1 //===--- Preprocess.cpp - C Language Family Preprocessor Implementation ---===//
3 // The LLVM Compiler Infrastructure
5 // This file is distributed under the University of Illinois Open Source
6 // License. See LICENSE.TXT for details.
8 //===----------------------------------------------------------------------===//
10 // This file implements the Preprocessor interface.
12 //===----------------------------------------------------------------------===//
14 // Options to support:
15 // -H - Print the name of each header file used.
16 // -d[DNI] - Dump various things.
17 // -fworking-directory - #line's with preprocessor's working dir.
19 // -dependency-file,-M,-MM,-MF,-MG,-MP,-MT,-MQ,-MD,-MMD
24 // "Multiple include guards may be useful for:\n"
26 //===----------------------------------------------------------------------===//
28 #include "clang/Lex/Preprocessor.h"
29 #include "MacroArgs.h"
30 #include "clang/Lex/ExternalPreprocessorSource.h"
31 #include "clang/Lex/HeaderSearch.h"
32 #include "clang/Lex/MacroInfo.h"
33 #include "clang/Lex/Pragma.h"
34 #include "clang/Lex/PreprocessingRecord.h"
35 #include "clang/Lex/ScratchBuffer.h"
36 #include "clang/Lex/LexDiagnostic.h"
37 #include "clang/Basic/SourceManager.h"
38 #include "clang/Basic/FileManager.h"
39 #include "clang/Basic/TargetInfo.h"
40 #include "llvm/ADT/APFloat.h"
41 #include "llvm/ADT/SmallVector.h"
42 #include "llvm/Support/MemoryBuffer.h"
43 #include "llvm/Support/raw_ostream.h"
44 using namespace clang;
46 //===----------------------------------------------------------------------===//
47 ExternalPreprocessorSource::~ExternalPreprocessorSource() { }
49 Preprocessor::Preprocessor(Diagnostic &diags, const LangOptions &opts,
50 const TargetInfo &target, SourceManager &SM,
51 HeaderSearch &Headers,
52 IdentifierInfoLookup* IILookup,
54 : Diags(&diags), Features(opts), Target(target),FileMgr(Headers.getFileMgr()),
55 SourceMgr(SM), HeaderInfo(Headers), ExternalSource(0),
56 Identifiers(opts, IILookup), BuiltinInfo(Target), CodeCompletionFile(0),
57 CurPPLexer(0), CurDirLookup(0), Callbacks(0), MacroArgCache(0), Record(0) {
58 ScratchBuf = new ScratchBuffer(SourceMgr);
59 CounterValue = 0; // __COUNTER__ starts at 0.
60 OwnsHeaderSearch = OwnsHeaders;
63 NumDirectives = NumDefined = NumUndefined = NumPragma = 0;
64 NumIf = NumElse = NumEndif = 0;
65 NumEnteredSourceFiles = 0;
66 NumMacroExpanded = NumFnMacroExpanded = NumBuiltinMacroExpanded = 0;
67 NumFastMacroExpanded = NumTokenPaste = NumFastTokenPaste = 0;
68 MaxIncludeStackDepth = 0;
71 // Default to discarding comments.
73 KeepMacroComments = false;
75 // Macro expansion is enabled.
76 DisableMacroExpansion = false;
78 NumCachedTokenLexers = 0;
82 // We haven't read anything from the external source.
83 ReadMacrosFromExternalSource = false;
85 // "Poison" __VA_ARGS__, which can only appear in the expansion of a macro.
86 // This gets unpoisoned where it is allowed.
87 (Ident__VA_ARGS__ = getIdentifierInfo("__VA_ARGS__"))->setIsPoisoned();
89 // Initialize the pragma handlers.
90 PragmaHandlers = new PragmaNamespace(0);
91 RegisterBuiltinPragmas();
93 // Initialize builtin macros like __LINE__ and friends.
94 RegisterBuiltinMacros();
97 Preprocessor::~Preprocessor() {
98 assert(BacktrackPositions.empty() && "EnableBacktrack/Backtrack imbalance!");
100 while (!IncludeMacroStack.empty()) {
101 delete IncludeMacroStack.back().TheLexer;
102 delete IncludeMacroStack.back().TheTokenLexer;
103 IncludeMacroStack.pop_back();
106 // Free any macro definitions.
