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/HeaderSearch.h"
31 #include "clang/Lex/MacroInfo.h"
32 #include "clang/Lex/Pragma.h"
33 #include "clang/Lex/ScratchBuffer.h"
34 #include "clang/Lex/LexDiagnostic.h"
35 #include "clang/Basic/SourceManager.h"
36 #include "clang/Basic/FileManager.h"
37 #include "clang/Basic/TargetInfo.h"
38 #include "llvm/ADT/APFloat.h"
39 #include "llvm/ADT/SmallVector.h"
40 #include "llvm/Support/MemoryBuffer.h"
41 #include "llvm/Support/raw_ostream.h"
43 using namespace clang;
45 //===----------------------------------------------------------------------===//
47 Preprocessor::Preprocessor(Diagnostic &diags, const LangOptions &opts,
48 const TargetInfo &target, SourceManager &SM,
49 HeaderSearch &Headers,
50 IdentifierInfoLookup* IILookup,
52 : Diags(&diags), Features(opts), Target(target),FileMgr(Headers.getFileMgr()),
53 SourceMgr(SM), HeaderInfo(Headers), Identifiers(opts, IILookup),
54 BuiltinInfo(Target), CodeCompletionFile(0), CurPPLexer(0), CurDirLookup(0),
55 Callbacks(0), MacroArgCache(0) {
56 ScratchBuf = new ScratchBuffer(SourceMgr);
57 CounterValue = 0; // __COUNTER__ starts at 0.
58 OwnsHeaderSearch = OwnsHeaders;
61 NumDirectives = NumDefined = NumUndefined = NumPragma = 0;
62 NumIf = NumElse = NumEndif = 0;
63 NumEnteredSourceFiles = 0;
64 NumMacroExpanded = NumFnMacroExpanded = NumBuiltinMacroExpanded = 0;
65 NumFastMacroExpanded = NumTokenPaste = NumFastTokenPaste = 0;
66 MaxIncludeStackDepth = 0;
69 // Default to discarding comments.
71 KeepMacroComments = false;
73 // Macro expansion is enabled.
74 DisableMacroExpansion = false;
76 NumCachedTokenLexers = 0;
80 // "Poison" __VA_ARGS__, which can only appear in the expansion of a macro.
81 // This gets unpoisoned where it is allowed.
82 (Ident__VA_ARGS__ = getIdentifierInfo("__VA_ARGS__"))->setIsPoisoned();
84 // Initialize the pragma handlers.
85 PragmaHandlers = new PragmaNamespace(0);
86 RegisterBuiltinPragmas();
88 // Initialize builtin macros like __LINE__ and friends.
89 RegisterBuiltinMacros();
92 Preprocessor::~Preprocessor() {
93 assert(BacktrackPositions.empty() && "EnableBacktrack/Backtrack imbalance!");
95 while (!IncludeMacroStack.empty()) {
96 delete IncludeMacroStack.back().TheLexer;
97 delete IncludeMacroStack.back().TheTokenLexer;
98 IncludeMacroStack.pop_back();
101 // Free any macro definitions.
102 for (llvm::DenseMap<IdentifierInfo*, MacroInfo*>::iterator I =
103 Macros.begin(), E = Macros.end(); I != E; ++I) {
104 // We don't need to free the MacroInfo objects directly. These
105 // will be released when the BumpPtrAllocator 'BP' object gets
106 // destroyed. We still need to run the dtor, however, to free
107 // memory alocated by MacroInfo.
108 I->second->Destroy(BP);
109 I->first->setHasMacroDefinition(false);
112 // Free any cached macro expanders.
113 for (unsigned i = 0, e = NumCachedTokenLexers; i != e; ++i)
114 delete TokenLexerCache[i];
116 // Free any cached MacroArgs.
117 for (MacroArgs *ArgList = MacroArgCache; ArgList; )
118 ArgList = ArgList->deallocate();
120 // Release pragma information.
121 delete PragmaHandlers;
123 // Delete the scratch buffer info.
126 // Delete the header search info, if we own it.
