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(llvm::StringRef());
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);
116 for (std::vector<MacroInfo*>::iterator I = MICache.begin(),
117 E = MICache.end(); I != E; ++I) {
118 // We don't need to free the MacroInfo objects directly. These
119 // will be released when the BumpPtrAllocator 'BP' object gets
120 // destroyed. We still need to run the dtor, however, to free
121 // memory alocated by MacroInfo.
125 // Free any cached macro expanders.
126 for (unsigned i = 0, e = NumCachedTokenLexers; i != e; ++i)
127 delete TokenLexerCache[i];
129 // Free any cached MacroArgs.
130 for (MacroArgs *ArgList = MacroArgCache; ArgList; )
131 ArgList = ArgList->deallocate();
133 // Release pragma information.
134 delete PragmaHandlers;
136 // Delete the scratch buffer info.
139 // Delete the header search info, if we own it.
140 if (OwnsHeaderSearch)
146 void Preprocessor::setPTHManager(PTHManager* pm) {
148 FileMgr.addStatCache(PTH->createStatCache());
151 void Preprocessor::DumpToken(const Token &Tok, bool DumpFlags) const {
152 llvm::errs() << tok::getTokenName(Tok.getKind()) << " '"
153 << getSpelling(Tok) << "'";
155 if (!DumpFlags) return;
157 llvm::errs() << "\t";
158 if (Tok.isAtStartOfLine())
159 llvm::errs() << " [StartOfLine]";
160 if (Tok.hasLeadingSpace())
161 llvm::errs() << " [LeadingSpace]";
162 if (Tok.isExpandDisabled())
163 llvm::errs() << " [ExpandDisabled]";
164 if (Tok.needsCleaning()) {
165 const char *Start = SourceMgr.getCharacterData(Tok.getLocation());
166 llvm::errs() << " [UnClean='" << std::string(Start, Start+Tok.getLength())
170 llvm::errs() << "\tLoc=<";
171 DumpLocation(Tok.getLocation());
175 void Preprocessor::DumpLocation(SourceLocation Loc) const {
179 void Preprocessor::DumpMacro(const MacroInfo &MI) const {
180 llvm::errs() << "MACRO: ";
181 for (unsigned i = 0, e = MI.getNumTokens(); i != e; ++i) {
182 DumpToken(MI.getReplacementToken(i));
185 llvm::errs() << "\n";
188 void Preprocessor::PrintStats() {
189 llvm::errs() << "\n*** Preprocessor Stats:\n";
190 llvm::errs() << NumDirectives << " directives found:\n";
191 llvm::errs() << " " << NumDefined << " #define.\n";
192 llvm::errs() << " " << NumUndefined << " #undef.\n";
193 llvm::errs() << " #include/#include_next/#import:\n";
194 llvm::errs() << " " << NumEnteredSourceFiles << " source files entered.\n";
195 llvm::errs() << " " << MaxIncludeStackDepth << " max include stack depth\n";
196 llvm::errs() << " " << NumIf << " #if/#ifndef/#ifdef.\n";
197 llvm::errs() << " " << NumElse << " #else/#elif.\n";
198 llvm::errs() << " " << NumEndif << " #endif.\n";
199 llvm::errs() << " " << NumPragma << " #pragma.\n";
200 llvm::errs() << NumSkipped << " #if/#ifndef#ifdef regions skipped\n";
202 llvm::errs() << NumMacroExpanded << "/" << NumFnMacroExpanded << "/"
203 << NumBuiltinMacroExpanded << " obj/fn/builtin macros expanded, "
204 << NumFastMacroExpanded << " on the fast path.\n";
205 llvm::errs() << (NumFastTokenPaste+NumTokenPaste)
206 << " token paste (##) operations performed, "
207 << NumFastTokenPaste << " on the fast path.\n";
210 Preprocessor::macro_iterator
211 Preprocessor::macro_begin(bool IncludeExternalMacros) const {
212 if (IncludeExternalMacros && ExternalSource &&
213 !ReadMacrosFromExternalSource) {
214 ReadMacrosFromExternalSource = true;
215 ExternalSource->ReadDefinedMacros();
218 return Macros.begin();
221 Preprocessor::macro_iterator
222 Preprocessor::macro_end(bool IncludeExternalMacros) const {
223 if (IncludeExternalMacros && ExternalSource &&
224 !ReadMacrosFromExternalSource) {
225 ReadMacrosFromExternalSource = true;
226 ExternalSource->ReadDefinedMacros();
232 bool Preprocessor::SetCodeCompletionPoint(const FileEntry *File,
233 unsigned TruncateAtLine,
234 unsigned TruncateAtColumn) {
235 using llvm::MemoryBuffer;
237 CodeCompletionFile = File;
239 // Okay to clear out the code-completion point by passing NULL.
