1 //===--- PrintPreprocessedOutput.cpp - Implement the -E mode --------------===// 2 // 3 // Part of the LLVM Project, under the Apache License v2.0 with LLVM Exceptions. 4 // See https://llvm.org/LICENSE.txt for license information. 5 // SPDX-License-Identifier: Apache-2.0 WITH LLVM-exception 6 // 7 //===----------------------------------------------------------------------===// 8 // 9 // This code simply runs the preprocessor on the input file and prints out the 10 // result. This is the traditional behavior of the -E option. 11 // 12 //===----------------------------------------------------------------------===// 13 14 #include "clang/Frontend/Utils.h" 15 #include "clang/Basic/CharInfo.h" 16 #include "clang/Basic/Diagnostic.h" 17 #include "clang/Basic/SourceManager.h" 18 #include "clang/Frontend/PreprocessorOutputOptions.h" 19 #include "clang/Lex/MacroInfo.h" 20 #include "clang/Lex/PPCallbacks.h" 21 #include "clang/Lex/Pragma.h" 22 #include "clang/Lex/Preprocessor.h" 23 #include "clang/Lex/TokenConcatenation.h" 24 #include "llvm/ADT/STLExtras.h" 25 #include "llvm/ADT/SmallString.h" 26 #include "llvm/ADT/StringRef.h" 27 #include "llvm/Support/ErrorHandling.h" 28 #include "llvm/Support/raw_ostream.h" 29 #include <cstdio> 30 using namespace clang; 31 32 /// PrintMacroDefinition - Print a macro definition in a form that will be 33 /// properly accepted back as a definition. 34 static void PrintMacroDefinition(const IdentifierInfo &II, const MacroInfo &MI, 35 Preprocessor &PP, raw_ostream &OS) { 36 OS << "#define " << II.getName(); 37 38 if (MI.isFunctionLike()) { 39 OS << '('; 40 if (!MI.param_empty()) { 41 MacroInfo::param_iterator AI = MI.param_begin(), E = MI.param_end(); 42 for (; AI+1 != E; ++AI) { 43 OS << (*AI)->getName(); 44 OS << ','; 45 } 46 47 // Last argument. 48 if ((*AI)->getName() == "__VA_ARGS__") 49 OS << "..."; 50 else 51 OS << (*AI)->getName(); 52 } 53 54 if (MI.isGNUVarargs()) 55 OS << "..."; // #define foo(x...) 56 57 OS << ')'; 58 } 59 60 // GCC always emits a space, even if the macro body is empty. However, do not 61 // want to emit two spaces if the first token has a leading space. 62 if (MI.tokens_empty() || !MI.tokens_begin()->hasLeadingSpace()) 63 OS << ' '; 64 65 SmallString<128> SpellingBuffer; 66 for (const auto &T : MI.tokens()) { 67 if (T.hasLeadingSpace()) 68 OS << ' '; 69 70 OS << PP.getSpelling(T, SpellingBuffer); 71 } 72 } 73 74 //===----------------------------------------------------------------------===// 75 // Preprocessed token printer 76 //===----------------------------------------------------------------------===// 77 78 namespace { 79 class PrintPPOutputPPCallbacks : public PPCallbacks { 80 Preprocessor &PP; 81 SourceManager &SM; 82 TokenConcatenation ConcatInfo; 83 public: 84 raw_ostream &OS; 85 private: 86 unsigned CurLine; 87 88 bool EmittedTokensOnThisLine; 89 bool EmittedDirectiveOnThisLine; 90 SrcMgr::CharacteristicKind FileType; 91 SmallString<512> CurFilename; 92 bool Initialized; 93 bool DisableLineMarkers; 94 bool DumpDefines; 95 bool DumpIncludeDirectives; 96 bool UseLineDirectives; 97 bool IsFirstFileEntered; 98 bool MinimizeWhitespace; 99 100 Token PrevTok; 101 Token PrevPrevTok; 102 103 public: 104 PrintPPOutputPPCallbacks(Preprocessor &pp, raw_ostream &os, bool lineMarkers, 105 bool defines, bool DumpIncludeDirectives, 106 bool UseLineDirectives, bool MinimizeWhitespace) 107 : PP(pp), SM(PP.getSourceManager()), ConcatInfo(PP), OS(os), 108 DisableLineMarkers(lineMarkers), DumpDefines(defines), 109 DumpIncludeDirectives(DumpIncludeDirectives), 110 UseLineDirectives(UseLineDirectives), 111 MinimizeWhitespace(MinimizeWhitespace) { 112 CurLine = 0; 113 CurFilename += "<uninit>"; 114 EmittedTokensOnThisLine = false; 115 EmittedDirectiveOnThisLine = false; 116 FileType = SrcMgr::C_User; 117 Initialized = false; 118 IsFirstFileEntered = false; 119 120 PrevTok.startToken(); 121 PrevPrevTok.startToken(); 122 } 123 124 bool isMinimizeWhitespace() const { return MinimizeWhitespace; } 125 126 void setEmittedTokensOnThisLine() { EmittedTokensOnThisLine = true; } 127 bool hasEmittedTokensOnThisLine() const { return EmittedTokensOnThisLine; } 128 129 void setEmittedDirectiveOnThisLine() { EmittedDirectiveOnThisLine = true; } 130 bool hasEmittedDirectiveOnThisLine() const { 131 return EmittedDirectiveOnThisLine; 132 } 133 134 /// Ensure that the output stream position is at the beginning of a new line 135 /// and inserts one if it does not. It is intended to ensure that directives 136 /// inserted by the directives not from the input source (such as #line) are 137 /// in the first column. To insert newlines that represent the input, use 138 /// MoveToLine(/*...