107 for (llvm::DenseMap<IdentifierInfo*, MacroInfo*>::iterator I =
108 Macros.begin(), E = Macros.end(); I != E; ++I) {
109 // We don't need to free the MacroInfo objects directly. These
110 // will be released when the BumpPtrAllocator 'BP' object gets
111 // destroyed. We still need to run the dtor, however, to free
112 // memory alocated by MacroInfo.
113 I->second->Destroy(BP);
114 I->first->setHasMacroDefinition(false);
117 // Free any cached macro expanders.
118 for (unsigned i = 0, e = NumCachedTokenLexers; i != e; ++i)
119 delete TokenLexerCache[i];
121 // Free any cached MacroArgs.
122 for (MacroArgs *ArgList = MacroArgCache; ArgList; )
123 ArgList = ArgList->deallocate();
125 // Release pragma information.
126 delete PragmaHandlers;
128 // Delete the scratch buffer info.
131 // Delete the header search info, if we own it.
132 if (OwnsHeaderSearch)
138 void Preprocessor::setPTHManager(PTHManager* pm) {
140 FileMgr.addStatCache(PTH->createStatCache());
143 void Preprocessor::DumpToken(const Token &Tok, bool DumpFlags) const {
144 llvm::errs() << tok::getTokenName(Tok.getKind()) << " '"
145 << getSpelling(Tok) << "'";
147 if (!DumpFlags) return;
149 llvm::errs() << "\t";
150 if (Tok.isAtStartOfLine())
151 llvm::errs() << " [StartOfLine]";
152 if (Tok.hasLeadingSpace())
153 llvm::errs() << " [LeadingSpace]";
154 if (Tok.isExpandDisabled())
155 llvm::errs() << " [ExpandDisabled]";
156 if (Tok.needsCleaning()) {
157 const char *Start = SourceMgr.getCharacterData(Tok.getLocation());
158 llvm::errs() << " [UnClean='" << std::string(Start, Start+Tok.getLength())
162 llvm::errs() << "\tLoc=<";
163 DumpLocation(Tok.getLocation());
167 void Preprocessor::DumpLocation(SourceLocation Loc) const {
171 void Preprocessor::DumpMacro(const MacroInfo &MI) const {
172 llvm::errs() << "MACRO: ";
173 for (unsigned i = 0, e = MI.getNumTokens(); i != e; ++i) {
174 DumpToken(MI.getReplacementToken(i));
177 llvm::errs() << "\n";
180 void Preprocessor::PrintStats() {
181 llvm::errs() << "\n*** Preprocessor Stats:\n";
182 llvm::errs() << NumDirectives << " directives found:\n";
183 llvm::errs() << " " << NumDefined << " #define.\n";
184 llvm::errs() << " " << NumUndefined << " #undef.\n";
185 llvm::errs() << " #include/#include_next/#import:\n";
186 llvm::errs() << " " << NumEnteredSourceFiles << " source files entered.\n";
187 llvm::errs() << " " << MaxIncludeStackDepth << " max include stack depth\n";
188 llvm::errs() << " " << NumIf << " #if/#ifndef/#ifdef.\n";
189 llvm::errs() << " " << NumElse << " #else/#elif.\n";
190 llvm::errs() << " " << NumEndif << " #endif.\n";
191 llvm::errs() << " " << NumPragma << " #pragma.\n";
192 llvm::errs() << NumSkipped << " #if/#ifndef#ifdef regions skipped\n";
194 llvm::errs() << NumMacroExpanded << "/" << NumFnMacroExpanded << "/"
195 << NumBuiltinMacroExpanded << " obj/fn/builtin macros expanded, "
196 << NumFastMacroExpanded << " on the fast path.\n";
197 llvm::errs() << (NumFastTokenPaste+NumTokenPaste)
198 << " token paste (##) operations performed, "
199 << NumFastTokenPaste << " on the fast path.\n";
202 Preprocessor::macro_iterator
203 Preprocessor::macro_begin(bool IncludeExternalMacros) const {
204 if (IncludeExternalMacros && ExternalSource &&
205 !ReadMacrosFromExternalSource) {
206 ReadMacrosFromExternalSource = true;
207 ExternalSource->ReadDefinedMacros();
210 return Macros.begin();
213 Preprocessor::macro_iterator
214 Preprocessor::macro_end(bool IncludeExternalMacros) const {
215 if (IncludeExternalMacros && ExternalSource &&
216 !ReadMacrosFromExternalSource) {
217 ReadMacrosFromExternalSource = true;
218 ExternalSource->ReadDefinedMacros();
224 bool Preprocessor::SetCodeCompletionPoint(const FileEntry *File,
225 unsigned TruncateAtLine,
226 unsigned TruncateAtColumn) {
227 using llvm::MemoryBuffer;
229 CodeCompletionFile = File;
231 // Okay to clear out the code-completion point by passing NULL.
