| 1 | //===--- InclusionRewriter.cpp - Rewrite includes into their expansions ---===// |
| 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 rewrites include invocations into their expansions. This gives you |
| 10 | // a file with all included files merged into it. |
| 11 | // |
| 12 | //===----------------------------------------------------------------------===// |
| 13 | |
| 14 | #include "clang/Rewrite/Frontend/Rewriters.h" |
| 15 | #include "clang/Basic/SourceManager.h" |
| 16 | #include "clang/Frontend/PreprocessorOutputOptions.h" |
| 17 | #include "clang/Lex/Pragma.h" |
| 18 | #include "clang/Lex/Preprocessor.h" |
| 19 | #include "llvm/ADT/SmallString.h" |
| 20 | #include "llvm/Support/raw_ostream.h" |
| 21 | #include <optional> |
| 22 | |
| 23 | using namespace clang; |
| 24 | using namespace llvm; |
| 25 | |
| 26 | namespace { |
| 27 | |
| 28 | class InclusionRewriter : public PPCallbacks { |
| 29 | /// Information about which #includes were actually performed, |
| 30 | /// created by preprocessor callbacks. |
| 31 | struct IncludedFile { |
| 32 | FileID Id; |
| 33 | SrcMgr::CharacteristicKind FileType; |
| 34 | IncludedFile(FileID Id, SrcMgr::CharacteristicKind FileType) |
| 35 | : Id(Id), FileType(FileType) {} |
| 36 | }; |
| 37 | Preprocessor &PP; ///< Used to find inclusion directives. |
| 38 | SourceManager &SM; ///< Used to read and manage source files. |
| 39 | raw_ostream &OS; ///< The destination stream for rewritten contents. |
| 40 | StringRef MainEOL; ///< The line ending marker to use. |
| 41 | llvm::MemoryBufferRef PredefinesBuffer; ///< The preprocessor predefines. |
| 42 | bool ShowLineMarkers; ///< Show #line markers. |
| 43 | bool UseLineDirectives; ///< Use of line directives or line markers. |
| 44 | /// Tracks where inclusions that change the file are found. |
| 45 | std::map<SourceLocation, IncludedFile> FileIncludes; |
| 46 | /// Tracks where inclusions that import modules are found. |
| 47 | std::map<SourceLocation, const Module *> ModuleIncludes; |
| 48 | /// Tracks where inclusions that enter modules (in a module build) are found. |
| 49 | std::map<SourceLocation, const Module *> ModuleEntryIncludes; |
| 50 | /// Tracks where #if and #elif directives get evaluated and whether to true. |
| 51 | std::map<SourceLocation, bool> IfConditions; |
| 52 | /// Used transitively for building up the FileIncludes mapping over the |
| 53 | /// various \c PPCallbacks callbacks. |
| 54 | SourceLocation LastInclusionLocation; |
| 55 | public: |
| 56 | InclusionRewriter(Preprocessor &PP, raw_ostream &OS, bool ShowLineMarkers, |
| 57 | bool UseLineDirectives); |
| 58 | void Process(FileID FileId, SrcMgr::CharacteristicKind FileType); |
| 59 | void setPredefinesBuffer(const llvm::MemoryBufferRef &Buf) { |
| 60 | PredefinesBuffer = Buf; |
| 61 | } |
| 62 | void detectMainFileEOL(); |
| 63 | void handleModuleBegin(Token &Tok) { |
| 64 | assert(Tok.getKind() == tok::annot_module_begin); |
| 65 | ModuleEntryIncludes.insert( |
| 66 | x: {Tok.getLocation(), (Module *)Tok.getAnnotationValue()}); |
| 67 | } |
| 68 | private: |
| 69 | void FileChanged(SourceLocation Loc, FileChangeReason Reason, |
| 70 | SrcMgr::CharacteristicKind FileType, |
| 71 | FileID PrevFID) override; |
| 72 | void FileSkipped(const FileEntryRef &SkippedFile, const Token &FilenameTok, |
| 73 | SrcMgr::CharacteristicKind FileType) override; |
| 74 | void InclusionDirective(SourceLocation HashLoc, const Token &IncludeTok, |