127 if (OwnsHeaderSearch)
133 void Preprocessor::setPTHManager(PTHManager* pm) {
135 FileMgr.addStatCache(PTH->createStatCache());
138 void Preprocessor::DumpToken(const Token &Tok, bool DumpFlags) const {
139 llvm::errs() << tok::getTokenName(Tok.getKind()) << " '"
140 << getSpelling(Tok) << "'";
142 if (!DumpFlags) return;
144 llvm::errs() << "\t";
145 if (Tok.isAtStartOfLine())
146 llvm::errs() << " [StartOfLine]";
147 if (Tok.hasLeadingSpace())
148 llvm::errs() << " [LeadingSpace]";
149 if (Tok.isExpandDisabled())
150 llvm::errs() << " [ExpandDisabled]";
151 if (Tok.needsCleaning()) {
152 const char *Start = SourceMgr.getCharacterData(Tok.getLocation());
153 llvm::errs() << " [UnClean='" << std::string(Start, Start+Tok.getLength())
157 llvm::errs() << "\tLoc=<";
158 DumpLocation(Tok.getLocation());
162 void Preprocessor::DumpLocation(SourceLocation Loc) const {
166 void Preprocessor::DumpMacro(const MacroInfo &MI) const {
167 llvm::errs() << "MACRO: ";
168 for (unsigned i = 0, e = MI.getNumTokens(); i != e; ++i) {
169 DumpToken(MI.getReplacementToken(i));
172 llvm::errs() << "\n";
175 void Preprocessor::PrintStats() {
176 llvm::errs() << "\n*** Preprocessor Stats:\n";
177 llvm::errs() << NumDirectives << " directives found:\n";
178 llvm::errs() << " " << NumDefined << " #define.\n";
179 llvm::errs() << " " << NumUndefined << " #undef.\n";
180 llvm::errs() << " #include/#include_next/#import:\n";
181 llvm::errs() << " " << NumEnteredSourceFiles << " source files entered.\n";
182 llvm::errs() << " " << MaxIncludeStackDepth << " max include stack depth\n";
183 llvm::errs() << " " << NumIf << " #if/#ifndef/#ifdef.\n";
184 llvm::errs() << " " << NumElse << " #else/#elif.\n";
185 llvm::errs() << " " << NumEndif << " #endif.\n";
186 llvm::errs() << " " << NumPragma << " #pragma.\n";
187 llvm::errs() << NumSkipped << " #if/#ifndef#ifdef regions skipped\n";
189 llvm::errs() << NumMacroExpanded << "/" << NumFnMacroExpanded << "/"
190 << NumBuiltinMacroExpanded << " obj/fn/builtin macros expanded, "
191 << NumFastMacroExpanded << " on the fast path.\n";
192 llvm::errs() << (NumFastTokenPaste+NumTokenPaste)
193 << " token paste (##) operations performed, "
194 << NumFastTokenPaste << " on the fast path.\n";
197 bool Preprocessor::SetCodeCompletionPoint(const FileEntry *File,
198 unsigned TruncateAtLine,
199 unsigned TruncateAtColumn) {
200 using llvm::MemoryBuffer;
202 CodeCompletionFile = File;
204 // Okay to clear out the code-completion point by passing NULL.
205 if (!CodeCompletionFile)
208 // Load the actual file's contents.
209 const MemoryBuffer *Buffer = SourceMgr.getMemoryBufferForFile(File);
213 // Find the byte position of the truncation point.
214 const char *Position = Buffer->getBufferStart();
215 for (unsigned Line = 1; Line < TruncateAtLine; ++Line) {
216 for (; *Position; ++Position) {
217 if (*Position != '\r' && *Position != '\n')
220 // Eat \r\n or \n\r as a single line.
221 if ((Position[1] == '\r' || Position[1] == '\n') &&
222 Position[0] != Position[1])
229 Position += TruncateAtColumn - 1;
231 // Truncate the buffer.
232 if (Position < Buffer->getBufferEnd()) {
233 MemoryBuffer *TruncatedBuffer
234 = MemoryBuffer::getMemBufferCopy(Buffer->getBufferStart(), Position,
235 Buffer->getBufferIdentifier());
236 SourceMgr.overrideFileContents(File, TruncatedBuffer);
242 bool Preprocessor::isCodeCompletionFile(SourceLocation FileLoc) const {
243 return CodeCompletionFile && FileLoc.isFileID() &&
244 SourceMgr.getFileEntryForID(SourceMgr.getFileID(FileLoc))
245 == CodeCompletionFile;
248 //===----------------------------------------------------------------------===//
250 //===----------------------------------------------------------------------===//
252 /// getSpelling() - Return the 'spelling' of this token. The spelling of a
253 /// token are the characters used to represent the token in the source file
254 /// after trigraph expansion and escaped-newline folding. In particular, this
255 /// wants to get the true, uncanonicalized, spelling of things like digraphs
257 std::string Preprocessor::getSpelling(const Token &Tok,
258 const SourceManager &SourceMgr,
259 const LangOptions &Features) {
260 assert((int)Tok.getLength() >= 0 && "Token character range is bogus!");
262 // If this token contains nothing interesting, return it directly.