232 if (!CodeCompletionFile)
235 // Load the actual file's contents.
236 bool Invalid = false;
237 const MemoryBuffer *Buffer = SourceMgr.getMemoryBufferForFile(File, &Invalid);
241 // Find the byte position of the truncation point.
242 const char *Position = Buffer->getBufferStart();
243 for (unsigned Line = 1; Line < TruncateAtLine; ++Line) {
244 for (; *Position; ++Position) {
245 if (*Position != '\r' && *Position != '\n')
248 // Eat \r\n or \n\r as a single line.
249 if ((Position[1] == '\r' || Position[1] == '\n') &&
250 Position[0] != Position[1])
257 Position += TruncateAtColumn - 1;
259 // Truncate the buffer.
260 if (Position < Buffer->getBufferEnd()) {
261 MemoryBuffer *TruncatedBuffer
262 = MemoryBuffer::getMemBufferCopy(Buffer->getBufferStart(), Position,
263 Buffer->getBufferIdentifier());
264 SourceMgr.overrideFileContents(File, TruncatedBuffer);
270 bool Preprocessor::isCodeCompletionFile(SourceLocation FileLoc) const {
271 return CodeCompletionFile && FileLoc.isFileID() &&
272 SourceMgr.getFileEntryForID(SourceMgr.getFileID(FileLoc))
273 == CodeCompletionFile;
276 //===----------------------------------------------------------------------===//
278 //===----------------------------------------------------------------------===//
280 /// getSpelling() - Return the 'spelling' of this token. The spelling of a
281 /// token are the characters used to represent the token in the source file
282 /// after trigraph expansion and escaped-newline folding. In particular, this
283 /// wants to get the true, uncanonicalized, spelling of things like digraphs
285 std::string Preprocessor::getSpelling(const Token &Tok,
286 const SourceManager &SourceMgr,
287 const LangOptions &Features,
289 assert((int)Tok.getLength() >= 0 && "Token character range is bogus!");
291 // If this token contains nothing interesting, return it directly.
292 bool CharDataInvalid = false;
293 const char* TokStart = SourceMgr.getCharacterData(Tok.getLocation(),
296 *Invalid = CharDataInvalid;
298 return std::string();
300 if (!Tok.needsCleaning())
301 return std::string(TokStart, TokStart+Tok.getLength());
304 Result.reserve(Tok.getLength());
306 // Otherwise, hard case, relex the characters into the string.
307 for (const char *Ptr = TokStart, *End = TokStart+Tok.getLength();
310 Result.push_back(Lexer::getCharAndSizeNoWarn(Ptr, CharSize, Features));
313 assert(Result.size() != unsigned(Tok.getLength()) &&
314 "NeedsCleaning flag set on something that didn't need cleaning!");
318 /// getSpelling() - Return the 'spelling' of this token. The spelling of a
319 /// token are the characters used to represent the token in the source file
320 /// after trigraph expansion and escaped-newline folding. In particular, this
321 /// wants to get the true, uncanonicalized, spelling of things like digraphs
323 std::string Preprocessor::getSpelling(const Token &Tok, bool *Invalid) const {
324 return getSpelling(Tok, SourceMgr, Features, Invalid);
327 /// getSpelling - This method is used to get the spelling of a token into a
328 /// preallocated buffer, instead of as an std::string. The caller is required
329 /// to allocate enough space for the token, which is guaranteed to be at least
330 /// Tok.getLength() bytes long. The actual length of the token is returned.