| 75 | StringRef FileName, bool IsAngled, |
| 76 | CharSourceRange FilenameRange, |
| 77 | OptionalFileEntryRef File, StringRef SearchPath, |
| 78 | StringRef RelativePath, const Module *SuggestedModule, |
| 79 | bool ModuleImported, |
| 80 | SrcMgr::CharacteristicKind FileType) override; |
| 81 | void If(SourceLocation Loc, SourceRange ConditionRange, |
| 82 | ConditionValueKind ConditionValue) override; |
| 83 | void Elif(SourceLocation Loc, SourceRange ConditionRange, |
| 84 | ConditionValueKind ConditionValue, SourceLocation IfLoc) override; |
| 85 | void WriteLineInfo(StringRef Filename, int Line, |
| 86 | SrcMgr::CharacteristicKind FileType, |
| 87 | StringRef = StringRef()); |
| 88 | void WriteImplicitModuleImport(const Module *Mod); |
| 89 | void OutputContentUpTo(const MemoryBufferRef &FromFile, unsigned &WriteFrom, |
| 90 | unsigned WriteTo, StringRef EOL, int &lines, |
| 91 | bool EnsureNewline); |
| 92 | void CommentOutDirective(Lexer &DirectivesLex, const Token &StartToken, |
| 93 | const MemoryBufferRef &FromFile, StringRef EOL, |
| 94 | unsigned &NextToWrite, int &Lines, |
| 95 | const IncludedFile *Inc = nullptr); |
| 96 | const IncludedFile *FindIncludeAtLocation(SourceLocation Loc) const; |
| 97 | StringRef getIncludedFileName(const IncludedFile *Inc) const; |
| 98 | const Module *FindModuleAtLocation(SourceLocation Loc) const; |
| 99 | const Module *FindEnteredModule(SourceLocation Loc) const; |
| 100 | bool IsIfAtLocationTrue(SourceLocation Loc) const; |
| 101 | StringRef NextIdentifierName(Lexer &RawLex, Token &RawToken); |
| 102 | }; |
| 103 | |
| 104 | } // end anonymous namespace |
| 105 | |
| 106 | /// Initializes an InclusionRewriter with a \p PP source and \p OS destination. |
| 107 | InclusionRewriter::InclusionRewriter(Preprocessor &PP, raw_ostream &OS, |
| 108 | bool ShowLineMarkers, |
| 109 | bool UseLineDirectives) |
| 110 | : PP(PP), SM(PP.getSourceManager()), OS(OS), MainEOL("\n" ), |
| 111 | ShowLineMarkers(ShowLineMarkers), UseLineDirectives(UseLineDirectives), |
| 112 | LastInclusionLocation(SourceLocation()) {} |
| 113 | |
| 114 | /// Write appropriate line information as either #line directives or GNU line |
| 115 | /// markers depending on what mode we're in, including the \p Filename and |
| 116 | /// \p Line we are located at, using the specified \p EOL line separator, and |
| 117 | /// any \p Extra context specifiers in GNU line directives. |
| 118 | void InclusionRewriter::WriteLineInfo(StringRef Filename, int Line, |
| 119 | SrcMgr::CharacteristicKind FileType, |
| 120 | StringRef ) { |
| 121 | if (!ShowLineMarkers) |
| 122 | return; |
| 123 | if (UseLineDirectives) { |
| 124 | OS << "#line" << ' ' << Line << ' ' << '"'; |
| 125 | OS.write_escaped(Str: Filename); |
| 126 | OS << '"'; |
| 127 | } else { |
| 128 | // Use GNU linemarkers as described here: |
| 129 | // http://gcc.gnu.org/onlinedocs/cpp/Preprocessor-Output.html |
| 130 | OS << '#' << ' ' << Line << ' ' << '"'; |
| 131 | OS.write_escaped(Str: Filename); |
| 132 | OS << '"'; |
| 133 | if (!Extra.empty()) |
| 134 | OS << Extra; |
| 135 | if (FileType == SrcMgr::C_System) |
| 136 | // "`3' This indicates that the following text comes from a system header |
| 137 | // file, so certain warnings should be suppressed." |
| 138 | OS << " 3" ; |
| 139 | else if (FileType == SrcMgr::C_ExternCSystem) |
| 140 | // as above for `3', plus "`4' This indicates that the following text |