240 if (!CodeCompletionFile)
243 // Load the actual file's contents.
244 bool Invalid = false;
245 const MemoryBuffer *Buffer = SourceMgr.getMemoryBufferForFile(File, &Invalid);
249 // Find the byte position of the truncation point.
250 const char *Position = Buffer->getBufferStart();
251 for (unsigned Line = 1; Line < TruncateAtLine; ++Line) {
252 for (; *Position; ++Position) {
253 if (*Position != '\r' && *Position != '\n')
256 // Eat \r\n or \n\r as a single line.
257 if ((Position[1] == '\r' || Position[1] == '\n') &&
258 Position[0] != Position[1])
265 Position += TruncateAtColumn - 1;
267 // Truncate the buffer.
268 if (Position < Buffer->getBufferEnd()) {
269 llvm::StringRef Data(Buffer->getBufferStart(),
270 Position-Buffer->getBufferStart());
271 MemoryBuffer *TruncatedBuffer
272 = MemoryBuffer::getMemBufferCopy(Data, Buffer->getBufferIdentifier());
273 SourceMgr.overrideFileContents(File, TruncatedBuffer);
279 bool Preprocessor::isCodeCompletionFile(SourceLocation FileLoc) const {
280 return CodeCompletionFile && FileLoc.isFileID() &&
281 SourceMgr.getFileEntryForID(SourceMgr.getFileID(FileLoc))
282 == CodeCompletionFile;
285 //===----------------------------------------------------------------------===//
287 //===----------------------------------------------------------------------===//
289 /// getSpelling() - Return the 'spelling' of this token. The spelling of a
290 /// token are the characters used to represent the token in the source file
291 /// after trigraph expansion and escaped-newline folding. In particular, this
292 /// wants to get the true, uncanonicalized, spelling of things like digraphs
294 std::string Preprocessor::getSpelling(const Token &Tok,
295 const SourceManager &SourceMgr,
296 const LangOptions &Features,
298 assert((int)Tok.getLength() >= 0 && "Token character range is bogus!");
300 // If this token contains nothing interesting, return it directly.
301 bool CharDataInvalid = false;
302 const char* TokStart = SourceMgr.getCharacterData(Tok.getLocation(),
305 *Invalid = CharDataInvalid;
307 return std::string();
309 if (!Tok.needsCleaning())
310 return std::string(TokStart, TokStart+Tok.getLength());
313 Result.reserve(Tok.getLength());
315 // Otherwise, hard case, relex the characters into the string.
316 for (const char *Ptr = TokStart, *End = TokStart+Tok.getLength();
319 Result.push_back(Lexer::getCharAndSizeNoWarn(Ptr, CharSize, Features));
322 assert(Result.size() != unsigned(Tok.getLength()) &&
323 "NeedsCleaning flag set on something that didn't need cleaning!");
327 /// getSpelling() - Return the 'spelling' of this token. The spelling of a
328 /// token are the characters used to represent the token in the source file
329 /// after trigraph expansion and escaped-newline folding. In particular, this
330 /// wants to get the true, uncanonicalized, spelling of things like digraphs
332 std::string Preprocessor::getSpelling(const Token &Tok, bool *Invalid) const {
333 return getSpelling(Tok, SourceMgr, Features, Invalid);
336 /// getSpelling - This method is used to get the spelling of a token into a
337 /// preallocated buffer, instead of as an std::string. The caller is required
338 /// to allocate enough space for the token, which is guaranteed to be at least
339 /// Tok.getLength() bytes long. The actual length of the token is returned.
