1 //===--- Preprocess.cpp - C Language Family Preprocessor Implementation ---===//
2 //
3 // The LLVM Compiler Infrastructure
4 //
5 // This file is distributed under the University of Illinois Open Source
6 // License. See LICENSE.TXT for details.
7 //
8 //===----------------------------------------------------------------------===//
9 //
10 // This file implements the Preprocessor interface.
11 //
12 //===----------------------------------------------------------------------===//
13 //
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.
18 // -fpreprocessed
19 // -dependency-file,-M,-MM,-MF,-MG,-MP,-MT,-MQ,-MD,-MMD
20 // -W*
21 // -w
22 //
23 // Messages to emit:
24 // "Multiple include guards may be useful for:\n"
25 //
26 //===----------------------------------------------------------------------===//
27
28 #include "clang/Lex/Preprocessor.h"
29 #include "clang/Lex/MacroArgs.h"
30 #include "clang/Basic/FileManager.h"
31 #include "clang/Basic/SourceManager.h"
32 #include "clang/Basic/TargetInfo.h"
33 #include "clang/Lex/CodeCompletionHandler.h"
34 #include "clang/Lex/ExternalPreprocessorSource.h"
35 #include "clang/Lex/HeaderSearch.h"
36 #include "clang/Lex/LexDiagnostic.h"
37 #include "clang/Lex/LiteralSupport.h"
38 #include "clang/Lex/MacroInfo.h"
39 #include "clang/Lex/ModuleLoader.h"
40 #include "clang/Lex/Pragma.h"
41 #include "clang/Lex/PreprocessingRecord.h"
42 #include "clang/Lex/PreprocessorOptions.h"
43 #include "clang/Lex/ScratchBuffer.h"
44 #include "llvm/ADT/APFloat.h"
45 #include "llvm/ADT/SmallString.h"
46 #include "llvm/ADT/STLExtras.h"
47 #include "llvm/ADT/StringExtras.h"
48 #include "llvm/Support/Capacity.h"
49 #include "llvm/Support/ConvertUTF.h"
50 #include "llvm/Support/MemoryBuffer.h"
51 #include "llvm/Support/raw_ostream.h"
52 using namespace clang;
53
54 //===----------------------------------------------------------------------===//
~ExternalPreprocessorSource()55 ExternalPreprocessorSource::~ExternalPreprocessorSource() { }
56
Preprocessor(IntrusiveRefCntPtr<PreprocessorOptions> PPOpts,DiagnosticsEngine & diags,LangOptions & opts,const TargetInfo * target,SourceManager & SM,HeaderSearch & Headers,ModuleLoader & TheModuleLoader,IdentifierInfoLookup * IILookup,bool OwnsHeaders,bool DelayInitialization,bool IncrProcessing)57 Preprocessor::Preprocessor(IntrusiveRefCntPtr<PreprocessorOptions> PPOpts,
58 DiagnosticsEngine &diags, LangOptions &opts,
59 const TargetInfo *target, SourceManager &SM,
60 HeaderSearch &Headers, ModuleLoader &TheModuleLoader,
61 IdentifierInfoLookup *IILookup, bool OwnsHeaders,
62 bool DelayInitialization, bool IncrProcessing)
63 : PPOpts(PPOpts), Diags(&diags), LangOpts(opts), Target(target),
64 FileMgr(Headers.getFileMgr()), SourceMgr(SM), HeaderInfo(Headers),
65 TheModuleLoader(TheModuleLoader), ExternalSource(0),
66 Identifiers(opts, IILookup), IncrementalProcessing(IncrProcessing),
67 CodeComplete(0), CodeCompletionFile(0), CodeCompletionOffset(0),
68 LastTokenWasAt(false), ModuleImportExpectsIdentifier(false),
69 CodeCompletionReached(0), SkipMainFilePreamble(0, true), CurPPLexer(0),
70 CurDirLookup(0), CurLexerKind(CLK_Lexer), Callbacks(0),
71 MacroArgCache(0), Record(0), MIChainHead(0), MICache(0),
72 DeserialMIChainHead(0) {
73 OwnsHeaderSearch = OwnsHeaders;
74
75 ScratchBuf = new ScratchBuffer(SourceMgr);
76 CounterValue = 0; // __COUNTER__ starts at 0.
77
78 // Clear stats.
79 NumDirectives = NumDefined = NumUndefined = NumPragma = 0;
80 NumIf = NumElse = NumEndif = 0;
81 NumEnteredSourceFiles = 0;
82 NumMacroExpanded = NumFnMacroExpanded = NumBuiltinMacroExpanded = 0;
83 NumFastMacroExpanded = NumTokenPaste = NumFastTokenPaste = 0;
84 MaxIncludeStackDepth = 0;
85 NumSkipped = 0;
86
87 // Default to discarding comments.
88 KeepComments = false;
89 KeepMacroComments = false;
90 SuppressIncludeNotFoundError = false;
91
92 // Macro expansion is enabled.