| 141 | // should be treated as being wrapped in an implicit extern "C" block." |
| 142 | OS << " 3 4" ; |
| 143 | } |
| 144 | OS << MainEOL; |
| 145 | } |
| 146 | |
| 147 | void InclusionRewriter::WriteImplicitModuleImport(const Module *Mod) { |
| 148 | OS << "#pragma clang module import " << Mod->getFullModuleName(AllowStringLiterals: true) |
| 149 | << " /* clang -frewrite-includes: implicit import */" << MainEOL; |
| 150 | } |
| 151 | |
| 152 | /// FileChanged - Whenever the preprocessor enters or exits a #include file |
| 153 | /// it invokes this handler. |
| 154 | void InclusionRewriter::FileChanged(SourceLocation Loc, |
| 155 | FileChangeReason Reason, |
| 156 | SrcMgr::CharacteristicKind NewFileType, |
| 157 | FileID) { |
| 158 | if (Reason != EnterFile) |
| 159 | return; |
| 160 | if (LastInclusionLocation.isInvalid()) |
| 161 | // we didn't reach this file (eg: the main file) via an inclusion directive |
| 162 | return; |
| 163 | FileID Id = FullSourceLoc(Loc, SM).getFileID(); |
| 164 | auto P = FileIncludes.insert( |
| 165 | x: std::make_pair(x&: LastInclusionLocation, y: IncludedFile(Id, NewFileType))); |
| 166 | (void)P; |
| 167 | assert(P.second && "Unexpected revisitation of the same include directive" ); |
| 168 | LastInclusionLocation = SourceLocation(); |
| 169 | } |
| 170 | |
| 171 | /// Called whenever an inclusion is skipped due to canonical header protection |
| 172 | /// macros. |
| 173 | void InclusionRewriter::FileSkipped(const FileEntryRef & /*SkippedFile*/, |
| 174 | const Token & /*FilenameTok*/, |
| 175 | SrcMgr::CharacteristicKind /*FileType*/) { |
| 176 | assert(LastInclusionLocation.isValid() && |
| 177 | "A file, that wasn't found via an inclusion directive, was skipped" ); |
| 178 | LastInclusionLocation = SourceLocation(); |
| 179 | } |
| 180 | |
| 181 | /// This should be called whenever the preprocessor encounters include |
| 182 | /// directives. It does not say whether the file has been included, but it |
| 183 | /// provides more information about the directive (hash location instead |
| 184 | /// of location inside the included file). It is assumed that the matching |
| 185 | /// FileChanged() or FileSkipped() is called after this (or neither is |
| 186 | /// called if this #include results in an error or does not textually include |
| 187 | /// anything). |
| 188 | void InclusionRewriter::InclusionDirective( |
| 189 | SourceLocation HashLoc, const Token & /*IncludeTok*/, |
| 190 | StringRef /*FileName*/, bool /*IsAngled*/, |
| 191 | CharSourceRange /*FilenameRange*/, OptionalFileEntryRef /*File*/, |
| 192 | StringRef /*SearchPath*/, StringRef /*RelativePath*/, |
| 193 | const Module *SuggestedModule, bool ModuleImported, |
| 194 | SrcMgr::CharacteristicKind FileType) { |
| 195 | if (ModuleImported) { |
| 196 | auto P = ModuleIncludes.insert(x: std::make_pair(x&: HashLoc, y&: SuggestedModule)); |
| 197 | (void)P; |
| 198 | assert(P.second && "Unexpected revisitation of the same include directive" ); |
| 199 | } else |
| 200 | LastInclusionLocation = HashLoc; |
| 201 | } |
| 202 | |
| 203 | void InclusionRewriter::If(SourceLocation Loc, SourceRange ConditionRange, |
| 204 | ConditionValueKind ConditionValue) { |
| 205 | auto P = IfConditions.insert(x: std::make_pair(x&: Loc, y: ConditionValue == CVK_True)); |
| 206 | (void)P; |
| 207 | assert(P.second && "Unexpected revisitation of the same if directive" ); |
| 208 | } |
| 209 | |