341 /// Note that this method may do two possible things: it may either fill in
342 /// the buffer specified with characters, or it may *change the input pointer*
343 /// to point to a constant buffer with the data already in it (avoiding a
344 /// copy). The caller is not allowed to modify the returned buffer pointer
345 /// if an internal buffer is returned.
346 unsigned Preprocessor::getSpelling(const Token &Tok,
347 const char *&Buffer, bool *Invalid) const {
348 assert((int)Tok.getLength() >= 0 && "Token character range is bogus!");
350 // If this token is an identifier, just return the string from the identifier
351 // table, which is very quick.
352 if (const IdentifierInfo *II = Tok.getIdentifierInfo()) {
353 Buffer = II->getNameStart();
354 return II->getLength();
357 // Otherwise, compute the start of the token in the input lexer buffer.
358 const char *TokStart = 0;
361 TokStart = Tok.getLiteralData();
364 bool CharDataInvalid = false;
365 TokStart = SourceMgr.getCharacterData(Tok.getLocation(), &CharDataInvalid);
367 *Invalid = CharDataInvalid;
368 if (CharDataInvalid) {
374 // If this token contains nothing interesting, return it directly.
375 if (!Tok.needsCleaning()) {
377 return Tok.getLength();
380 // Otherwise, hard case, relex the characters into the string.
381 char *OutBuf = const_cast<char*>(Buffer);
382 for (const char *Ptr = TokStart, *End = TokStart+Tok.getLength();
385 *OutBuf++ = Lexer::getCharAndSizeNoWarn(Ptr, CharSize, Features);
388 assert(unsigned(OutBuf-Buffer) != Tok.getLength() &&
389 "NeedsCleaning flag set on something that didn't need cleaning!");
391 return OutBuf-Buffer;
394 /// getSpelling - This method is used to get the spelling of a token into a
395 /// SmallVector. Note that the returned StringRef may not point to the
396 /// supplied buffer if a copy can be avoided.
397 llvm::StringRef Preprocessor::getSpelling(const Token &Tok,
398 llvm::SmallVectorImpl<char> &Buffer,
399 bool *Invalid) const {
400 // Try the fast path.
401 if (const IdentifierInfo *II = Tok.getIdentifierInfo())
402 return II->getName();
404 // Resize the buffer if we need to copy into it.
405 if (Tok.needsCleaning())
406 Buffer.resize(Tok.getLength());
408 const char *Ptr = Buffer.data();
409 unsigned Len = getSpelling(Tok, Ptr, Invalid);
410 return llvm::StringRef(Ptr, Len);
413 /// CreateString - Plop the specified string into a scratch buffer and return a
414 /// location for it. If specified, the source location provides a source
415 /// location for the token.
416 void Preprocessor::CreateString(const char *Buf, unsigned Len, Token &Tok,
417 SourceLocation InstantiationLoc) {
421 SourceLocation Loc = ScratchBuf->getToken(Buf, Len, DestPtr);
423 if (InstantiationLoc.isValid())
424 Loc = SourceMgr.createInstantiationLoc(Loc, InstantiationLoc,
425 InstantiationLoc, Len);
426 Tok.setLocation(Loc);
428 // If this is a literal token, set the pointer data.
430 Tok.setLiteralData(DestPtr);
434 /// AdvanceToTokenCharacter - Given a location that specifies the start of a
435 /// token, return a new location that specifies a character within the token.
436 SourceLocation Preprocessor::AdvanceToTokenCharacter(SourceLocation TokStart,
438 // Figure out how many physical characters away the specified instantiation
439 // character is. This needs to take into consideration newlines and
441 bool Invalid = false;
442 const char *TokPtr = SourceMgr.getCharacterData(TokStart, &Invalid);
444 // If they request the first char of the token, we're trivially done.