332 /// Note that this method may do two possible things: it may either fill in
333 /// the buffer specified with characters, or it may *change the input pointer*
334 /// to point to a constant buffer with the data already in it (avoiding a
335 /// copy). The caller is not allowed to modify the returned buffer pointer
336 /// if an internal buffer is returned.
337 unsigned Preprocessor::getSpelling(const Token &Tok,
338 const char *&Buffer, bool *Invalid) const {
339 assert((int)Tok.getLength() >= 0 && "Token character range is bogus!");
341 // If this token is an identifier, just return the string from the identifier
342 // table, which is very quick.
343 if (const IdentifierInfo *II = Tok.getIdentifierInfo()) {
344 Buffer = II->getNameStart();
345 return II->getLength();
348 // Otherwise, compute the start of the token in the input lexer buffer.
349 const char *TokStart = 0;
352 TokStart = Tok.getLiteralData();
355 bool CharDataInvalid = false;
356 TokStart = SourceMgr.getCharacterData(Tok.getLocation(), &CharDataInvalid);
358 *Invalid = CharDataInvalid;
359 if (CharDataInvalid) {
365 // If this token contains nothing interesting, return it directly.
366 if (!Tok.needsCleaning()) {
368 return Tok.getLength();
371 // Otherwise, hard case, relex the characters into the string.
372 char *OutBuf = const_cast<char*>(Buffer);
373 for (const char *Ptr = TokStart, *End = TokStart+Tok.getLength();
376 *OutBuf++ = Lexer::getCharAndSizeNoWarn(Ptr, CharSize, Features);
379 assert(unsigned(OutBuf-Buffer) != Tok.getLength() &&
380 "NeedsCleaning flag set on something that didn't need cleaning!");
382 return OutBuf-Buffer;
385 /// getSpelling - This method is used to get the spelling of a token into a
386 /// SmallVector. Note that the returned StringRef may not point to the
387 /// supplied buffer if a copy can be avoided.
388 llvm::StringRef Preprocessor::getSpelling(const Token &Tok,
389 llvm::SmallVectorImpl<char> &Buffer,
390 bool *Invalid) const {
391 // Try the fast path.
392 if (const IdentifierInfo *II = Tok.getIdentifierInfo())
393 return II->getName();
395 // Resize the buffer if we need to copy into it.
396 if (Tok.needsCleaning())
397 Buffer.resize(Tok.getLength());
399 const char *Ptr = Buffer.data();
400 unsigned Len = getSpelling(Tok, Ptr, Invalid);
401 return llvm::StringRef(Ptr, Len);
404 /// CreateString - Plop the specified string into a scratch buffer and return a
405 /// location for it. If specified, the source location provides a source
406 /// location for the token.
407 void Preprocessor::CreateString(const char *Buf, unsigned Len, Token &Tok,
408 SourceLocation InstantiationLoc) {
412 SourceLocation Loc = ScratchBuf->getToken(Buf, Len, DestPtr);
414 if (InstantiationLoc.isValid())
415 Loc = SourceMgr.createInstantiationLoc(Loc, InstantiationLoc,
416 InstantiationLoc, Len);
417 Tok.setLocation(Loc);
419 // If this is a literal token, set the pointer data.
421 Tok.setLiteralData(DestPtr);
425 /// AdvanceToTokenCharacter - Given a location that specifies the start of a
426 /// token, return a new location that specifies a character within the token.
427 SourceLocation Preprocessor::AdvanceToTokenCharacter(SourceLocation TokStart,
429 // Figure out how many physical characters away the specified instantiation
430 // character is. This needs to take into consideration newlines and
432 bool Invalid = false;
433 const char *TokPtr = SourceMgr.getCharacterData(TokStart, &Invalid);
435 // If they request the first char of the token, we're trivially done.