| 210 | void InclusionRewriter::Elif(SourceLocation Loc, SourceRange ConditionRange, |
| 211 | ConditionValueKind ConditionValue, |
| 212 | SourceLocation IfLoc) { |
| 213 | auto P = IfConditions.insert(x: std::make_pair(x&: Loc, y: ConditionValue == CVK_True)); |
| 214 | (void)P; |
| 215 | assert(P.second && "Unexpected revisitation of the same elif directive" ); |
| 216 | } |
| 217 | |
| 218 | /// Simple lookup for a SourceLocation (specifically one denoting the hash in |
| 219 | /// an inclusion directive) in the map of inclusion information, FileChanges. |
| 220 | const InclusionRewriter::IncludedFile * |
| 221 | InclusionRewriter::FindIncludeAtLocation(SourceLocation Loc) const { |
| 222 | const auto I = FileIncludes.find(x: Loc); |
| 223 | if (I != FileIncludes.end()) |
| 224 | return &I->second; |
| 225 | return nullptr; |
| 226 | } |
| 227 | |
| 228 | /// Simple lookup for a SourceLocation (specifically one denoting the hash in |
| 229 | /// an inclusion directive) in the map of module inclusion information. |
| 230 | const Module * |
| 231 | InclusionRewriter::FindModuleAtLocation(SourceLocation Loc) const { |
| 232 | const auto I = ModuleIncludes.find(x: Loc); |
| 233 | if (I != ModuleIncludes.end()) |
| 234 | return I->second; |
| 235 | return nullptr; |
| 236 | } |
| 237 | |
| 238 | /// Simple lookup for a SourceLocation (specifically one denoting the hash in |
| 239 | /// an inclusion directive) in the map of module entry information. |
| 240 | const Module * |
| 241 | InclusionRewriter::FindEnteredModule(SourceLocation Loc) const { |
| 242 | const auto I = ModuleEntryIncludes.find(x: Loc); |
| 243 | if (I != ModuleEntryIncludes.end()) |
| 244 | return I->second; |
| 245 | return nullptr; |
| 246 | } |
| 247 | |
| 248 | bool InclusionRewriter::IsIfAtLocationTrue(SourceLocation Loc) const { |
| 249 | const auto I = IfConditions.find(x: Loc); |
| 250 | if (I != IfConditions.end()) |
| 251 | return I->second; |
| 252 | return false; |
| 253 | } |
| 254 | |
| 255 | void InclusionRewriter::detectMainFileEOL() { |
| 256 | std::optional<MemoryBufferRef> FromFile = |
| 257 | *SM.getBufferOrNone(FID: SM.getMainFileID()); |
| 258 | assert(FromFile); |
| 259 | if (!FromFile) |
| 260 | return; // Should never happen, but whatever. |
| 261 | MainEOL = FromFile->getBuffer().detectEOL(); |
| 262 | } |
| 263 | |
| 264 | /// Writes out bytes from \p FromFile, starting at \p NextToWrite and ending at |
| 265 | /// \p WriteTo - 1. |
| 266 | void InclusionRewriter::OutputContentUpTo(const MemoryBufferRef &FromFile, |
| 267 | unsigned &WriteFrom, unsigned WriteTo, |
| 268 | StringRef LocalEOL, int &Line, |
| 269 | bool EnsureNewline) { |
| 270 | if (WriteTo <= WriteFrom) |
| 271 | return; |
| 272 | if (FromFile == PredefinesBuffer) { |
| 273 | // Ignore the #defines of the predefines buffer. |
| 274 | WriteFrom = WriteTo; |
| 275 | return; |
| 276 | } |
| 277 | |
| 278 | // If we would output half of a line ending, advance one character to output |
| 279 | // the whole line ending. All buffers are null terminated, so looking ahead |
| 280 | // one byte is safe. |
| 281 | if (LocalEOL.size() == 2 && |
| 282 | LocalEOL[0] == (FromFile.getBufferStart() + WriteTo)[-1] && |
| 283 | LocalEOL[1] == (FromFile.getBufferStart() + WriteTo)[0]) |
| 284 | WriteTo++; |
| 285 | |
| 286 | StringRef TextToWrite(FromFile.getBufferStart() + WriteFrom, |
| 287 | WriteTo - WriteFrom); |
| 288 | // count lines manually, it's faster than getPresumedLoc() |