263 const char* TokStart = SourceMgr.getCharacterData(Tok.getLocation());
264 if (!Tok.needsCleaning())
265 return std::string(TokStart, TokStart+Tok.getLength());
268 Result.reserve(Tok.getLength());
270 // Otherwise, hard case, relex the characters into the string.
271 for (const char *Ptr = TokStart, *End = TokStart+Tok.getLength();
274 Result.push_back(Lexer::getCharAndSizeNoWarn(Ptr, CharSize, Features));
277 assert(Result.size() != unsigned(Tok.getLength()) &&
278 "NeedsCleaning flag set on something that didn't need cleaning!");
282 /// getSpelling() - Return the 'spelling' of this token. The spelling of a
283 /// token are the characters used to represent the token in the source file
284 /// after trigraph expansion and escaped-newline folding. In particular, this
285 /// wants to get the true, uncanonicalized, spelling of things like digraphs
287 std::string Preprocessor::getSpelling(const Token &Tok) const {
288 return getSpelling(Tok, SourceMgr, Features);
291 /// getSpelling - This method is used to get the spelling of a token into a
292 /// preallocated buffer, instead of as an std::string. The caller is required
293 /// to allocate enough space for the token, which is guaranteed to be at least
294 /// Tok.getLength() bytes long. The actual length of the token is returned.
296 /// Note that this method may do two possible things: it may either fill in
297 /// the buffer specified with characters, or it may *change the input pointer*
298 /// to point to a constant buffer with the data already in it (avoiding a
299 /// copy). The caller is not allowed to modify the returned buffer pointer
300 /// if an internal buffer is returned.
301 unsigned Preprocessor::getSpelling(const Token &Tok,
302 const char *&Buffer) const {
303 assert((int)Tok.getLength() >= 0 && "Token character range is bogus!");
305 // If this token is an identifier, just return the string from the identifier
306 // table, which is very quick.
307 if (const IdentifierInfo *II = Tok.getIdentifierInfo()) {
308 Buffer = II->getNameStart();
309 return II->getLength();
312 // Otherwise, compute the start of the token in the input lexer buffer.
313 const char *TokStart = 0;
316 TokStart = Tok.getLiteralData();
319 TokStart = SourceMgr.getCharacterData(Tok.getLocation());
321 // If this token contains nothing interesting, return it directly.
322 if (!Tok.needsCleaning()) {
324 return Tok.getLength();
327 // Otherwise, hard case, relex the characters into the string.
328 char *OutBuf = const_cast<char*>(Buffer);
329 for (const char *Ptr = TokStart, *End = TokStart+Tok.getLength();
332 *OutBuf++ = Lexer::getCharAndSizeNoWarn(Ptr, CharSize, Features);
335 assert(unsigned(OutBuf-Buffer) != Tok.getLength() &&
336 "NeedsCleaning flag set on something that didn't need cleaning!");
338 return OutBuf-Buffer;
341 /// CreateString - Plop the specified string into a scratch buffer and return a
342 /// location for it. If specified, the source location provides a source
343 /// location for the token.
344 void Preprocessor::CreateString(const char *Buf, unsigned Len, Token &Tok,
345 SourceLocation InstantiationLoc) {
349 SourceLocation Loc = ScratchBuf->getToken(Buf, Len, DestPtr);
351 if (InstantiationLoc.isValid())
352 Loc = SourceMgr.createInstantiationLoc(Loc, InstantiationLoc,
353 InstantiationLoc, Len);
354 Tok.setLocation(Loc);
356 // If this is a literal token, set the pointer data.
358 Tok.setLiteralData(DestPtr);
362 /// AdvanceToTokenCharacter - Given a location that specifies the start of a
363 /// token, return a new location that specifies a character within the token.
364 SourceLocation Preprocessor::AdvanceToTokenCharacter(SourceLocation TokStart,
366 // Figure out how many physical characters away the specified instantiation
367 // character is. This needs to take into consideration newlines and
369 const char *TokPtr = SourceMgr.getCharacterData(TokStart);
371 // If they request the first char of the token, we're trivially done.