436 if (Invalid || (CharNo == 0 && Lexer::isObviouslySimpleCharacter(*TokPtr)))
439 unsigned PhysOffset = 0;
441 // The usual case is that tokens don't contain anything interesting. Skip
442 // over the uninteresting characters. If a token only consists of simple
443 // chars, this method is extremely fast.
444 while (Lexer::isObviouslySimpleCharacter(*TokPtr)) {
446 return TokStart.getFileLocWithOffset(PhysOffset);
447 ++TokPtr, --CharNo, ++PhysOffset;
450 // If we have a character that may be a trigraph or escaped newline, use a
451 // lexer to parse it correctly.
452 for (; CharNo; --CharNo) {
454 Lexer::getCharAndSizeNoWarn(TokPtr, Size, Features);
459 // Final detail: if we end up on an escaped newline, we want to return the
460 // location of the actual byte of the token. For example foo\<newline>bar
461 // advanced by 3 should return the location of b, not of \\. One compounding
462 // detail of this is that the escape may be made by a trigraph.
463 if (!Lexer::isObviouslySimpleCharacter(*TokPtr))
464 PhysOffset += Lexer::SkipEscapedNewLines(TokPtr)-TokPtr;
466 return TokStart.getFileLocWithOffset(PhysOffset);
469 SourceLocation Preprocessor::getLocForEndOfToken(SourceLocation Loc,
471 if (Loc.isInvalid() || !Loc.isFileID())
472 return SourceLocation();
474 unsigned Len = Lexer::MeasureTokenLength(Loc, getSourceManager(), Features);
480 return AdvanceToTokenCharacter(Loc, Len);
485 //===----------------------------------------------------------------------===//
486 // Preprocessor Initialization Methods
487 //===----------------------------------------------------------------------===//
490 /// EnterMainSourceFile - Enter the specified FileID as the main source file,
491 /// which implicitly adds the builtin defines etc.
492 bool Preprocessor::EnterMainSourceFile() {
493 // We do not allow the preprocessor to reenter the main file. Doing so will
494 // cause FileID's to accumulate information from both runs (e.g. #line
495 // information) and predefined macros aren't guaranteed to be set properly.
496 assert(NumEnteredSourceFiles == 0 && "Cannot reenter the main file!");
497 FileID MainFileID = SourceMgr.getMainFileID();
499 // Enter the main file source buffer.
500 std::string ErrorStr;
501 if (EnterSourceFile(MainFileID, 0, ErrorStr))
504 // Tell the header info that the main file was entered. If the file is later
505 // #imported, it won't be re-entered.
506 if (const FileEntry *FE = SourceMgr.getFileEntryForID(MainFileID))
507 HeaderInfo.IncrementIncludeCount(FE);
509 // Preprocess Predefines to populate the initial preprocessor state.
510 llvm::MemoryBuffer *SB =
511 llvm::MemoryBuffer::getMemBufferCopy(Predefines.data(),
512 Predefines.data() + Predefines.size(),
514 assert(SB && "Cannot fail to create predefined source buffer");
515 FileID FID = SourceMgr.createFileIDForMemBuffer(SB);
516 assert(!FID.isInvalid() && "Could not create FileID for predefines?");
518 // Start parsing the predefines.
519 return EnterSourceFile(FID, 0, ErrorStr);
522 void Preprocessor::EndSourceFile() {
523 // Notify the client that we reached the end of the source file.
525 Callbacks->EndOfMainFile();
528 //===----------------------------------------------------------------------===//
529 // Lexer Event Handling.
530 //===----------------------------------------------------------------------===//
532 /// LookUpIdentifierInfo - Given a tok::identifier token, look up the
533 /// identifier information for the token and install it into the token.
534 IdentifierInfo *Preprocessor::LookUpIdentifierInfo(Token &Identifier,
535 const char *BufPtr) const {
536 assert(Identifier.is(tok::identifier) && "Not an identifier!");
537 assert(Identifier.getIdentifierInfo() == 0 && "Identinfo already exists!");
539 // Look up this token, see if it is a macro, or if it is a language keyword.