| 289 | Line += TextToWrite.count(Str: LocalEOL); |
| 290 | |
| 291 | if (MainEOL == LocalEOL) { |
| 292 | OS << TextToWrite; |
| 293 | } else { |
| 294 | // Output the file one line at a time, rewriting the line endings as we go. |
| 295 | StringRef Rest = TextToWrite; |
| 296 | while (!Rest.empty()) { |
| 297 | // Identify and output the next line excluding an EOL sequence if present. |
| 298 | size_t Idx = Rest.find(Str: LocalEOL); |
| 299 | StringRef LineText = Rest.substr(Start: 0, N: Idx); |
| 300 | OS << LineText; |
| 301 | if (Idx != StringRef::npos) { |
| 302 | // An EOL sequence was present, output the EOL sequence for the |
| 303 | // main source file and skip past the local EOL sequence. |
| 304 | OS << MainEOL; |
| 305 | Idx += LocalEOL.size(); |
| 306 | } |
| 307 | // Strip the line just handled. If Idx is npos or matches the end of the |
| 308 | // text, Rest will be set to an empty string and the loop will terminate. |
| 309 | Rest = Rest.substr(Start: Idx); |
| 310 | } |
| 311 | } |
| 312 | if (EnsureNewline && !TextToWrite.ends_with(Suffix: LocalEOL)) |
| 313 | OS << MainEOL; |
| 314 | |
| 315 | WriteFrom = WriteTo; |
| 316 | } |
| 317 | |
| 318 | StringRef |
| 319 | InclusionRewriter::getIncludedFileName(const IncludedFile *Inc) const { |
| 320 | if (Inc) { |
| 321 | auto B = SM.getBufferOrNone(FID: Inc->Id); |
| 322 | assert(B && "Attempting to process invalid inclusion" ); |
| 323 | if (B) |
| 324 | return llvm::sys::path::filename(path: B->getBufferIdentifier()); |
| 325 | } |
| 326 | return StringRef(); |
| 327 | } |
| 328 | |
| 329 | /// Print characters from \p FromFile starting at \p NextToWrite up until the |
| 330 | /// inclusion directive at \p StartToken, then print out the inclusion |
| 331 | /// inclusion directive disabled by a #if directive, updating \p NextToWrite |
| 332 | /// and \p Line to track the number of source lines visited and the progress |
| 333 | /// through the \p FromFile buffer. |
| 334 | void InclusionRewriter::(Lexer &DirectiveLex, |
| 335 | const Token &StartToken, |
| 336 | const MemoryBufferRef &FromFile, |
| 337 | StringRef LocalEOL, |
| 338 | unsigned &NextToWrite, int &Line, |
| 339 | const IncludedFile *Inc) { |
| 340 | OutputContentUpTo(FromFile, WriteFrom&: NextToWrite, |
| 341 | WriteTo: SM.getFileOffset(SpellingLoc: StartToken.getLocation()), LocalEOL, Line, |
| 342 | EnsureNewline: false); |
| 343 | Token DirectiveToken; |
| 344 | do { |
| 345 | DirectiveLex.LexFromRawLexer(Result&: DirectiveToken); |
| 346 | } while (!DirectiveToken.is(K: tok::eod) && DirectiveToken.isNot(K: tok::eof)); |
| 347 | if (FromFile == PredefinesBuffer) { |
| 348 | // OutputContentUpTo() would not output anything anyway. |
| 349 | return; |
| 350 | } |
| 351 | if (Inc) { |
| 352 | OS << "#if defined(__CLANG_REWRITTEN_INCLUDES) " ; |
| 353 | if (isSystem(CK: Inc->FileType)) |
| 354 | OS << "|| defined(__CLANG_REWRITTEN_SYSTEM_INCLUDES) " ; |
| 355 | OS << "/* " << getIncludedFileName(Inc); |
| 356 | } else { |
| 357 | OS << "#if 0 /*" ; |
| 358 | } |
| 359 | OS << " expanded by -frewrite-includes */" << MainEOL; |
| 360 | OutputContentUpTo(FromFile, WriteFrom&: NextToWrite, |
| 361 | WriteTo: SM.getFileOffset(SpellingLoc: DirectiveToken.getLocation()) + |
| 362 | DirectiveToken.getLength(), |
| 363 | LocalEOL, Line, EnsureNewline: true); |
| 364 | OS << (Inc ? "#else /* " : "#endif /*" ) << getIncludedFileName(Inc) |