*/, /*RequireStartOfLine=*/true). 139 void startNewLineIfNeeded(); 140 141 void FileChanged(SourceLocation Loc, FileChangeReason Reason, 142 SrcMgr::CharacteristicKind FileType, 143 FileID PrevFID) override; 144 void InclusionDirective(SourceLocation HashLoc, const Token &IncludeTok, 145 StringRef FileName, bool IsAngled, 146 CharSourceRange FilenameRange, const FileEntry *File, 147 StringRef SearchPath, StringRef RelativePath, 148 const Module *Imported, 149 SrcMgr::CharacteristicKind FileType) override; 150 void Ident(SourceLocation Loc, StringRef str) override; 151 void PragmaMessage(SourceLocation Loc, StringRef Namespace, 152 PragmaMessageKind Kind, StringRef Str) override; 153 void PragmaDebug(SourceLocation Loc, StringRef DebugType) override; 154 void PragmaDiagnosticPush(SourceLocation Loc, StringRef Namespace) override; 155 void PragmaDiagnosticPop(SourceLocation Loc, StringRef Namespace) override; 156 void PragmaDiagnostic(SourceLocation Loc, StringRef Namespace, 157 diag::Severity Map, StringRef Str) override; 158 void PragmaWarning(SourceLocation Loc, StringRef WarningSpec, 159 ArrayRef<int> Ids) override; 160 void PragmaWarningPush(SourceLocation Loc, int Level) override; 161 void PragmaWarningPop(SourceLocation Loc) override; 162 void PragmaExecCharsetPush(SourceLocation Loc, StringRef Str) override; 163 void PragmaExecCharsetPop(SourceLocation Loc) override; 164 void PragmaAssumeNonNullBegin(SourceLocation Loc) override; 165 void PragmaAssumeNonNullEnd(SourceLocation Loc) override; 166 167 /// Insert whitespace before emitting the next token. 168 /// 169 /// @param Tok Next token to be emitted. 170 /// @param RequireSpace Ensure at least one whitespace is emitted. Useful 171 /// if non-tokens have been emitted to the stream. 172 /// @param RequireSameLine Never emit newlines. Useful when semantics depend 173 /// on being on the same line, such as directives. 174 void HandleWhitespaceBeforeTok(const Token &Tok, bool RequireSpace, 175 bool RequireSameLine); 176 177 /// Move to the line of the provided source location. This will 178 /// return true if a newline was inserted or if 179 /// the requested location is the first token on the first line. 180 /// In these cases the next output will be the first column on the line and 181 /// make it possible to insert indention. The newline was inserted 182 /// implicitly when at the beginning of the file. 183 /// 184 /// @param Tok Token where to move to. 185 /// @param RequiresStartOfLine Whether the next line depends on being in the 186 /// first column, such as a directive. 187 /// 188 /// @return Whether column adjustments are necessary. 189 bool MoveToLine(const Token &Tok, bool RequireStartOfLine) { 190 PresumedLoc PLoc = SM.getPresumedLoc(Tok.getLocation()); 191 if (PLoc.isInvalid()) 192 return false; 193 bool IsFirstInFile = Tok.isAtStartOfLine() && PLoc.getLine() == 1; 194 return MoveToLine(PLoc.getLine(), RequireStartOfLine) || IsFirstInFile; 195 } 196 197 /// Move to the line of the provided source location. Returns true if a new 198 /// line was inserted. 199 bool MoveToLine(SourceLocation Loc, bool RequireStartOfLine) { 200 PresumedLoc PLoc = SM.getPresumedLoc(Loc); 201 if (PLoc.isInvalid()) 202 return false; 203 return MoveToLine(PLoc.getLine(), RequireStartOfLine); 204 } 205 bool MoveToLine(unsigned LineNo, bool RequireStartOfLine); 206 207 bool AvoidConcat(const Token &PrevPrevTok, const Token &PrevTok, 208 const Token &Tok) { 209 return ConcatInfo.AvoidConcat(PrevPrevTok, PrevTok, Tok); 210 } 211 void WriteLineInfo(unsigned LineNo, const char *Extra=nullptr, 212 unsigned ExtraLen=0); 213 bool LineMarkersAreDisabled() const { return DisableLineMarkers; } 214 void HandleNewlinesInToken(const char *TokStr, unsigned Len); 215 216 /// MacroDefined - This hook is called whenever a macro definition is seen. 217 void MacroDefined(const Token &MacroNameTok, 218 const MacroDirective *MD) override; 219 220 /// MacroUndefined - This hook is called whenever a macro #undef is seen. 221 void MacroUndefined(const Token &MacroNameTok, 222 const MacroDefinition &MD, 223 const MacroDirective *Undef) override; 224 225 void BeginModule(const Module *M); 226 void EndModule(const Module *M); 227 }; 228 } // end anonymous namespace 229 230 void PrintPPOutputPPCallbacks::WriteLineInfo(unsigned LineNo, 231 const char *Extra, 232 unsigned ExtraLen) { 233 startNewLineIfNeeded(); 234 235 // Emit #line directives or GNU line markers