445 if (Invalid || (CharNo == 0 && Lexer::isObviouslySimpleCharacter(*TokPtr)))
448 unsigned PhysOffset = 0;
450 // The usual case is that tokens don't contain anything interesting. Skip
451 // over the uninteresting characters. If a token only consists of simple
452 // chars, this method is extremely fast.
453 while (Lexer::isObviouslySimpleCharacter(*TokPtr)) {
455 return TokStart.getFileLocWithOffset(PhysOffset);
456 ++TokPtr, --CharNo, ++PhysOffset;
459 // If we have a character that may be a trigraph or escaped newline, use a
460 // lexer to parse it correctly.
461 for (; CharNo; --CharNo) {
463 Lexer::getCharAndSizeNoWarn(TokPtr, Size, Features);
468 // Final detail: if we end up on an escaped newline, we want to return the
469 // location of the actual byte of the token. For example foo\<newline>bar
470 // advanced by 3 should return the location of b, not of \\. One compounding
471 // detail of this is that the escape may be made by a trigraph.
472 if (!Lexer::isObviouslySimpleCharacter(*TokPtr))
473 PhysOffset += Lexer::SkipEscapedNewLines(TokPtr)-TokPtr;
475 return TokStart.getFileLocWithOffset(PhysOffset);
478 SourceLocation Preprocessor::getLocForEndOfToken(SourceLocation Loc,
480 if (Loc.isInvalid() || !Loc.isFileID())
481 return SourceLocation();
483 unsigned Len = Lexer::MeasureTokenLength(Loc, getSourceManager(), Features);
489 return AdvanceToTokenCharacter(Loc, Len);
494 //===----------------------------------------------------------------------===//
495 // Preprocessor Initialization Methods
496 //===----------------------------------------------------------------------===//
499 /// EnterMainSourceFile - Enter the specified FileID as the main source file,
500 /// which implicitly adds the builtin defines etc.
501 void Preprocessor::EnterMainSourceFile() {
502 // We do not allow the preprocessor to reenter the main file. Doing so will
503 // cause FileID's to accumulate information from both runs (e.g. #line
504 // information) and predefined macros aren't guaranteed to be set properly.
505 assert(NumEnteredSourceFiles == 0 && "Cannot reenter the main file!");
506 FileID MainFileID = SourceMgr.getMainFileID();
508 // Enter the main file source buffer.
509 EnterSourceFile(MainFileID, 0, SourceLocation());
511 // Tell the header info that the main file was entered. If the file is later
512 // #imported, it won't be re-entered.
513 if (const FileEntry *FE = SourceMgr.getFileEntryForID(MainFileID))
514 HeaderInfo.IncrementIncludeCount(FE);
516 // Preprocess Predefines to populate the initial preprocessor state.
517 llvm::MemoryBuffer *SB =
518 llvm::MemoryBuffer::getMemBufferCopy(Predefines, "<built-in>");
519 assert(SB && "Cannot fail to create predefined source buffer");
520 FileID FID = SourceMgr.createFileIDForMemBuffer(SB);
521 assert(!FID.isInvalid() && "Could not create FileID for predefines?");
523 // Start parsing the predefines.
524 EnterSourceFile(FID, 0, SourceLocation());
527 void Preprocessor::EndSourceFile() {
528 // Notify the client that we reached the end of the source file.
530 Callbacks->EndOfMainFile();
533 //===----------------------------------------------------------------------===//
534 // Lexer Event Handling.
535 //===----------------------------------------------------------------------===//
537 /// LookUpIdentifierInfo - Given a tok::identifier token, look up the
538 /// identifier information for the token and install it into the token.
539 IdentifierInfo *Preprocessor::LookUpIdentifierInfo(Token &Identifier,
540 const char *BufPtr) const {
541 assert(Identifier.is(tok::identifier) && "Not an identifier!");
542 assert(Identifier.getIdentifierInfo() == 0 && "Identinfo already exists!");
544 // Look up this token, see if it is a macro, or if it is a language keyword.
546 if (BufPtr && !Identifier.needsCleaning()) {
547 // No cleaning needed, just use the characters from the lexed buffer.