93 DisableMacroExpansion = false;
94 MacroExpansionInDirectivesOverride = false;
95 InMacroArgs = false;
96 InMacroArgPreExpansion = false;
97 NumCachedTokenLexers = 0;
98 PragmasEnabled = true;
99 ParsingIfOrElifDirective = false;
100 PreprocessedOutput = false;
101
102 CachedLexPos = 0;
103
104 // We haven't read anything from the external source.
105 ReadMacrosFromExternalSource = false;
106
107 // "Poison" __VA_ARGS__, which can only appear in the expansion of a macro.
108 // This gets unpoisoned where it is allowed.
109 (Ident__VA_ARGS__ = getIdentifierInfo("__VA_ARGS__"))->setIsPoisoned();
110 SetPoisonReason(Ident__VA_ARGS__,diag::ext_pp_bad_vaargs_use);
111
112 // Initialize the pragma handlers.
113 PragmaHandlers = new PragmaNamespace(StringRef());
114 RegisterBuiltinPragmas();
115
116 // Initialize builtin macros like __LINE__ and friends.
117 RegisterBuiltinMacros();
118
119 if(LangOpts.Borland) {
120 Ident__exception_info = getIdentifierInfo("_exception_info");
121 Ident___exception_info = getIdentifierInfo("__exception_info");
122 Ident_GetExceptionInfo = getIdentifierInfo("GetExceptionInformation");
123 Ident__exception_code = getIdentifierInfo("_exception_code");
124 Ident___exception_code = getIdentifierInfo("__exception_code");
125 Ident_GetExceptionCode = getIdentifierInfo("GetExceptionCode");
126 Ident__abnormal_termination = getIdentifierInfo("_abnormal_termination");
127 Ident___abnormal_termination = getIdentifierInfo("__abnormal_termination");
128 Ident_AbnormalTermination = getIdentifierInfo("AbnormalTermination");
129 } else {
130 Ident__exception_info = Ident__exception_code = Ident__abnormal_termination = 0;
131 Ident___exception_info = Ident___exception_code = Ident___abnormal_termination = 0;
132 Ident_GetExceptionInfo = Ident_GetExceptionCode = Ident_AbnormalTermination = 0;
133 }
134
135 if (!DelayInitialization) {
136 assert(Target && "Must provide target information for PP initialization");
137 Initialize(*Target);
138 }
139 }
140
~Preprocessor()141 Preprocessor::~Preprocessor() {
142 assert(BacktrackPositions.empty() && "EnableBacktrack/Backtrack imbalance!");
143
144 while (!IncludeMacroStack.empty()) {
145 delete IncludeMacroStack.back().TheLexer;
146 delete IncludeMacroStack.back().TheTokenLexer;
147 IncludeMacroStack.pop_back();
148 }
149
150 // Free any macro definitions.
151 for (MacroInfoChain *I = MIChainHead ; I ; I = I->Next)
152 I->MI.Destroy();
153
154 // Free any cached macro expanders.
155 for (unsigned i = 0, e = NumCachedTokenLexers; i != e; ++i)
156 delete TokenLexerCache[i];
157
158 for (DeserializedMacroInfoChain *I = DeserialMIChainHead ; I ; I = I->Next)
159 I->MI.Destroy();
160
161 // Free any cached MacroArgs.
162 for (MacroArgs *ArgList = MacroArgCache; ArgList; )
163 ArgList = ArgList->deallocate();
164
165 // Release pragma information.
166 delete PragmaHandlers;
167
168 // Delete the scratch buffer info.
169 delete ScratchBuf;
170
171 // Delete the header search info, if we own it.
172 if (OwnsHeaderSearch)
173 delete &HeaderInfo;
174
175 delete Callbacks;
176 }
177
Initialize(const TargetInfo & Target)178 void Preprocessor::Initialize(const TargetInfo &Target) {
179 assert((!this->Target || this->Target == &Target) &&
180 "Invalid override of target information");
181 this->Target = &Target;
182
183 // Initialize information about built-ins.