depending on what mode we're in. 236 if (UseLineDirectives) { 237 OS << "#line" << ' ' << LineNo << ' ' << '"'; 238 OS.write_escaped(CurFilename); 239 OS << '"'; 240 } else { 241 OS << '#' << ' ' << LineNo << ' ' << '"'; 242 OS.write_escaped(CurFilename); 243 OS << '"'; 244 245 if (ExtraLen) 246 OS.write(Extra, ExtraLen); 247 248 if (FileType == SrcMgr::C_System) 249 OS.write(" 3", 2); 250 else if (FileType == SrcMgr::C_ExternCSystem) 251 OS.write(" 3 4", 4); 252 } 253 OS << '\n'; 254 } 255 256 /// MoveToLine - Move the output to the source line specified by the location 257 /// object. We can do this by emitting some number of \n's, or be emitting a 258 /// #line directive. This returns false if already at the specified line, true 259 /// if some newlines were emitted. 260 bool PrintPPOutputPPCallbacks::MoveToLine(unsigned LineNo, 261 bool RequireStartOfLine) { 262 // If it is required to start a new line or finish the current, insert 263 // vertical whitespace now and take it into account when moving to the 264 // expected line. 265 bool StartedNewLine = false; 266 if ((RequireStartOfLine && EmittedTokensOnThisLine) || 267 EmittedDirectiveOnThisLine) { 268 OS << '\n'; 269 StartedNewLine = true; 270 CurLine += 1; 271 EmittedTokensOnThisLine = false; 272 EmittedDirectiveOnThisLine = false; 273 } 274 275 // If this line is "close enough" to the original line, just print newlines, 276 // otherwise print a #line directive. 277 if (CurLine == LineNo) { 278 // Nothing to do if we are already on the correct line. 279 } else if (MinimizeWhitespace && DisableLineMarkers) { 280 // With -E -P -fminimize-whitespace, don't emit anything if not necessary. 281 } else if (!StartedNewLine && LineNo - CurLine == 1) { 282 // Printing a single line has priority over printing a #line directive, even 283 // when minimizing whitespace which otherwise would print #line directives 284 // for every single line. 285 OS << '\n'; 286 StartedNewLine = true; 287 } else if (!DisableLineMarkers) { 288 if (LineNo - CurLine <= 8) { 289 const char *NewLines = "\n\n\n\n\n\n\n\n"; 290 OS.write(NewLines, LineNo - CurLine); 291 } else { 292 // Emit a #line or line marker. 293 WriteLineInfo(LineNo, nullptr, 0); 294 } 295 StartedNewLine = true; 296 } else if (!StartedNewLine) { 297 // If we are not on the correct line and don't need to be line-correct, 298 // at least ensure we start on a new line. 299 OS << '\n'; 300 StartedNewLine = true; 301 } 302 303 if (StartedNewLine) { 304 EmittedTokensOnThisLine = false; 305 EmittedDirectiveOnThisLine = false; 306 } 307 308 CurLine = LineNo; 309 return StartedNewLine; 310 } 311 312 void PrintPPOutputPPCallbacks::startNewLineIfNeeded() { 313 if (EmittedTokensOnThisLine || EmittedDirectiveOnThisLine) { 314 OS << '\n'; 315 EmittedTokensOnThisLine = false; 316 EmittedDirectiveOnThisLine = false; 317 } 318 } 319 320 /// FileChanged - Whenever the preprocessor enters or exits a #include file 321 /// it invokes this handler. Update our conception of the current source 322 /// position. 323 void PrintPPOutputPPCallbacks::FileChanged(SourceLocation Loc, 324 FileChangeReason Reason, 325 SrcMgr::CharacteristicKind NewFileType, 326 FileID PrevFID) { 327 // Unless we are exiting a #include, make sure to skip ahead to the line the 328 // #include directive was at. 329 SourceManager &SourceMgr = SM; 330 331 PresumedLoc UserLoc = SourceMgr.getPresumedLoc(Loc); 332 if (UserLoc.isInvalid()) 333 return; 334 335 unsigned NewLine = UserLoc.getLine(); 336 337 if (Reason == PPCallbacks::EnterFile) { 338 SourceLocation IncludeLoc = UserLoc.getIncludeLoc(); 339 if (IncludeLoc.isValid()) 340 MoveToLine(IncludeLoc, /*RequireStartOfLine=*/false); 341 } else if (Reason == PPCallbacks::SystemHeaderPragma) { 342 // GCC emits the # directive for this directive on the line AFTER the 343 // directive and emits a bunch of spaces that aren't needed. This is because 344 // otherwise we will emit a line marker for THIS line, which requires an 345 // extra blank line after the directive to avoid making all following lines 346 // off by one. We can do better by simply incrementing NewLine here. 347 NewLine += 1; 348 } 349 350 CurLine = NewLine; 351 352 CurFilename.clear(); 353 CurFilename += UserLoc.getFilename(); 354 FileType = NewFileType; 355 356 if (DisableLineMarkers) { 357 if (!MinimizeWhitespace) 358 startNewLineIfNeeded(); 359 return; 360 } 361 362 if (!Initialized) { 363 WriteLineInfo(CurLine); 364 Initialized = true; 365 } 366 367 // Do not emit an enter marker for the main file (which we expect is the first 368 // entered file). This matches gcc, and improves compatibility with some tools 369 // which track the # line markers as a way to determine when the preprocessed 370 // output is in the context of the main file. 371 if (Reason == PPCallbacks::EnterFile && !IsFirstFileEntered) { 372 IsFirstFileEntered = true; 373 return; 374 } 375 376 switch (Reason) { 377 case PPCallbacks::EnterFile: 378 WriteLineInfo(CurLine, " 1", 2); 379 break; 380 case PPCallbacks::ExitFile: 381 WriteLineInfo(CurLine, " 2", 2); 382 break; 383 case PPCallbacks::SystemHeaderPragma: 384 case PPCallbacks::RenameFile: 385 WriteLineInfo(CurLine); 386 break; 387 } 388 } 389 390 void PrintPPOutputPPCallbacks::InclusionDirective( 391 SourceLocation HashLoc, 392 const Token &IncludeTok, 393 StringRef FileName, 394 bool IsAngled, 395 CharSourceRange FilenameRange, 396 const FileEntry *File, 397 StringRef SearchPath, 398 StringRef RelativePath, 399 const Module *Imported, 400 SrcMgr::CharacteristicKind FileType) { 401 // In -dI mode, dump #include directives prior to dumping their content or 402 // interpretation. 403 if (DumpIncludeDirectives) { 404 MoveToLine(HashLoc, /*RequireStartOfLine=*/true); 405 const std::string TokenText = PP.getSpelling(IncludeTok); 406 assert(!TokenText.empty()); 407 OS << "#" << TokenText << " " 408 << (IsAngled ? '<' : '"') << FileName << (IsAngled ? '>' : '"') 409 << " /* clang -E -dI */"; 410 setEmittedDirectiveOnThisLine(); 411 } 412 413 // When preprocessing, turn implicit imports into module import pragmas. 414 if (Imported) { 415 switch (IncludeTok.getIdentifierInfo()->getPPKeywordID()) { 416 case tok::pp_include: 417 case tok::pp_import: 418 case tok::pp_include_next: 419 MoveToLine(HashLoc, /*RequireStartOfLine=*/true); 420 OS << "#pragma clang module import " << Imported->getFullModuleName(true) 421 << " /* clang -E: implicit import for " 422 << "#" << PP.getSpelling(IncludeTok) << " " 423 << (IsAngled ? '<' : '"') << FileName << (IsAngled ? '>' : '"') 424 << " */"; 425 setEmittedDirectiveOnThisLine(); 426 break; 427 428 case tok::pp___include_macros: 429 // #__include_macros has no effect on a user of a preprocessed source 430 // file; the only effect is on preprocessing. 431 // 432 // FIXME: That's not *quite* true: it causes the module in question to 433 // be loaded, which can affect downstream diagnostics. 434 break; 435 436 default: 437 llvm_unreachable("unknown include directive kind"); 438 break; 439 } 440 } 441 } 442 443 /// Handle entering the scope of a module during a module compilation. 444 void PrintPPOutputPPCallbacks::BeginModule(const Module *M) { 445 startNewLineIfNeeded(); 446 OS << "#pragma clang module begin " << M->getFullModuleName(true); 447 setEmittedDirectiveOnThisLine(); 448 } 449 450 /// Handle leaving the scope of a module during a module compilation. 451 void PrintPPOutputPPCallbacks::EndModule(const Module *M) { 452 startNewLineIfNeeded(); 453 OS << "#pragma clang module end /*" << M->getFullModuleName(true) << "*/"; 454 setEmittedDirectiveOnThisLine(); 455 } 456 457 /// Ident - Handle #ident directives when read by the preprocessor. 458 /// 459 void PrintPPOutputPPCallbacks::Ident(SourceLocation Loc, StringRef S) { 460 MoveToLine(Loc, /*RequireStartOfLine=*/true); 461 462 OS.write("#ident ", strlen("#ident ")); 463 OS.write(S.begin(), S.size()); 464 setEmittedTokensOnThisLine(); 465 } 466 467 /// MacroDefined - This hook is called whenever a macro definition is seen. 468 void PrintPPOutputPPCallbacks::MacroDefined(const Token &MacroNameTok, 469 const MacroDirective *MD) { 470 const MacroInfo *MI = MD->getMacroInfo(); 471 // Only print out macro definitions in -dD mode. 472 if (!DumpDefines || 473 // Ignore __FILE__ etc. 474 MI->isBuiltinMacro()) return; 475 476 MoveToLine(MI->getDefinitionLoc(), /*RequireStartOfLine=*/true); 477 PrintMacroDefinition(*MacroNameTok.getIdentifierInfo(), *MI, PP, OS); 478 setEmittedDirectiveOnThisLine(); 479 } 480 481 void PrintPPOutputPPCallbacks::MacroUndefined(const Token &MacroNameTok, 482 const MacroDefinition &MD, 483 const MacroDirective *Undef) { 484 // Only print out macro definitions in -dD mode. 485 if (!DumpDefines) return; 486 487 MoveToLine(MacroNameTok.getLocation(), /*RequireStartOfLine=*/true); 488 OS << "#undef " << MacroNameTok.getIdentifierInfo()->getName(); 489 setEmittedDirectiveOnThisLine(); 490 } 491 492 static void outputPrintable(raw_ostream &OS, StringRef Str) { 493 for (unsigned char Char : Str) { 494 if (isPrintable(Char) && Char != '\\' && Char != '"') 495 OS << (char)Char; 496 else // Output anything hard as an octal escape. 