuf+Size); 728 } 729 Identifier.setIdentifierInfo(II); 730 return II; 731} 732 733 734/// HandleIdentifier - This callback is invoked when the lexer reads an 735/// identifier. This callback looks up the identifier in the map and/or 736/// potentially macro expands it or turns it into a named token (like 'for'). 737/// 738/// Note that callers of this method are guarded by checking the 739/// IdentifierInfo's 'isHandleIdentifierCase' bit. If this method changes, the 740/// IdentifierInfo methods that compute these properties will need to change to 741/// match. 742void Preprocessor::HandleIdentifier(Token &Identifier) { 743 assert(Identifier.getIdentifierInfo() && 744 "Can't handle identifiers without identifier info!"); 745 746 IdentifierInfo &II = *Identifier.getIdentifierInfo(); 747 748 // If this identifier was poisoned, and if it was not produced from a macro 749 // expansion, emit an error. 750 if (II.isPoisoned() && CurPPLexer) { 751 if (&II != Ident__VA_ARGS__) // We warn about __VA_ARGS__ with poisoning. 752 Diag(Identifier, diag::err_pp_used_poisoned_id); 753 else 754 Diag(Identifier, diag::ext_pp_bad_vaargs_use); 755 } 756 757 // If this is a macro to be expanded, do it. 758 if (MacroInfo *MI = getMacroInfo(&II)) { 759 if (!DisableMacroExpansion && !Identifier.isExpandDisabled()) { 760 if (MI->isEnabled()) { 761 if (!HandleMacroExpandedIdentifier(Identifier, MI)) 762 return; 763 } else { 764 // C99 6.10.3.4p2 says that a disabled macro may never again be 765 // expanded, even if it's in a context where it could be expanded in the 766 // future. 767 Identifier.setFlag(Token::DisableExpand); 768 } 769 } 770 } 771 772 // C++ 2.11p2: If this is an alternative representation of a C++ operator, 773 // then we act as if it is the actual operator and not the textual 774 // representation of it. 775 if (II.isCPlusPlusOperatorKeyword()) 776 Identifier.setIdentifierInfo(0); 777 778 // If this is an extension token, diagnose its use. 779 // We avoid diagnosing tokens that originate from macro definitions. 780 if (II.isExtensionToken() && Features.C99 && !DisableMacroExpansion) 781 Diag(Identifier, diag::ext_token_used); 782} 783