184 BuiltinInfo.InitializeTarget(Target);
185 HeaderInfo.setTarget(Target);
186 }
187
setPTHManager(PTHManager * pm)188 void Preprocessor::setPTHManager(PTHManager* pm) {
189 PTH.reset(pm);
190 FileMgr.addStatCache(PTH->createStatCache());
191 }
192
DumpToken(const Token & Tok,bool DumpFlags) const193 void Preprocessor::DumpToken(const Token &Tok, bool DumpFlags) const {
194 llvm::errs() << tok::getTokenName(Tok.getKind()) << " '"
195 << getSpelling(Tok) << "'";
196
197 if (!DumpFlags) return;
198
199 llvm::errs() << "\t";
200 if (Tok.isAtStartOfLine())
201 llvm::errs() << " [StartOfLine]";
202 if (Tok.hasLeadingSpace())
203 llvm::errs() << " [LeadingSpace]";
204 if (Tok.isExpandDisabled())
205 llvm::errs() << " [ExpandDisabled]";
206 if (Tok.needsCleaning()) {
207 const char *Start = SourceMgr.getCharacterData(Tok.getLocation());
208 llvm::errs() << " [UnClean='" << StringRef(Start, Tok.getLength())
209 << "']";
210 }
211
212 llvm::errs() << "\tLoc=<";
213 DumpLocation(Tok.getLocation());
214 llvm::errs() << ">";
215 }
216
DumpLocation(SourceLocation Loc) const217 void Preprocessor::DumpLocation(SourceLocation Loc) const {
218 Loc.dump(SourceMgr);
219 }
220
DumpMacro(const MacroInfo & MI) const221 void Preprocessor::DumpMacro(const MacroInfo &MI) const {
222 llvm::errs() << "MACRO: ";
223 for (unsigned i = 0, e = MI.getNumTokens(); i != e; ++i) {
224 DumpToken(MI.getReplacementToken(i));
225 llvm::errs() << " ";
226 }
227 llvm::errs() << "\n";
228 }
229
PrintStats()230 void Preprocessor::PrintStats() {
231 llvm::errs() << "\n*** Preprocessor Stats:\n";
232 llvm::errs() << NumDirectives << " directives found:\n";
233 llvm::errs() << " " << NumDefined << " #define.\n";
234 llvm::errs() << " " << NumUndefined << " #undef.\n";
235 llvm::errs() << " #include/#include_next/#import:\n";
236 llvm::errs() << " " << NumEnteredSourceFiles << " source files entered.\n";
237 llvm::errs() << " " << MaxIncludeStackDepth << " max include stack depth\n";
238 llvm::errs() << " " << NumIf << " #if/#ifndef/#ifdef.\n";
239 llvm::errs() << " " << NumElse << " #else/#elif.\n";
240 llvm::errs() << " " << NumEndif << " #endif.\n";
241 llvm::errs() << " " << NumPragma << " #pragma.\n";
242 llvm::errs() << NumSkipped << " #if/#ifndef#ifdef regions skipped\n";
243
244 llvm::errs() << NumMacroExpanded << "/" << NumFnMacroExpanded << "/"
245 << NumBuiltinMacroExpanded << " obj/fn/builtin macros expanded, "
246 << NumFastMacroExpanded << " on the fast path.\n";
247 llvm::errs() << (NumFastTokenPaste+NumTokenPaste)
248 << " token paste (##) operations performed, "
249 << NumFastTokenPaste << " on the fast path.\n";
250
251 llvm::errs() << "\nPreprocessor Memory: " << getTotalMemory() << "B total";
252
253 llvm::errs() << "\n BumpPtr: " << BP.getTotalMemory();
254 llvm::errs() << "\n Macro Expanded Tokens: "
255 << llvm::capacity_in_bytes(MacroExpandedTokens);
256 llvm::errs() << "\n Predefines Buffer: " << Predefines.capacity();
257 llvm::errs() << "\n Macros: " << llvm::capacity_in_bytes(Macros);
258 llvm::errs() << "\n #pragma push_macro Info: "
259 << llvm::capacity_in_bytes(PragmaPushMacroInfo);
260 llvm::errs() << "\n Poison Reasons: "
261 << llvm::capacity_in_bytes(PoisonReasons);
262 llvm::errs() << "\n Comment Handlers: "
263 << llvm::capacity_in_bytes(CommentHandlers) << "\n";
264 }
265
266 Preprocessor::macro_iterator
macro_begin(bool IncludeExternalMacros) const267 Preprocessor::macro_begin(bool IncludeExternalMacros) const {
268 if (IncludeExternalMacros && ExternalSource &&
269 !ReadMacrosFromExternalSource) {
270 ReadMacrosFromExternalSource = true;
271 ExternalSource->ReadDefinedMacros();
272 }
273
274 return Macros.begin();
275 }
276
getTotalMemory() const277 size_t Preprocessor::getTotalMemory() const {
278 return BP.getTotalMemory()
279 + llvm::capacity_in_bytes(MacroExpandedTokens)
280 + Predefines.capacity() /* Predefines buffer. */
281 + llvm::capacity_in_bytes(Macros)
282 + llvm::capacity_in_bytes(PragmaPushMacroInfo)
283 + llvm::capacity_in_bytes(PoisonReasons)
284 + llvm::capacity_in_bytes(CommentHandlers);
285 }
286
287 Preprocessor::macro_iterator
macro_end(bool IncludeExternalMacros) const288 Preprocessor::macro_end(bool IncludeExternalMacros) const {
289 if (IncludeExternalMacros && ExternalSource &&
290 !ReadMacrosFromExternalSource) {
291 ReadMacrosFromExternalSource = true;
292 ExternalSource->ReadDefinedMacros();
293 }
294
295 return Macros.end();
296 }
297
298 /// \brief Compares macro tokens with a specified token value sequence.