497 OS << '\\' 498 << (char)('0' + ((Char >> 6) & 7)) 499 << (char)('0' + ((Char >> 3) & 7)) 500 << (char)('0' + ((Char >> 0) & 7)); 501 } 502 } 503 504 void PrintPPOutputPPCallbacks::PragmaMessage(SourceLocation Loc, 505 StringRef Namespace, 506 PragmaMessageKind Kind, 507 StringRef Str) { 508 MoveToLine(Loc, /*RequireStartOfLine=*/true); 509 OS << "#pragma "; 510 if (!Namespace.empty()) 511 OS << Namespace << ' '; 512 switch (Kind) { 513 case PMK_Message: 514 OS << "message(\""; 515 break; 516 case PMK_Warning: 517 OS << "warning \""; 518 break; 519 case PMK_Error: 520 OS << "error \""; 521 break; 522 } 523 524 outputPrintable(OS, Str); 525 OS << '"'; 526 if (Kind == PMK_Message) 527 OS << ')'; 528 setEmittedDirectiveOnThisLine(); 529 } 530 531 void PrintPPOutputPPCallbacks::PragmaDebug(SourceLocation Loc, 532 StringRef DebugType) { 533 MoveToLine(Loc, /*RequireStartOfLine=*/true); 534 535 OS << "#pragma clang __debug "; 536 OS << DebugType; 537 538 setEmittedDirectiveOnThisLine(); 539 } 540 541 void PrintPPOutputPPCallbacks:: 542 PragmaDiagnosticPush(SourceLocation Loc, StringRef Namespace) { 543 MoveToLine(Loc, /*RequireStartOfLine=*/true); 544 OS << "#pragma " << Namespace << " diagnostic push"; 545 setEmittedDirectiveOnThisLine(); 546 } 547 548 void PrintPPOutputPPCallbacks:: 549 PragmaDiagnosticPop(SourceLocation Loc, StringRef Namespace) { 550 MoveToLine(Loc, /*RequireStartOfLine=*/true); 551 OS << "#pragma " << Namespace << " diagnostic pop"; 552 setEmittedDirectiveOnThisLine(); 553 } 554 555 void PrintPPOutputPPCallbacks::PragmaDiagnostic(SourceLocation Loc, 556 StringRef Namespace, 557 diag::Severity Map, 558 StringRef Str) { 559 MoveToLine(Loc, /*RequireStartOfLine=*/true); 560 OS << "#pragma " << Namespace << " diagnostic "; 561 switch (Map) { 562 case diag::Severity::Remark: 563 OS << "remark"; 564 break; 565 case diag::Severity::Warning: 566 OS << "warning"; 567 break; 568 case diag::Severity::Error: 569 OS << "error"; 570 break; 571 case diag::Severity::Ignored: 572 OS << "ignored"; 573 break; 574 case diag::Severity::Fatal: 575 OS << "fatal"; 576 break; 577 } 578 OS << " \"" << Str << '"'; 579 setEmittedDirectiveOnThisLine(); 580 } 581 582 void PrintPPOutputPPCallbacks::PragmaWarning(SourceLocation Loc, 583 StringRef WarningSpec, 584 ArrayRef<int> Ids) { 585 MoveToLine(Loc, /*RequireStartOfLine=*/true); 586 OS << "#pragma warning(" << WarningSpec << ':'; 587 for (ArrayRef<int>::iterator I = Ids.begin(), E = Ids.end(); I != E; ++I) 588 OS << ' ' << *I; 589 OS << ')'; 590 setEmittedDirectiveOnThisLine(); 591 } 592 593 void PrintPPOutputPPCallbacks::PragmaWarningPush(SourceLocation Loc, 594 int Level) { 595 MoveToLine(Loc, /*RequireStartOfLine=*/true); 596 OS << "#pragma warning(push"; 597 if (Level >= 0) 598 OS << ", " << Level; 599 OS << ')'; 600 setEmittedDirectiveOnThisLine(); 601 } 602 603 void PrintPPOutputPPCallbacks::PragmaWarningPop(SourceLocation Loc) { 604 MoveToLine(Loc, /*RequireStartOfLine=*/true); 605 OS << "#pragma warning(pop)"; 606 setEmittedDirectiveOnThisLine(); 607 } 608 609 void PrintPPOutputPPCallbacks::PragmaExecCharsetPush(SourceLocation Loc, 610 StringRef Str) { 611 MoveToLine(Loc, /*RequireStartOfLine=*/true); 612 OS << "#pragma character_execution_set(push"; 613 if (!Str.empty()) 614 OS << ", " << Str; 615 OS << ')'; 616 setEmittedDirectiveOnThisLine(); 617 } 618 619 void PrintPPOutputPPCallbacks::PragmaExecCharsetPop(SourceLocation Loc) { 620 MoveToLine(Loc, /*RequireStartOfLine=*/true); 621 OS << "#pragma character_execution_set(pop)"; 622 setEmittedDirectiveOnThisLine(); 623 } 624 625 void PrintPPOutputPPCallbacks:: 626 PragmaAssumeNonNullBegin(SourceLocation Loc) { 627 MoveToLine(Loc, /*RequireStartOfLine=*/true); 628 OS << "#pragma clang assume_nonnull begin"; 629 setEmittedDirectiveOnThisLine(); 630 } 631 632 void PrintPPOutputPPCallbacks:: 633 PragmaAssumeNonNullEnd(SourceLocation Loc) { 634 MoveToLine(Loc, /*RequireStartOfLine=*/true); 635 OS << "#pragma clang assume_nonnull end"; 636 setEmittedDirectiveOnThisLine(); 637 } 638 639 void PrintPPOutputPPCallbacks::HandleWhitespaceBeforeTok(const Token &Tok, 640 bool RequireSpace, 641 bool RequireSameLine) { 642 // These tokens are not expanded to anything and don't need whitespace before 643 // them. 644 if (Tok.is(tok::eof) || 645 (Tok.isAnnotation() && !Tok.is(tok::annot_header_unit) && 646 !Tok.is(tok::annot_module_begin) && !Tok.is(tok::annot_module_end))) 647 return; 648 649 if (!RequireSameLine && MoveToLine(Tok, /*RequireStartOfLine=*/false)) { 650 if (MinimizeWhitespace) { 651 // Avoid interpreting hash as a directive under -fpreprocessed. 