| 365 | << " expanded by -frewrite-includes */" << MainEOL; |
| 366 | } |
| 367 | |
| 368 | /// Find the next identifier in the pragma directive specified by \p RawToken. |
| 369 | StringRef InclusionRewriter::NextIdentifierName(Lexer &RawLex, |
| 370 | Token &RawToken) { |
| 371 | RawLex.LexFromRawLexer(Result&: RawToken); |
| 372 | if (RawToken.is(K: tok::raw_identifier)) |
| 373 | PP.LookUpIdentifierInfo(Identifier&: RawToken); |
| 374 | if (RawToken.is(K: tok::identifier)) |
| 375 | return RawToken.getIdentifierInfo()->getName(); |
| 376 | return StringRef(); |
| 377 | } |
| 378 | |
| 379 | /// Use a raw lexer to analyze \p FileId, incrementally copying parts of it |
| 380 | /// and including content of included files recursively. |
| 381 | void InclusionRewriter::Process(FileID FileId, |
| 382 | SrcMgr::CharacteristicKind FileType) { |
| 383 | MemoryBufferRef FromFile; |
| 384 | { |
| 385 | auto B = SM.getBufferOrNone(FID: FileId); |
| 386 | assert(B && "Attempting to process invalid inclusion" ); |
| 387 | if (B) |
| 388 | FromFile = *B; |
| 389 | } |
| 390 | StringRef FileName = FromFile.getBufferIdentifier(); |
| 391 | Lexer RawLex(FileId, FromFile, PP.getSourceManager(), PP.getLangOpts()); |
| 392 | RawLex.SetCommentRetentionState(false); |
| 393 | |
| 394 | StringRef LocalEOL = FromFile.getBuffer().detectEOL(); |
| 395 | |
| 396 | // Per the GNU docs: "1" indicates entering a new file. |
| 397 | if (FileId == SM.getMainFileID() || FileId == PP.getPredefinesFileID()) |
| 398 | WriteLineInfo(Filename: FileName, Line: 1, FileType, Extra: "" ); |
| 399 | else |
| 400 | WriteLineInfo(Filename: FileName, Line: 1, FileType, Extra: " 1" ); |
| 401 | |
| 402 | if (SM.getFileIDSize(FID: FileId) == 0) |
| 403 | return; |
| 404 | |
| 405 | // The next byte to be copied from the source file, which may be non-zero if |
| 406 | // the lexer handled a BOM. |
| 407 | unsigned NextToWrite = SM.getFileOffset(SpellingLoc: RawLex.getSourceLocation()); |
| 408 | assert(SM.getLineNumber(FileId, NextToWrite) == 1); |
| 409 | int Line = 1; // The current input file line number. |
| 410 | |
| 411 | Token RawToken; |
| 412 | RawLex.LexFromRawLexer(Result&: RawToken); |
| 413 | |
| 414 | // TODO: Consider adding a switch that strips possibly unimportant content, |
| 415 | // such as comments, to reduce the size of repro files. |
| 416 | while (RawToken.isNot(K: tok::eof)) { |
| 417 | if (RawToken.is(K: tok::hash) && RawToken.isAtStartOfLine()) { |
| 418 | RawLex.setParsingPreprocessorDirective(true); |
| 419 | Token HashToken = RawToken; |
| 420 | RawLex.LexFromRawLexer(Result&: RawToken); |
| 421 | if (RawToken.is(K: tok::raw_identifier)) |
| 422 | PP.LookUpIdentifierInfo(Identifier&: RawToken); |
| 423 | if (RawToken.getIdentifierInfo() != nullptr) { |
| 424 | switch (RawToken.getIdentifierInfo()->getPPKeywordID()) { |
| 425 | case tok::pp_include: |
| 426 | case tok::pp_include_next: |
| 427 | case tok::pp_import: { |
| 428 | SourceLocation Loc = HashToken.getLocation(); |
| 429 | const IncludedFile *Inc = FindIncludeAtLocation(Loc); |
| 430 | CommentOutDirective(DirectiveLex&: RawLex, StartToken: HashToken, FromFile, LocalEOL, |
| 431 | NextToWrite, Line, Inc); |
| 432 | if (FileId != PP.getPredefinesFileID()) |
| 433 | WriteLineInfo(Filename: FileName, Line: Line - 1, FileType, Extra: "" ); |
| 434 | StringRef ; |
| 435 | if (const Module *Mod = FindModuleAtLocation(Loc)) |
| 436 | WriteImplicitModuleImport(Mod); |