MacroDefinitionEquals(const MacroInfo * MI,ArrayRef<TokenValue> Tokens)299 static bool MacroDefinitionEquals(const MacroInfo *MI,
300 ArrayRef<TokenValue> Tokens) {
301 return Tokens.size() == MI->getNumTokens() &&
302 std::equal(Tokens.begin(), Tokens.end(), MI->tokens_begin());
303 }
304
getLastMacroWithSpelling(SourceLocation Loc,ArrayRef<TokenValue> Tokens) const305 StringRef Preprocessor::getLastMacroWithSpelling(
306 SourceLocation Loc,
307 ArrayRef<TokenValue> Tokens) const {
308 SourceLocation BestLocation;
309 StringRef BestSpelling;
310 for (Preprocessor::macro_iterator I = macro_begin(), E = macro_end();
311 I != E; ++I) {
312 if (!I->second->getMacroInfo()->isObjectLike())
313 continue;
314 const MacroDirective::DefInfo
315 Def = I->second->findDirectiveAtLoc(Loc, SourceMgr);
316 if (!Def)
317 continue;
318 if (!MacroDefinitionEquals(Def.getMacroInfo(), Tokens))
319 continue;
320 SourceLocation Location = Def.getLocation();
321 // Choose the macro defined latest.
322 if (BestLocation.isInvalid() ||
323 (Location.isValid() &&
324 SourceMgr.isBeforeInTranslationUnit(BestLocation, Location))) {
325 BestLocation = Location;
326 BestSpelling = I->first->getName();
327 }
328 }
329 return BestSpelling;
330 }
331
recomputeCurLexerKind()332 void Preprocessor::recomputeCurLexerKind() {
333 if (CurLexer)
334 CurLexerKind = CLK_Lexer;
335 else if (CurPTHLexer)
336 CurLexerKind = CLK_PTHLexer;
337 else if (CurTokenLexer)
338 CurLexerKind = CLK_TokenLexer;
339 else
340 CurLexerKind = CLK_CachingLexer;
341 }
342
SetCodeCompletionPoint(const FileEntry * File,unsigned CompleteLine,unsigned CompleteColumn)343 bool Preprocessor::SetCodeCompletionPoint(const FileEntry *File,
344 unsigned CompleteLine,
345 unsigned CompleteColumn) {
346 assert(File);
347 assert(CompleteLine && CompleteColumn && "Starts from 1:1");
348 assert(!CodeCompletionFile && "Already set");
349
350 using llvm::MemoryBuffer;
351
352 // Load the actual file's contents.
353 bool Invalid = false;
354 const MemoryBuffer *Buffer = SourceMgr.getMemoryBufferForFile(File, &Invalid);
355 if (Invalid)
356 return true;
357
358 // Find the byte position of the truncation point.
359 const char *Position = Buffer->getBufferStart();
360 for (unsigned Line = 1; Line < CompleteLine; ++Line) {
361 for (; *Position; ++Position) {
362 if (*Position != '\r' && *Position != '\n')
363 continue;
364
365 // Eat \r\n or \n\r as a single line.
366 if ((Position[1] == '\r' || Position[1] == '\n') &&
367 Position[0] != Position[1])
368 ++Position;
369 ++Position;
370 break;
371 }
372 }
373
374 Position += CompleteColumn - 1;
375
376 // Insert '\0' at the code-completion point.
377 if (Position < Buffer->getBufferEnd()) {
378 CodeCompletionFile = File;
379 CodeCompletionOffset = Position - Buffer->getBufferStart();
380
381 MemoryBuffer *NewBuffer =
382 MemoryBuffer::getNewUninitMemBuffer(Buffer->getBufferSize() + 1,
383 Buffer->getBufferIdentifier());
384 char *NewBuf = const_cast<char*>(NewBuffer->getBufferStart());
385 char *NewPos = std::copy(Buffer->getBufferStart(), Position, NewBuf);
386 *NewPos = '\0';
387 std::copy(Position, Buffer->getBufferEnd(), NewPos+1);
388 SourceMgr.overrideFileContents(File, NewBuffer);
389 }
390
391 return false;
392 }
393
CodeCompleteNaturalLanguage()394 void Preprocessor::CodeCompleteNaturalLanguage() {
395 if (CodeComplete)
396 CodeComplete->CodeCompleteNaturalLanguage();
397 setCodeCompletionReached();
398 }
399
400 /// getSpelling - This method is used to get the spelling of a token into a
401 /// SmallVector. Note that the returned StringRef may not point to the
402 /// supplied buffer if a copy can be avoided.
getSpelling(const Token & Tok,SmallVectorImpl<char> & Buffer,bool * Invalid) const403 StringRef Preprocessor::getSpelling(const Token &Tok,
404 SmallVectorImpl<char> &Buffer,
405 bool *Invalid) const {
406 // NOTE: this has to be checked *before* testing for an IdentifierInfo.
407 if (Tok.isNot(tok::raw_identifier) && !Tok.hasUCN()) {
408 // Try the fast path.
409 if (const IdentifierInfo *II = Tok.getIdentifierInfo())
410 return II->getName();
411 }
412
413 // Resize the buffer if we need to copy into it.