372 if (CharNo == 0 && Lexer::isObviouslySimpleCharacter(*TokPtr))
375 unsigned PhysOffset = 0;
377 // The usual case is that tokens don't contain anything interesting. Skip
378 // over the uninteresting characters. If a token only consists of simple
379 // chars, this method is extremely fast.
380 while (Lexer::isObviouslySimpleCharacter(*TokPtr)) {
382 return TokStart.getFileLocWithOffset(PhysOffset);
383 ++TokPtr, --CharNo, ++PhysOffset;
386 // If we have a character that may be a trigraph or escaped newline, use a
387 // lexer to parse it correctly.
388 for (; CharNo; --CharNo) {
390 Lexer::getCharAndSizeNoWarn(TokPtr, Size, Features);
395 // Final detail: if we end up on an escaped newline, we want to return the
396 // location of the actual byte of the token. For example foo\<newline>bar
397 // advanced by 3 should return the location of b, not of \\. One compounding
398 // detail of this is that the escape may be made by a trigraph.
399 if (!Lexer::isObviouslySimpleCharacter(*TokPtr))
400 PhysOffset = Lexer::SkipEscapedNewLines(TokPtr)-TokPtr;
402 return TokStart.getFileLocWithOffset(PhysOffset);
405 /// \brief Computes the source location just past the end of the
406 /// token at this source location.
408 /// This routine can be used to produce a source location that
409 /// points just past the end of the token referenced by \p Loc, and
410 /// is generally used when a diagnostic needs to point just after a
411 /// token where it expected something different that it received. If
412 /// the returned source location would not be meaningful (e.g., if
413 /// it points into a macro), this routine returns an invalid
415 SourceLocation Preprocessor::getLocForEndOfToken(SourceLocation Loc) {
416 if (Loc.isInvalid() || !Loc.isFileID())
417 return SourceLocation();
419 unsigned Len = Lexer::MeasureTokenLength(Loc, getSourceManager(), Features);
420 return AdvanceToTokenCharacter(Loc, Len);
425 //===----------------------------------------------------------------------===//
426 // Preprocessor Initialization Methods
427 //===----------------------------------------------------------------------===//
430 /// EnterMainSourceFile - Enter the specified FileID as the main source file,
431 /// which implicitly adds the builtin defines etc.
432 void Preprocessor::EnterMainSourceFile() {
433 // We do not allow the preprocessor to reenter the main file. Doing so will
434 // cause FileID's to accumulate information from both runs (e.g. #line
435 // information) and predefined macros aren't guaranteed to be set properly.
436 assert(NumEnteredSourceFiles == 0 && "Cannot reenter the main file!");
437 FileID MainFileID = SourceMgr.getMainFileID();
439 // Enter the main file source buffer.
440 std::string ErrorStr;
441 bool Res = EnterSourceFile(MainFileID, 0, ErrorStr);
442 assert(!Res && "Entering main file should not fail!");
444 // Tell the header info that the main file was entered. If the file is later
445 // #imported, it won't be re-entered.
446 if (const FileEntry *FE = SourceMgr.getFileEntryForID(MainFileID))
447 HeaderInfo.IncrementIncludeCount(FE);
449 std::vector<char> PrologFile;
450 PrologFile.reserve(4080);
452 // FIXME: Don't make a copy.
453 PrologFile.insert(PrologFile.end(), Predefines.begin(), Predefines.end());
455 // Memory buffer must end with a null byte!
456 PrologFile.push_back(0);
458 // Now that we have emitted the predefined macros, #includes, etc into
459 // PrologFile, preprocess it to populate the initial preprocessor state.
460 llvm::MemoryBuffer *SB =
461 llvm::MemoryBuffer::getMemBufferCopy(&PrologFile.front(),&PrologFile.back(),
463 assert(SB && "Cannot fail to create predefined source buffer");
464 FileID FID = SourceMgr.createFileIDForMemBuffer(SB);
465 assert(!FID.isInvalid() && "Could not create FileID for predefines?");
467 // Start parsing the predefines.
468 Res = EnterSourceFile(FID, 0, ErrorStr);
469 assert(!Res && "Entering predefines should not fail!");
473 //===----------------------------------------------------------------------===//
474 // Lexer Event Handling.
475 //===----------------------------------------------------------------------===//
477 /// LookUpIdentifierInfo - Given a tok::identifier token, look up the
478 /// identifier information for the token and install it into the token.