| 437 | else if (Inc) { |
| 438 | const Module *Mod = FindEnteredModule(Loc); |
| 439 | if (Mod) |
| 440 | OS << "#pragma clang module begin " |
| 441 | << Mod->getFullModuleName(AllowStringLiterals: true) << "\n" ; |
| 442 | |
| 443 | // Include and recursively process the file. |
| 444 | Process(FileId: Inc->Id, FileType: Inc->FileType); |
| 445 | |
| 446 | if (Mod) |
| 447 | OS << "#pragma clang module end /*" |
| 448 | << Mod->getFullModuleName(AllowStringLiterals: true) << "*/\n" ; |
| 449 | // There's no #include, therefore no #if, for -include files. |
| 450 | if (FromFile != PredefinesBuffer) { |
| 451 | OS << "#endif /* " << getIncludedFileName(Inc) |
| 452 | << " expanded by -frewrite-includes */" << LocalEOL; |
| 453 | } |
| 454 | |
| 455 | // Add line marker to indicate we're returning from an included |
| 456 | // file. |
| 457 | LineInfoExtra = " 2" ; |
| 458 | } |
| 459 | // fix up lineinfo (since commented out directive changed line |
| 460 | // numbers) for inclusions that were skipped due to header guards |
| 461 | WriteLineInfo(Filename: FileName, Line, FileType, Extra: LineInfoExtra); |
| 462 | break; |
| 463 | } |
| 464 | case tok::pp_pragma: { |
| 465 | StringRef Identifier = NextIdentifierName(RawLex, RawToken); |
| 466 | if (Identifier == "clang" || Identifier == "GCC" ) { |
| 467 | if (NextIdentifierName(RawLex, RawToken) == "system_header" ) { |
| 468 | // keep the directive in, commented out |
| 469 | CommentOutDirective(DirectiveLex&: RawLex, StartToken: HashToken, FromFile, LocalEOL, |
| 470 | NextToWrite, Line); |
| 471 | // update our own type |
| 472 | FileType = SM.getFileCharacteristic(Loc: RawToken.getLocation()); |
| 473 | WriteLineInfo(Filename: FileName, Line, FileType); |
| 474 | } |
| 475 | } else if (Identifier == "once" ) { |
| 476 | // keep the directive in, commented out |
| 477 | CommentOutDirective(DirectiveLex&: RawLex, StartToken: HashToken, FromFile, LocalEOL, |
| 478 | NextToWrite, Line); |
| 479 | WriteLineInfo(Filename: FileName, Line, FileType); |
| 480 | } |
| 481 | break; |
| 482 | } |
| 483 | case tok::pp_if: |
| 484 | case tok::pp_elif: { |
| 485 | bool elif = (RawToken.getIdentifierInfo()->getPPKeywordID() == |
| 486 | tok::pp_elif); |
| 487 | bool isTrue = IsIfAtLocationTrue(Loc: RawToken.getLocation()); |
| 488 | OutputContentUpTo(FromFile, WriteFrom&: NextToWrite, |
| 489 | WriteTo: SM.getFileOffset(SpellingLoc: HashToken.getLocation()), |
| 490 | LocalEOL, Line, /*EnsureNewline=*/true); |
| 491 | do { |
| 492 | RawLex.LexFromRawLexer(Result&: RawToken); |
| 493 | } while (!RawToken.is(K: tok::eod) && RawToken.isNot(K: tok::eof)); |
| 494 | // We need to disable the old condition, but that is tricky. |
| 495 | // Trying to comment it out can easily lead to comment nesting. |
| 496 | // So instead make the condition harmless by making it enclose |
| 497 | // and empty block. Moreover, put it itself inside an #if 0 block |
| 498 | // to disable it from getting evaluated (e.g. __has_include_next |
| 499 | // warns if used from the primary source file). |
| 500 | OS << "#if 0 /* disabled by -frewrite-includes */" << MainEOL; |
| 501 | if (elif) { |
| 502 | OS << "#if 0" << MainEOL; |
| 503 | } |
| 504 | OutputContentUpTo(FromFile, WriteFrom&: NextToWrite, |
| 505 | WriteTo: SM.getFileOffset(SpellingLoc: RawToken.getLocation()) + |
| 506 | RawToken.getLength(), |
| 507 | LocalEOL, Line, /*EnsureNewline=*/true); |