414 if (Tok.needsCleaning())
415 Buffer.resize(Tok.getLength());
416
417 const char *Ptr = Buffer.data();
418 unsigned Len = getSpelling(Tok, Ptr, Invalid);
419 return StringRef(Ptr, Len);
420 }
421
422 /// CreateString - Plop the specified string into a scratch buffer and return a
423 /// location for it. If specified, the source location provides a source
424 /// location for the token.
CreateString(StringRef Str,Token & Tok,SourceLocation ExpansionLocStart,SourceLocation ExpansionLocEnd)425 void Preprocessor::CreateString(StringRef Str, Token &Tok,
426 SourceLocation ExpansionLocStart,
427 SourceLocation ExpansionLocEnd) {
428 Tok.setLength(Str.size());
429
430 const char *DestPtr;
431 SourceLocation Loc = ScratchBuf->getToken(Str.data(), Str.size(), DestPtr);
432
433 if (ExpansionLocStart.isValid())
434 Loc = SourceMgr.createExpansionLoc(Loc, ExpansionLocStart,
435 ExpansionLocEnd, Str.size());
436 Tok.setLocation(Loc);
437
438 // If this is a raw identifier or a literal token, set the pointer data.
439 if (Tok.is(tok::raw_identifier))
440 Tok.setRawIdentifierData(DestPtr);
441 else if (Tok.isLiteral())
442 Tok.setLiteralData(DestPtr);
443 }
444
getCurrentModule()445 Module *Preprocessor::getCurrentModule() {
446 if (getLangOpts().CurrentModule.empty())
447 return 0;
448
449 return getHeaderSearchInfo().lookupModule(getLangOpts().CurrentModule);
450 }
451
452 //===----------------------------------------------------------------------===//
453 // Preprocessor Initialization Methods
454 //===----------------------------------------------------------------------===//
455
456
457 /// EnterMainSourceFile - Enter the specified FileID as the main source file,
458 /// which implicitly adds the builtin defines etc.
EnterMainSourceFile()459 void Preprocessor::EnterMainSourceFile() {
460 // We do not allow the preprocessor to reenter the main file. Doing so will
461 // cause FileID's to accumulate information from both runs (e.g. #line
462 // information) and predefined macros aren't guaranteed to be set properly.
463 assert(NumEnteredSourceFiles == 0 && "Cannot reenter the main file!");
464 FileID MainFileID = SourceMgr.getMainFileID();
465
466 // If MainFileID is loaded it means we loaded an AST file, no need to enter
467 // a main file.
468 if (!SourceMgr.isLoadedFileID(MainFileID)) {
469 // Enter the main file source buffer.
470 EnterSourceFile(MainFileID, 0, SourceLocation());
471
472 // If we've been asked to skip bytes in the main file (e.g., as part of a
473 // precompiled preamble), do so now.
474 if (SkipMainFilePreamble.first > 0)
475 CurLexer->SkipBytes(SkipMainFilePreamble.first,
476 SkipMainFilePreamble.second);
477
478 // Tell the header info that the main file was entered. If the file is later
479 // #imported, it won't be re-entered.
480 if (const FileEntry *FE = SourceMgr.getFileEntryForID(MainFileID))
481 HeaderInfo.IncrementIncludeCount(FE);
482 }
483
484 // Preprocess Predefines to populate the initial preprocessor state.
485 llvm::MemoryBuffer *SB =
486 llvm::MemoryBuffer::getMemBufferCopy(Predefines, "<built-in>");
487 assert(SB && "Cannot create predefined source buffer");
488 FileID FID = SourceMgr.createFileIDForMemBuffer(SB);
489 assert(!FID.isInvalid() && "Could not create FileID for predefines?");
490 setPredefinesFileID(FID);
491
492 // Start parsing the predefines.
493 EnterSourceFile(FID, 0, SourceLocation());
494 }
495
EndSourceFile()496 void Preprocessor::EndSourceFile() {
497 // Notify the client that we reached the end of the source file.
498 if (Callbacks)
499 Callbacks->EndOfMainFile();
500 }
501
502 //===----------------------------------------------------------------------===//
503 // Lexer Event Handling.