652 if (Tok.is(tok::hash)) 653 OS << ' '; 654 } else { 655 // Print out space characters so that the first token on a line is 656 // indented for easy reading. 657 unsigned ColNo = SM.getExpansionColumnNumber(Tok.getLocation()); 658 659 // The first token on a line can have a column number of 1, yet still 660 // expect leading white space, if a macro expansion in column 1 starts 661 // with an empty macro argument, or an empty nested macro expansion. In 662 // this case, move the token to column 2. 663 if (ColNo == 1 && Tok.hasLeadingSpace()) 664 ColNo = 2; 665 666 // This hack prevents stuff like: 667 // #define HASH # 668 // HASH define foo bar 669 // From having the # character end up at column 1, which makes it so it 670 // is not handled as a #define next time through the preprocessor if in 671 // -fpreprocessed mode. 672 if (ColNo <= 1 && Tok.is(tok::hash)) 673 OS << ' '; 674 675 // Otherwise, indent the appropriate number of spaces. 676 for (; ColNo > 1; --ColNo) 677 OS << ' '; 678 } 679 } else { 680 // Insert whitespace between the previous and next token if either 681 // - The caller requires it 682 // - The input had whitespace between them and we are not in 683 // whitespace-minimization mode 684 // - The whitespace is necessary to keep the tokens apart and there is not 685 // already a newline between them 686 if (RequireSpace || (!MinimizeWhitespace && Tok.hasLeadingSpace()) || 687 ((EmittedTokensOnThisLine || EmittedTokensOnThisLine) && 688 AvoidConcat(PrevPrevTok, PrevTok, Tok))) 689 OS << ' '; 690 } 691 692 PrevPrevTok = PrevTok; 693 PrevTok = Tok; 694 } 695 696 void PrintPPOutputPPCallbacks::HandleNewlinesInToken(const char *TokStr, 697 unsigned Len) { 698 unsigned NumNewlines = 0; 699 for (; Len; --Len, ++TokStr) { 700 if (*TokStr != '\n' && 701 *TokStr != '\r') 702 continue; 703 704 ++NumNewlines; 705 706 // If we have \n\r or \r\n, skip both and count as one line. 707 if (Len != 1 && 708 (TokStr[1] == '\n' || TokStr[1] == '\r') && 709 TokStr[0] != TokStr[1]) { 710 ++TokStr; 711 --Len; 712 } 713 } 714 715 if (NumNewlines == 0) return; 716 717 CurLine += NumNewlines; 718 } 719 720 721 namespace { 722 struct UnknownPragmaHandler : public PragmaHandler { 723 const char *Prefix; 724 PrintPPOutputPPCallbacks *Callbacks; 725 726 // Set to true if tokens should be expanded 727 bool ShouldExpandTokens; 728 729 UnknownPragmaHandler(const char *prefix, PrintPPOutputPPCallbacks *callbacks, 730 bool RequireTokenExpansion) 731 : Prefix(prefix), Callbacks(callbacks), 732 ShouldExpandTokens(RequireTokenExpansion) {} 733 void HandlePragma(Preprocessor &PP, PragmaIntroducer Introducer, 734 Token &PragmaTok) override { 735 // Figure out what line we went to and insert the appropriate number of 736 // newline characters. 737 Callbacks->MoveToLine(PragmaTok.getLocation(), /*RequireStartOfLine=*/true); 738 Callbacks->OS.write(Prefix, strlen(Prefix)); 739 Callbacks->setEmittedTokensOnThisLine(); 740 741 if (ShouldExpandTokens) { 742 // The first token does not have expanded macros. Expand them, if 743 // required. 744 auto Toks = std::make_unique<Token[]>(1); 745 Toks[0] = PragmaTok; 746 PP.EnterTokenStream(std::move(Toks), /*NumToks=*/1, 747 /*DisableMacroExpansion=*/false, 748 /*IsReinject=*/false); 749 PP.Lex(PragmaTok); 750 } 751 752 // Read and print all of the pragma tokens. 753 bool IsFirst = true; 754 while (PragmaTok.isNot(tok::eod)) { 755 Callbacks->HandleWhitespaceBeforeTok(PragmaTok, /*RequireSpace=*/IsFirst, 756 /*RequireSameLine=*/true); 757 IsFirst = false; 758 std::string TokSpell = PP.getSpelling(PragmaTok); 759 Callbacks->OS.write(&TokSpell[0], TokSpell.size()); 760 Callbacks->setEmittedTokensOnThisLine(); 761 762 if (ShouldExpandTokens) 763 PP.Lex(PragmaTok); 764 else 765 PP.LexUnexpandedToken(PragmaTok); 766 } 767 Callbacks->setEmittedDirectiveOnThisLine(); 768 } 769 }; 770 } // end anonymous namespace 771 772 773 static void PrintPreprocessedTokens(Preprocessor &PP, Token &Tok, 774 PrintPPOutputPPCallbacks *Callbacks, 775 raw_ostream &OS) { 776 bool DropComments = PP.getLangOpts().TraditionalCPP && 777 !PP.getCommentRetentionState(); 778 779 bool IsStartOfLine = false; 780 char Buffer[256]; 781 while (1) { 782 // Two lines joined with line continuation ('\' as last character on the 783 // line) must be emitted as one line even though Tok.getLine() returns two 784 // different values. In this situation Tok.isAtStartOfLine() is false even 785 // though it may be the first token on the lexical line. When 786 // dropping/skipping a token that is at the start of a line, propagate the 787 // start-of-line-ness to the next token to not append it to the previous 788 // line. 789 IsStartOfLine = IsStartOfLine || Tok.isAtStartOfLine(); 790 791 Callbacks->HandleWhitespaceBeforeTok(Tok, /*RequireSpace=*/false, 792 /*RequireSameLine=*/!IsStartOfLine); 793 794 if (DropComments && Tok.is(tok::comment)) { 795 // Skip comments. Normally the preprocessor does not generate 796 // tok::comment nodes at all when not keeping comments, but under 797 // -traditional-cpp the lexer keeps /all/ whitespace, including comments. 