548 II = getIdentifierInfo(llvm::StringRef(BufPtr, Identifier.getLength()));
550 // Cleaning needed, alloca a buffer, clean into it, then use the buffer.
551 llvm::SmallString<64> IdentifierBuffer;
552 llvm::StringRef CleanedStr = getSpelling(Identifier, IdentifierBuffer);
553 II = getIdentifierInfo(CleanedStr);
555 Identifier.setIdentifierInfo(II);
560 /// HandleIdentifier - This callback is invoked when the lexer reads an
561 /// identifier. This callback looks up the identifier in the map and/or
562 /// potentially macro expands it or turns it into a named token (like 'for').
564 /// Note that callers of this method are guarded by checking the
565 /// IdentifierInfo's 'isHandleIdentifierCase' bit. If this method changes, the
566 /// IdentifierInfo methods that compute these properties will need to change to
568 void Preprocessor::HandleIdentifier(Token &Identifier) {
569 assert(Identifier.getIdentifierInfo() &&
570 "Can't handle identifiers without identifier info!");
572 IdentifierInfo &II = *Identifier.getIdentifierInfo();
574 // If this identifier was poisoned, and if it was not produced from a macro
575 // expansion, emit an error.
576 if (II.isPoisoned() && CurPPLexer) {
577 if (&II != Ident__VA_ARGS__) // We warn about __VA_ARGS__ with poisoning.
578 Diag(Identifier, diag::err_pp_used_poisoned_id);
580 Diag(Identifier, diag::ext_pp_bad_vaargs_use);
583 // If this is a macro to be expanded, do it.
584 if (MacroInfo *MI = getMacroInfo(&II)) {
585 if (!DisableMacroExpansion && !Identifier.isExpandDisabled()) {
586 if (MI->isEnabled()) {
587 if (!HandleMacroExpandedIdentifier(Identifier, MI))
590 // C99 6.10.3.4p2 says that a disabled macro may never again be
591 // expanded, even if it's in a context where it could be expanded in the
593 Identifier.setFlag(Token::DisableExpand);
598 // C++ 2.11p2: If this is an alternative representation of a C++ operator,
599 // then we act as if it is the actual operator and not the textual
600 // representation of it.
601 if (II.isCPlusPlusOperatorKeyword())
602 Identifier.setIdentifierInfo(0);
604 // If this is an extension token, diagnose its use.
605 // We avoid diagnosing tokens that originate from macro definitions.
606 // FIXME: This warning is disabled in cases where it shouldn't be,
607 // like "#define TY typeof", "TY(1) x".
608 if (II.isExtensionToken() && !DisableMacroExpansion)
609 Diag(Identifier, diag::ext_token_used);
612 void Preprocessor::AddCommentHandler(CommentHandler *Handler) {
613 assert(Handler && "NULL comment handler");
614 assert(std::find(CommentHandlers.begin(), CommentHandlers.end(), Handler) ==
615 CommentHandlers.end() && "Comment handler already registered");
616 CommentHandlers.push_back(Handler);
619 void Preprocessor::RemoveCommentHandler(CommentHandler *Handler) {
620 std::vector<CommentHandler *>::iterator Pos
621 = std::find(CommentHandlers.begin(), CommentHandlers.end(), Handler);
622 assert(Pos != CommentHandlers.end() && "Comment handler not registered");
623 CommentHandlers.erase(Pos);
626 bool Preprocessor::HandleComment(Token &result, SourceRange Comment) {
627 bool AnyPendingTokens = false;
628 for (std::vector<CommentHandler *>::iterator H = CommentHandlers.begin(),
629 HEnd = CommentHandlers.end();
631 if ((*H)->HandleComment(*this, Comment))
632 AnyPendingTokens = true;
634 if (!AnyPendingTokens || getCommentRetentionState())
640 CommentHandler::~CommentHandler() { }
642 void Preprocessor::createPreprocessingRecord() {
646 Record = new PreprocessingRecord;
647 addPPCallbacks(Record);