504 //===----------------------------------------------------------------------===//
505
appendCodePoint(unsigned Codepoint,llvm::SmallVectorImpl<char> & Str)506 static void appendCodePoint(unsigned Codepoint,
507 llvm::SmallVectorImpl<char> &Str) {
508 char ResultBuf[4];
509 char *ResultPtr = ResultBuf;
510 bool Res = llvm::ConvertCodePointToUTF8(Codepoint, ResultPtr);
511 (void)Res;
512 assert(Res && "Unexpected conversion failure");
513 Str.append(ResultBuf, ResultPtr);
514 }
515
expandUCNs(SmallVectorImpl<char> & Buf,StringRef Input)516 static void expandUCNs(SmallVectorImpl<char> &Buf, StringRef Input) {
517 for (StringRef::iterator I = Input.begin(), E = Input.end(); I != E; ++I) {
518 if (*I != '\\') {
519 Buf.push_back(*I);
520 continue;
521 }
522
523 ++I;
524 assert(*I == 'u' || *I == 'U');
525
526 unsigned NumHexDigits;
527 if (*I == 'u')
528 NumHexDigits = 4;
529 else
530 NumHexDigits = 8;
531
532 assert(I + NumHexDigits <= E);
533
534 uint32_t CodePoint = 0;
535 for (++I; NumHexDigits != 0; ++I, --NumHexDigits) {
536 unsigned Value = llvm::hexDigitValue(*I);
537 assert(Value != -1U);
538
539 CodePoint <<= 4;
540 CodePoint += Value;
541 }
542
543 appendCodePoint(CodePoint, Buf);
544 --I;
545 }
546 }
547
548 /// LookUpIdentifierInfo - Given a tok::raw_identifier token, look up the
549 /// identifier information for the token and install it into the token,
550 /// updating the token kind accordingly.
LookUpIdentifierInfo(Token & Identifier) const551 IdentifierInfo *Preprocessor::LookUpIdentifierInfo(Token &Identifier) const {
552 assert(Identifier.getRawIdentifierData() != 0 && "No raw identifier data!");
553
554 // Look up this token, see if it is a macro, or if it is a language keyword.
555 IdentifierInfo *II;
556 if (!Identifier.needsCleaning() && !Identifier.hasUCN()) {
557 // No cleaning needed, just use the characters from the lexed buffer.
558 II = getIdentifierInfo(StringRef(Identifier.getRawIdentifierData(),
559 Identifier.getLength()));
560 } else {
561 // Cleaning needed, alloca a buffer, clean into it, then use the buffer.
562 SmallString<64> IdentifierBuffer;
563 StringRef CleanedStr = getSpelling(Identifier, IdentifierBuffer);
564
565 if (Identifier.hasUCN()) {
566 SmallString<64> UCNIdentifierBuffer;
567 expandUCNs(UCNIdentifierBuffer, CleanedStr);
568 II = getIdentifierInfo(UCNIdentifierBuffer);
569 } else {
570 II = getIdentifierInfo(CleanedStr);
571 }
572 }
573
574 // Update the token info (identifier info and appropriate token kind).
575 Identifier.setIdentifierInfo(II);
576 Identifier.setKind(II->getTokenID());
577
578 return II;
579 }
580
SetPoisonReason(IdentifierInfo * II,unsigned DiagID)581 void Preprocessor::SetPoisonReason(IdentifierInfo *II, unsigned DiagID) {
582 PoisonReasons[II] = DiagID;
583 }
584
PoisonSEHIdentifiers(bool Poison)585 void Preprocessor::PoisonSEHIdentifiers(bool Poison) {
586 assert(Ident__exception_code && Ident__exception_info);
587 assert(Ident___exception_code && Ident___exception_info);
588 Ident__exception_code->setIsPoisoned(Poison);
589 Ident___exception_code->setIsPoisoned(Poison);
590 Ident_GetExceptionCode->setIsPoisoned(Poison);
591 Ident__exception_info->setIsPoisoned(Poison);
592 Ident___exception_info->setIsPoisoned(Poison);
593 Ident_GetExceptionInfo->setIsPoisoned(Poison);
594 Ident__abnormal_termination->setIsPoisoned(Poison);
595 Ident___abnormal_termination->setIsPoisoned(Poison);
596 Ident_AbnormalTermination->setIsPoisoned(Poison);
597 }
598
HandlePoisonedIdentifier(Token & Identifier)599 void Preprocessor::HandlePoisonedIdentifier(Token & Identifier) {
600 assert(Identifier.getIdentifierInfo() &&
601 "Can't handle identifiers without identifier info!");
602 llvm::DenseMap<IdentifierInfo*,unsigned>::const_iterator it =
603 PoisonReasons.find(Identifier.getIdentifierInfo());
604 if(it == PoisonReasons.end())
605 Diag(Identifier, diag::err_pp_used_poisoned_id);
606 else
607 Diag(Identifier,it->second) << Identifier.getIdentifierInfo();
608 }
609
610 /// HandleIdentifier - This callback is invoked when the lexer reads an
611 /// identifier. This callback looks up the identifier in the map and/or
612 /// potentially macro expands it or turns it into a named token (like 'for').
613 ///
614 /// Note that callers of this method are guarded by checking the
615 /// IdentifierInfo's 'isHandleIdentifierCase' bit. If this method changes, the
616 /// IdentifierInfo methods that compute these properties will need to change to
617 /// match.
HandleIdentifier(Token & Identifier)618 bool Preprocessor::HandleIdentifier(Token &Identifier) {
619 assert(Identifier.getIdentifierInfo() &&
620 "Can't handle identifiers without identifier info!");
621
622 IdentifierInfo &II = *Identifier.getIdentifierInfo();
623
624 // If the information about this identifier is out of date, update it from
625 // the external source.