798 PP.Lex(Tok); 799 continue; 800 } else if (Tok.is(tok::eod)) { 801 // Don't print end of directive tokens, since they are typically newlines 802 // that mess up our line tracking. These come from unknown pre-processor 803 // directives or hash-prefixed comments in standalone assembly files. 804 PP.Lex(Tok); 805 // FIXME: The token on the next line after #include should have 806 // Tok.isAtStartOfLine() set. 807 IsStartOfLine = true; 808 continue; 809 } else if (Tok.is(tok::annot_module_include)) { 810 // PrintPPOutputPPCallbacks::InclusionDirective handles producing 811 // appropriate output here. Ignore this token entirely. 812 PP.Lex(Tok); 813 IsStartOfLine = true; 814 continue; 815 } else if (Tok.is(tok::annot_module_begin)) { 816 // FIXME: We retrieve this token after the FileChanged callback, and 817 // retrieve the module_end token before the FileChanged callback, so 818 // we render this within the file and render the module end outside the 819 // file, but this is backwards from the token locations: the module_begin 820 // token is at the include location (outside the file) and the module_end 821 // token is at the EOF location (within the file). 822 Callbacks->BeginModule( 823 reinterpret_cast<Module *>(Tok.getAnnotationValue())); 824 PP.Lex(Tok); 825 IsStartOfLine = true; 826 continue; 827 } else if (Tok.is(tok::annot_module_end)) { 828 Callbacks->EndModule( 829 reinterpret_cast<Module *>(Tok.getAnnotationValue())); 830 PP.Lex(Tok); 831 IsStartOfLine = true; 832 continue; 833 } else if (Tok.is(tok::annot_header_unit)) { 834 // This is a header-name that has been (effectively) converted into a 835 // module-name. 836 // FIXME: The module name could contain non-identifier module name 837 // components. We don't have a good way to round-trip those. 838 Module *M = reinterpret_cast<Module *>(Tok.getAnnotationValue()); 839 std::string Name = M->getFullModuleName(); 840 OS.write(Name.data(), Name.size()); 841 Callbacks->HandleNewlinesInToken(Name.data(), Name.size()); 842 } else if (Tok.isAnnotation()) { 843 // Ignore annotation tokens created by pragmas - the pragmas themselves 844 // will be reproduced in the preprocessed output. 845 PP.Lex(Tok); 846 continue; 847 } else if (IdentifierInfo *II = Tok.getIdentifierInfo()) { 848 OS << II->getName(); 849 } else if (Tok.isLiteral() && !Tok.needsCleaning() && 850 Tok.getLiteralData()) { 851 OS.write(Tok.getLiteralData(), Tok.getLength()); 852 } else if (Tok.getLength() < llvm::array_lengthof(Buffer)) { 853 const char *TokPtr = Buffer; 854 unsigned Len = PP.getSpelling(Tok, TokPtr); 855 OS.write(TokPtr, Len); 856 857 // Tokens that can contain embedded newlines need to adjust our current 858 // line number. 859 // FIXME: The token may end with a newline in which case 860 // setEmittedDirectiveOnThisLine/setEmittedTokensOnThisLine afterwards is 861 // wrong. 862 if (Tok.getKind() == tok::comment || Tok.getKind() == tok::unknown) 863 Callbacks->HandleNewlinesInToken(TokPtr, Len); 864 if (Tok.is(tok::comment) && Len >= 2 && TokPtr[0] == '/' && 865 TokPtr[1] == '/') { 866 // It's a line comment; 867 // Ensure that we don't concatenate anything behind it. 868 Callbacks->setEmittedDirectiveOnThisLine(); 869 } 870 } else { 871 std::string S = PP.getSpelling(Tok); 872 OS.write(S.data(), S.size()); 873 874 // Tokens that can contain embedded newlines need to adjust our current 875 // line number. 876 if (Tok.getKind() == tok::comment || Tok.getKind() == tok::unknown) 877 Callbacks->HandleNewlinesInToken(S.data(), S.size()); 878 if (Tok.is(tok::comment) && S.size() >= 2 && S[0] == '/' && S[1] == '/') { 879 // It's a line comment; 880 // Ensure that we don't concatenate anything behind it. 881 Callbacks->setEmittedDirectiveOnThisLine(); 882 } 883 } 884 Callbacks->setEmittedTokensOnThisLine(); 885 IsStartOfLine = false; 886 887 if (Tok.is(tok::eof)) break; 888 889 PP.Lex(Tok); 890 } 891 } 892 893 typedef std::pair<const IdentifierInfo *, MacroInfo *> id_macro_pair; 894 static int MacroIDCompare(const id_macro_pair *LHS, const id_macro_pair *RHS) { 895 return LHS->first->getName().compare(RHS->first->getName()); 896 } 897 898 static void DoPrintMacros(Preprocessor &PP, raw_ostream *OS) { 899 // Ignore unknown pragmas. 