| 508 | // Close the empty block and the disabling block. |
| 509 | OS << "#endif" << MainEOL; |
| 510 | OS << "#endif /* disabled by -frewrite-includes */" << MainEOL; |
| 511 | OS << (elif ? "#elif " : "#if " ) << (isTrue ? "1" : "0" ) |
| 512 | << " /* evaluated by -frewrite-includes */" << MainEOL; |
| 513 | WriteLineInfo(Filename: FileName, Line, FileType); |
| 514 | break; |
| 515 | } |
| 516 | case tok::pp_endif: |
| 517 | case tok::pp_else: { |
| 518 | // We surround every #include by #if 0 to comment it out, but that |
| 519 | // changes line numbers. These are fixed up right after that, but |
| 520 | // the whole #include could be inside a preprocessor conditional |
| 521 | // that is not processed. So it is necessary to fix the line |
| 522 | // numbers one the next line after each #else/#endif as well. |
| 523 | RawLex.SetKeepWhitespaceMode(true); |
| 524 | do { |
| 525 | RawLex.LexFromRawLexer(Result&: RawToken); |
| 526 | } while (RawToken.isNot(K: tok::eod) && RawToken.isNot(K: tok::eof)); |
| 527 | OutputContentUpTo(FromFile, WriteFrom&: NextToWrite, |
| 528 | WriteTo: SM.getFileOffset(SpellingLoc: RawToken.getLocation()) + |
| 529 | RawToken.getLength(), |
| 530 | LocalEOL, Line, /*EnsureNewline=*/ true); |
| 531 | WriteLineInfo(Filename: FileName, Line, FileType); |
| 532 | RawLex.SetKeepWhitespaceMode(false); |
| 533 | break; |
| 534 | } |
| 535 | default: |
| 536 | break; |
| 537 | } |
| 538 | } |
| 539 | RawLex.setParsingPreprocessorDirective(false); |
| 540 | } |
| 541 | RawLex.LexFromRawLexer(Result&: RawToken); |
| 542 | } |
| 543 | OutputContentUpTo(FromFile, WriteFrom&: NextToWrite, |
| 544 | WriteTo: SM.getFileOffset(SpellingLoc: SM.getLocForEndOfFile(FID: FileId)), LocalEOL, |
| 545 | Line, /*EnsureNewline=*/true); |
| 546 | } |
| 547 | |
| 548 | /// InclusionRewriterInInput - Implement -frewrite-includes mode. |
| 549 | void clang::RewriteIncludesInInput(Preprocessor &PP, raw_ostream *OS, |
| 550 | const PreprocessorOutputOptions &Opts) { |
| 551 | SourceManager &SM = PP.getSourceManager(); |
| 552 | InclusionRewriter *Rewrite = new InclusionRewriter( |
| 553 | PP, *OS, Opts.ShowLineMarkers, Opts.UseLineDirectives); |
| 554 | Rewrite->detectMainFileEOL(); |
| 555 | |
| 556 | PP.addPPCallbacks(C: std::unique_ptr<PPCallbacks>(Rewrite)); |
| 557 | PP.IgnorePragmas(); |
| 558 | |
| 559 | // First let the preprocessor process the entire file and call callbacks. |
| 560 | // Callbacks will record which #include's were actually performed. |
| 561 | PP.EnterMainSourceFile(); |
| 562 | Token Tok; |
| 563 | // Only preprocessor directives matter here, so disable macro expansion |
| 564 | // everywhere else as an optimization. |
| 565 | // TODO: It would be even faster if the preprocessor could be switched |
| 566 | // to a mode where it would parse only preprocessor directives and comments, |
| 567 | // nothing else matters for parsing or processing. |
| 568 | PP.SetMacroExpansionOnlyInDirectives(); |
| 569 | do { |
| 570 | PP.Lex(Result&: Tok); |
| 571 | if (Tok.is(K: tok::annot_module_begin)) |
| 572 | Rewrite->handleModuleBegin(Tok); |
| 573 | } while (Tok.isNot(K: tok::eof)); |
| 574 | Rewrite->setPredefinesBuffer(SM.getBufferOrFake(FID: PP.getPredefinesFileID())); |
| 575 | Rewrite->Process(FileId: PP.getPredefinesFileID(), FileType: SrcMgr::C_User); |
| 576 | Rewrite->Process(FileId: SM.getMainFileID(), FileType: SrcMgr::C_User); |
| 577 | OS->flush(); |
| 578 | } |
| 579 | |