626 // We have to treat __VA_ARGS__ in a special way, since it gets
627 // serialized with isPoisoned = true, but our preprocessor may have
628 // unpoisoned it if we're defining a C99 macro.
629 if (II.isOutOfDate()) {
630 bool CurrentIsPoisoned = false;
631 if (&II == Ident__VA_ARGS__)
632 CurrentIsPoisoned = Ident__VA_ARGS__->isPoisoned();
633
634 ExternalSource->updateOutOfDateIdentifier(II);
635 Identifier.setKind(II.getTokenID());
636
637 if (&II == Ident__VA_ARGS__)
638 II.setIsPoisoned(CurrentIsPoisoned);
639 }
640
641 // If this identifier was poisoned, and if it was not produced from a macro
642 // expansion, emit an error.
643 if (II.isPoisoned() && CurPPLexer) {
644 HandlePoisonedIdentifier(Identifier);
645 }
646
647 // If this is a macro to be expanded, do it.
648 if (MacroDirective *MD = getMacroDirective(&II)) {
649 MacroInfo *MI = MD->getMacroInfo();
650 if (!DisableMacroExpansion) {
651 if (!Identifier.isExpandDisabled() && MI->isEnabled()) {
652 // C99 6.10.3p10: If the preprocessing token immediately after the
653 // macro name isn't a '(', this macro should not be expanded.
654 if (!MI->isFunctionLike() || isNextPPTokenLParen())
655 return HandleMacroExpandedIdentifier(Identifier, MD);
656 } else {
657 // C99 6.10.3.4p2 says that a disabled macro may never again be
658 // expanded, even if it's in a context where it could be expanded in the
659 // future.
660 Identifier.setFlag(Token::DisableExpand);
661 if (MI->isObjectLike() || isNextPPTokenLParen())
662 Diag(Identifier, diag::pp_disabled_macro_expansion);
663 }
664 }
665 }
666
667 // If this identifier is a keyword in C++11, produce a warning. Don't warn if
668 // we're not considering macro expansion, since this identifier might be the
669 // name of a macro.
670 // FIXME: This warning is disabled in cases where it shouldn't be, like
671 // "#define constexpr constexpr", "int constexpr;"
672 if (II.isCXX11CompatKeyword() & !DisableMacroExpansion) {
673 Diag(Identifier, diag::warn_cxx11_keyword) << II.getName();
674 // Don't diagnose this keyword again in this translation unit.
675 II.setIsCXX11CompatKeyword(false);
676 }
677
678 // C++ 2.11p2: If this is an alternative representation of a C++ operator,
679 // then we act as if it is the actual operator and not the textual
680 // representation of it.
681 if (II.isCPlusPlusOperatorKeyword())
682 Identifier.setIdentifierInfo(0);
683
684 // If this is an extension token, diagnose its use.
685 // We avoid diagnosing tokens that originate from macro definitions.
686 // FIXME: This warning is disabled in cases where it shouldn't be,
687 // like "#define TY typeof", "TY(1) x".
688 if (II.isExtensionToken() && !DisableMacroExpansion)
689 Diag(Identifier, diag::ext_token_used);
690
691 // If this is the 'import' contextual keyword following an '@', note
692 // that the next token indicates a module name.
693 //
694 // Note that we do not treat 'import' as a contextual
695 // keyword when we're in a caching lexer, because caching lexers only get
696 // used in contexts where import declarations are disallowed.
697 if (LastTokenWasAt && II.isModulesImport() && !InMacroArgs &&
698 !DisableMacroExpansion && getLangOpts().Modules &&
699 CurLexerKind != CLK_CachingLexer) {
700 ModuleImportLoc = Identifier.getLocation();
701 ModuleImportPath.clear();
702 ModuleImportExpectsIdentifier = true;
703 CurLexerKind = CLK_LexAfterModuleImport;
704 }
705 return true;
706 }
707
Lex(Token & Result)708 void Preprocessor::Lex(Token &Result) {
709 // We loop here until a lex function retuns a token; this avoids recursion.
710 bool ReturnedToken;
711 do {
712 switch (CurLexerKind) {
713 case CLK_Lexer:
714 ReturnedToken = CurLexer->Lex(Result);
715 break;
716 case CLK_PTHLexer:
717 ReturnedToken = CurPTHLexer->Lex(Result);
718 break;
719 case CLK_TokenLexer:
720 ReturnedToken = CurTokenLexer->Lex(Result);
721 break;
722 case CLK_CachingLexer:
723 CachingLex(Result);
724 ReturnedToken = true;
725 break;
726 case CLK_LexAfterModuleImport:
727 LexAfterModuleImport(Result);
728 ReturnedToken = true;
729 break;
730 }
731 } while (!ReturnedToken);
732
733 LastTokenWasAt = Result.is(tok::at);
734 }
735
736
737 /// \brief Lex a token following the 'import' contextual keyword.
738 ///
LexAfterModuleImport(Token & Result)739 void Preprocessor::LexAfterModuleImport(Token &Result) {
740 // Figure out what kind of lexer we actually have.