900 PP.IgnorePragmas(); 901 902 // -dM mode just scans and ignores all tokens in the files, then dumps out 903 // the macro table at the end. 904 PP.EnterMainSourceFile(); 905 906 Token Tok; 907 do PP.Lex(Tok); 908 while (Tok.isNot(tok::eof)); 909 910 SmallVector<id_macro_pair, 128> MacrosByID; 911 for (Preprocessor::macro_iterator I = PP.macro_begin(), E = PP.macro_end(); 912 I != E; ++I) { 913 auto *MD = I->second.getLatest(); 914 if (MD && MD->isDefined()) 915 MacrosByID.push_back(id_macro_pair(I->first, MD->getMacroInfo())); 916 } 917 llvm::array_pod_sort(MacrosByID.begin(), MacrosByID.end(), MacroIDCompare); 918 919 for (unsigned i = 0, e = MacrosByID.size(); i != e; ++i) { 920 MacroInfo &MI = *MacrosByID[i].second; 921 // Ignore computed macros like __LINE__ and friends. 922 if (MI.isBuiltinMacro()) continue; 923 924 PrintMacroDefinition(*MacrosByID[i].first, MI, PP, *OS); 925 *OS << '\n'; 926 } 927 } 928 929 /// DoPrintPreprocessedInput - This implements -E mode. 930 /// 931 void clang::DoPrintPreprocessedInput(Preprocessor &PP, raw_ostream *OS, 932 const PreprocessorOutputOptions &Opts) { 933 // Show macros with no output is handled specially. 934 if (!Opts.ShowCPP) { 935 assert(Opts.ShowMacros && "Not yet implemented!"); 936 DoPrintMacros(PP, OS); 937 return; 938 } 939 940 // Inform the preprocessor whether we want it to retain comments or not, due 941 // to -C or -CC. 942 PP.SetCommentRetentionState(Opts.ShowComments, Opts.ShowMacroComments); 943 944 PrintPPOutputPPCallbacks *Callbacks = new PrintPPOutputPPCallbacks( 945 PP, *OS, !Opts.ShowLineMarkers, Opts.ShowMacros, 946 Opts.ShowIncludeDirectives, Opts.UseLineDirectives, 947 Opts.MinimizeWhitespace); 948 949 // Expand macros in pragmas with -fms-extensions. The assumption is that 950 // the majority of pragmas in such a file will be Microsoft pragmas. 951 // Remember the handlers we will add so that we can remove them later. 952 std::unique_ptr<UnknownPragmaHandler> MicrosoftExtHandler( 953 new UnknownPragmaHandler( 954 "#pragma", Callbacks, 955 /*RequireTokenExpansion=*/PP.getLangOpts().MicrosoftExt)); 956 957 std::unique_ptr<UnknownPragmaHandler> GCCHandler(new UnknownPragmaHandler( 958 "#pragma GCC", Callbacks, 959 /*RequireTokenExpansion=*/PP.getLangOpts().MicrosoftExt)); 960 961 std::unique_ptr<UnknownPragmaHandler> ClangHandler(new UnknownPragmaHandler( 962 "#pragma clang", Callbacks, 963 /*RequireTokenExpansion=*/PP.getLangOpts().MicrosoftExt)); 964 965 PP.AddPragmaHandler(MicrosoftExtHandler.get()); 966 PP.AddPragmaHandler("GCC", GCCHandler.get()); 967 PP.AddPragmaHandler("clang", ClangHandler.get()); 968 969 // The tokens after pragma omp need to be expanded. 970 // 971 // OpenMP [2.1, Directive format] 972 // Preprocessing tokens following the #pragma omp are subject to macro 973 // replacement. 974 std::unique_ptr<UnknownPragmaHandler> OpenMPHandler( 975 new UnknownPragmaHandler("#pragma omp", Callbacks, 976 /*RequireTokenExpansion=*/true)); 977 PP.AddPragmaHandler("omp", OpenMPHandler.get()); 978 979 PP.addPPCallbacks(std::unique_ptr<PPCallbacks>(Callbacks)); 980 981 // After we have configured the preprocessor, enter the main file. 982 PP.EnterMainSourceFile(); 983 984 // Consume all of the tokens that come from the predefines buffer. Those 985 // should not be emitted into the output and are guaranteed to be at the 986 // start. 987 const SourceManager &SourceMgr = PP.getSourceManager(); 988 Token Tok; 989 do { 990 PP.Lex(Tok); 991 if (Tok.is(tok::eof) || !Tok.getLocation().isFileID()) 992 break; 993 994 PresumedLoc PLoc = SourceMgr.getPresumedLoc(Tok.getLocation()); 995 if (PLoc.isInvalid()) 996 break; 997 998 if (strcmp(PLoc.getFilename(), "<built-in>")) 999 break; 1000 } while (true); 1001 1002 // Read all the preprocessed tokens, printing them out to the stream. 1003 PrintPreprocessedTokens(PP, Tok, Callbacks, *OS); 1004 *OS << '\n'; 1005 1006 // Remove the handlers we just added to leave the preprocessor in a sane state 1007 // so that it can be reused (for example by a clang::Parser instance). 1008 PP.RemovePragmaHandler(MicrosoftExtHandler.get()); 1009 PP.RemovePragmaHandler("GCC", GCCHandler.get()); 1010 PP.RemovePragmaHandler("clang", ClangHandler.get()); 1011 PP.RemovePragmaHandler("omp", OpenMPHandler.get()); 1012 } 1013