541 if (BufPtr && !Identifier.needsCleaning()) {
542 // No cleaning needed, just use the characters from the lexed buffer.
543 II = getIdentifierInfo(llvm::StringRef(BufPtr, Identifier.getLength()));
545 // Cleaning needed, alloca a buffer, clean into it, then use the buffer.
546 llvm::SmallString<64> IdentifierBuffer;
547 llvm::StringRef CleanedStr = getSpelling(Identifier, IdentifierBuffer);
548 II = getIdentifierInfo(CleanedStr);
550 Identifier.setIdentifierInfo(II);
555 /// HandleIdentifier - This callback is invoked when the lexer reads an
556 /// identifier. This callback looks up the identifier in the map and/or
557 /// potentially macro expands it or turns it into a named token (like 'for').
559 /// Note that callers of this method are guarded by checking the
560 /// IdentifierInfo's 'isHandleIdentifierCase' bit. If this method changes, the
561 /// IdentifierInfo methods that compute these properties will need to change to
563 void Preprocessor::HandleIdentifier(Token &Identifier) {
564 assert(Identifier.getIdentifierInfo() &&
565 "Can't handle identifiers without identifier info!");
567 IdentifierInfo &II = *Identifier.getIdentifierInfo();
569 // If this identifier was poisoned, and if it was not produced from a macro
570 // expansion, emit an error.
571 if (II.isPoisoned() && CurPPLexer) {
572 if (&II != Ident__VA_ARGS__) // We warn about __VA_ARGS__ with poisoning.
573 Diag(Identifier, diag::err_pp_used_poisoned_id);
575 Diag(Identifier, diag::ext_pp_bad_vaargs_use);
578 // If this is a macro to be expanded, do it.
579 if (MacroInfo *MI = getMacroInfo(&II)) {
580 if (!DisableMacroExpansion && !Identifier.isExpandDisabled()) {
581 if (MI->isEnabled()) {
582 if (!HandleMacroExpandedIdentifier(Identifier, MI))
585 // C99 6.10.3.4p2 says that a disabled macro may never again be
586 // expanded, even if it's in a context where it could be expanded in the
588 Identifier.setFlag(Token::DisableExpand);
593 // C++ 2.11p2: If this is an alternative representation of a C++ operator,
594 // then we act as if it is the actual operator and not the textual
595 // representation of it.
596 if (II.isCPlusPlusOperatorKeyword())
597 Identifier.setIdentifierInfo(0);
599 // If this is an extension token, diagnose its use.
600 // We avoid diagnosing tokens that originate from macro definitions.
601 // FIXME: This warning is disabled in cases where it shouldn't be,
602 // like "#define TY typeof", "TY(1) x".
603 if (II.isExtensionToken() && !DisableMacroExpansion)
604 Diag(Identifier, diag::ext_token_used);
607 void Preprocessor::AddCommentHandler(CommentHandler *Handler) {
608 assert(Handler && "NULL comment handler");
609 assert(std::find(CommentHandlers.begin(), CommentHandlers.end(), Handler) ==
610 CommentHandlers.end() && "Comment handler already registered");
611 CommentHandlers.push_back(Handler);
614 void Preprocessor::RemoveCommentHandler(CommentHandler *Handler) {
615 std::vector<CommentHandler *>::iterator Pos
616 = std::find(CommentHandlers.begin(), CommentHandlers.end(), Handler);
617 assert(Pos != CommentHandlers.end() && "Comment handler not registered");
618 CommentHandlers.erase(Pos);
621 bool Preprocessor::HandleComment(Token &result, SourceRange Comment) {
622 bool AnyPendingTokens = false;
623 for (std::vector<CommentHandler *>::iterator H = CommentHandlers.begin(),
624 HEnd = CommentHandlers.end();
626 if ((*H)->HandleComment(*this, Comment))
627 AnyPendingTokens = true;
629 if (!AnyPendingTokens || getCommentRetentionState())
635 CommentHandler::~CommentHandler() { }
637 void Preprocessor::createPreprocessingRecord() {
641 Record = new PreprocessingRecord;
642 addPPCallbacks(Record);