741 recomputeCurLexerKind();
742
743 // Lex the next token.
744 Lex(Result);
745
746 // The token sequence
747 //
748 // import identifier (. identifier)*
749 //
750 // indicates a module import directive. We already saw the 'import'
751 // contextual keyword, so now we're looking for the identifiers.
752 if (ModuleImportExpectsIdentifier && Result.getKind() == tok::identifier) {
753 // We expected to see an identifier here, and we did; continue handling
754 // identifiers.
755 ModuleImportPath.push_back(std::make_pair(Result.getIdentifierInfo(),
756 Result.getLocation()));
757 ModuleImportExpectsIdentifier = false;
758 CurLexerKind = CLK_LexAfterModuleImport;
759 return;
760 }
761
762 // If we're expecting a '.' or a ';', and we got a '.', then wait until we
763 // see the next identifier.
764 if (!ModuleImportExpectsIdentifier && Result.getKind() == tok::period) {
765 ModuleImportExpectsIdentifier = true;
766 CurLexerKind = CLK_LexAfterModuleImport;
767 return;
768 }
769
770 // If we have a non-empty module path, load the named module.
771 if (!ModuleImportPath.empty() && getLangOpts().Modules) {
772 Module *Imported = TheModuleLoader.loadModule(ModuleImportLoc,
773 ModuleImportPath,
774 Module::MacrosVisible,
775 /*IsIncludeDirective=*/false);
776 if (Callbacks)
777 Callbacks->moduleImport(ModuleImportLoc, ModuleImportPath, Imported);
778 }
779 }
780
FinishLexStringLiteral(Token & Result,std::string & String,const char * DiagnosticTag,bool AllowMacroExpansion)781 bool Preprocessor::FinishLexStringLiteral(Token &Result, std::string &String,
782 const char *DiagnosticTag,
783 bool AllowMacroExpansion) {
784 // We need at least one string literal.
785 if (Result.isNot(tok::string_literal)) {
786 Diag(Result, diag::err_expected_string_literal)
787 << /*Source='in...'*/0 << DiagnosticTag;
788 return false;
789 }
790
791 // Lex string literal tokens, optionally with macro expansion.
792 SmallVector<Token, 4> StrToks;
793 do {
794 StrToks.push_back(Result);
795
796 if (Result.hasUDSuffix())
797 Diag(Result, diag::err_invalid_string_udl);
798
799 if (AllowMacroExpansion)
800 Lex(Result);
801 else
802 LexUnexpandedToken(Result);
803 } while (Result.is(tok::string_literal));
804
805 // Concatenate and parse the strings.
806 StringLiteralParser Literal(&StrToks[0], StrToks.size(), *this);
807 assert(Literal.isAscii() && "Didn't allow wide strings in");
808
809 if (Literal.hadError)
810 return false;
811
812 if (Literal.Pascal) {
813 Diag(StrToks[0].getLocation(), diag::err_expected_string_literal)
814 << /*Source='in...'*/0 << DiagnosticTag;
815 return false;
816 }
817
818 String = Literal.GetString();
819 return true;
820 }
821
addCommentHandler(CommentHandler * Handler)822 void Preprocessor::addCommentHandler(CommentHandler *Handler) {
823 assert(Handler && "NULL comment handler");
824 assert(std::find(CommentHandlers.begin(), CommentHandlers.end(), Handler) ==
825 CommentHandlers.end() && "Comment handler already registered");
826 CommentHandlers.push_back(Handler);
827 }
828
removeCommentHandler(CommentHandler * Handler)829 void Preprocessor::removeCommentHandler(CommentHandler *Handler) {
830 std::vector<CommentHandler *>::iterator Pos
831 = std::find(CommentHandlers.begin(), CommentHandlers.end(), Handler);
832 assert(Pos != CommentHandlers.end() && "Comment handler not registered");
833 CommentHandlers.erase(Pos);
834 }
835
HandleComment(Token & result,SourceRange Comment)836 bool Preprocessor::HandleComment(Token &result, SourceRange Comment) {
837 bool AnyPendingTokens = false;
838 for (std::vector<CommentHandler *>::iterator H = CommentHandlers.begin(),
839 HEnd = CommentHandlers.end();
840 H != HEnd; ++H) {
841 if ((*H)->HandleComment(*this, Comment))
842 AnyPendingTokens = true;
843 }
844 if (!AnyPendingTokens || getCommentRetentionState())
845 return false;
846 Lex(result);
847 return true;
848 }
849
~ModuleLoader()850 ModuleLoader::~ModuleLoader() { }
851
~CommentHandler()852 CommentHandler::~CommentHandler() { }
853
~CodeCompletionHandler()854 CodeCompletionHandler::~CodeCompletionHandler() { }
855
createPreprocessingRecord()856 void Preprocessor::createPreprocessingRecord() {
857 if (Record)
858 return;
859
860 Record = new PreprocessingRecord(getSourceManager());
861 addPPCallbacks(Record);
862 }
863