479 IdentifierInfo *Preprocessor::LookUpIdentifierInfo(Token &Identifier,
480 const char *BufPtr) const {
481 assert(Identifier.is(tok::identifier) && "Not an identifier!");
482 assert(Identifier.getIdentifierInfo() == 0 && "Identinfo already exists!");
484 // Look up this token, see if it is a macro, or if it is a language keyword.
486 if (BufPtr && !Identifier.needsCleaning()) {
487 // No cleaning needed, just use the characters from the lexed buffer.
488 II = getIdentifierInfo(llvm::StringRef(BufPtr, Identifier.getLength()));
490 // Cleaning needed, alloca a buffer, clean into it, then use the buffer.
491 llvm::SmallVector<char, 64> IdentifierBuffer;
492 IdentifierBuffer.resize(Identifier.getLength());
493 const char *TmpBuf = &IdentifierBuffer[0];
494 unsigned Size = getSpelling(Identifier, TmpBuf);
495 II = getIdentifierInfo(llvm::StringRef(TmpBuf, Size));
497 Identifier.setIdentifierInfo(II);
502 /// HandleIdentifier - This callback is invoked when the lexer reads an
503 /// identifier. This callback looks up the identifier in the map and/or
504 /// potentially macro expands it or turns it into a named token (like 'for').
506 /// Note that callers of this method are guarded by checking the
507 /// IdentifierInfo's 'isHandleIdentifierCase' bit. If this method changes, the
508 /// IdentifierInfo methods that compute these properties will need to change to
510 void Preprocessor::HandleIdentifier(Token &Identifier) {
511 assert(Identifier.getIdentifierInfo() &&
512 "Can't handle identifiers without identifier info!");
514 IdentifierInfo &II = *Identifier.getIdentifierInfo();
516 // If this identifier was poisoned, and if it was not produced from a macro
517 // expansion, emit an error.
518 if (II.isPoisoned() && CurPPLexer) {
519 if (&II != Ident__VA_ARGS__) // We warn about __VA_ARGS__ with poisoning.
520 Diag(Identifier, diag::err_pp_used_poisoned_id);
522 Diag(Identifier, diag::ext_pp_bad_vaargs_use);
525 // If this is a macro to be expanded, do it.
526 if (MacroInfo *MI = getMacroInfo(&II)) {
527 if (!DisableMacroExpansion && !Identifier.isExpandDisabled()) {
528 if (MI->isEnabled()) {
529 if (!HandleMacroExpandedIdentifier(Identifier, MI))
532 // C99 6.10.3.4p2 says that a disabled macro may never again be
533 // expanded, even if it's in a context where it could be expanded in the
535 Identifier.setFlag(Token::DisableExpand);
540 // C++ 2.11p2: If this is an alternative representation of a C++ operator,
541 // then we act as if it is the actual operator and not the textual
542 // representation of it.
543 if (II.isCPlusPlusOperatorKeyword())
544 Identifier.setIdentifierInfo(0);
546 // If this is an extension token, diagnose its use.
547 // We avoid diagnosing tokens that originate from macro definitions.
548 // FIXME: This warning is disabled in cases where it shouldn't be,
549 // like "#define TY typeof", "TY(1) x".
550 if (II.isExtensionToken() && !DisableMacroExpansion)
551 Diag(Identifier, diag::ext_token_used);
554 void Preprocessor::AddCommentHandler(CommentHandler *Handler) {
555 assert(Handler && "NULL comment handler");
556 assert(std::find(CommentHandlers.begin(), CommentHandlers.end(), Handler) ==
557 CommentHandlers.end() && "Comment handler already registered");
558 CommentHandlers.push_back(Handler);
561 void Preprocessor::RemoveCommentHandler(CommentHandler *Handler) {
562 std::vector<CommentHandler *>::iterator Pos
563 = std::find(CommentHandlers.begin(), CommentHandlers.end(), Handler);
564 assert(Pos != CommentHandlers.end() && "Comment handler not registered");
565 CommentHandlers.erase(Pos);
568 void Preprocessor::HandleComment(SourceRange Comment) {
569 for (std::vector<CommentHandler *>::iterator H = CommentHandlers.begin(),
570 HEnd = CommentHandlers.end();
572 (*H)->HandleComment(*this, Comment);
575 CommentHandler::~CommentHandler() { }