1 //===- Preprocessor.cpp - C Language Family Preprocessor Implementation ---===//
2 //
3 // Part of the LLVM Project, under the Apache License v2.0 with LLVM Exceptions.
4 // See https://llvm.org/LICENSE.txt for license information.
5 // SPDX-License-Identifier: Apache-2.0 WITH LLVM-exception
6 //
7 //===----------------------------------------------------------------------===//
8 //
9 // This file implements the Preprocessor interface.
10 //
11 //===----------------------------------------------------------------------===//
12 //
13 // Options to support:
14 // -H - Print the name of each header file used.
15 // -d[DNI] - Dump various things.
16 // -fworking-directory - #line's with preprocessor's working dir.
17 // -fpreprocessed
18 // -dependency-file,-M,-MM,-MF,-MG,-MP,-MT,-MQ,-MD,-MMD
19 // -W*
20 // -w
21 //
22 // Messages to emit:
23 // "Multiple include guards may be useful for:\n"
24 //
25 //===----------------------------------------------------------------------===//
26
27 #include "clang/Lex/Preprocessor.h"
28 #include "clang/Basic/Builtins.h"
29 #include "clang/Basic/FileManager.h"
30 #include "clang/Basic/FileSystemStatCache.h"
31 #include "clang/Basic/IdentifierTable.h"
32 #include "clang/Basic/LLVM.h"
33 #include "clang/Basic/LangOptions.h"
34 #include "clang/Basic/Module.h"
35 #include "clang/Basic/SourceLocation.h"
36 #include "clang/Basic/SourceManager.h"
37 #include "clang/Basic/TargetInfo.h"
38 #include "clang/Lex/CodeCompletionHandler.h"
39 #include "clang/Lex/ExternalPreprocessorSource.h"
40 #include "clang/Lex/HeaderSearch.h"
41 #include "clang/Lex/LexDiagnostic.h"
42 #include "clang/Lex/Lexer.h"
43 #include "clang/Lex/LiteralSupport.h"
44 #include "clang/Lex/MacroArgs.h"
45 #include "clang/Lex/MacroInfo.h"
46 #include "clang/Lex/ModuleLoader.h"
47 #include "clang/Lex/Pragma.h"
48 #include "clang/Lex/PreprocessingRecord.h"
49 #include "clang/Lex/PreprocessorLexer.h"
50 #include "clang/Lex/PreprocessorOptions.h"
51 #include "clang/Lex/ScratchBuffer.h"
52 #include "clang/Lex/Token.h"
53 #include "clang/Lex/TokenLexer.h"
54 #include "llvm/ADT/APInt.h"
55 #include "llvm/ADT/ArrayRef.h"
56 #include "llvm/ADT/DenseMap.h"
57 #include "llvm/ADT/STLExtras.h"
58 #include "llvm/ADT/SmallString.h"
59 #include "llvm/ADT/SmallVector.h"
60 #include "llvm/ADT/StringRef.h"
61 #include "llvm/ADT/StringSwitch.h"
62 #include "llvm/Support/Capacity.h"
63 #include "llvm/Support/ErrorHandling.h"
64 #include "llvm/Support/MemoryBuffer.h"
65 #include "llvm/Support/raw_ostream.h"
66 #include <algorithm>
67 #include <cassert>
68 #include <memory>
69 #include <string>
70 #include <utility>
71 #include <vector>
72
73 using namespace clang;
74
75 LLVM_INSTANTIATE_REGISTRY(PragmaHandlerRegistry)
76
77 ExternalPreprocessorSource::~ExternalPreprocessorSource() = default;
78
Preprocessor(std::shared_ptr<PreprocessorOptions> PPOpts,DiagnosticsEngine & diags,LangOptions & opts,SourceManager & SM,HeaderSearch & Headers,ModuleLoader & TheModuleLoader,IdentifierInfoLookup * IILookup,bool OwnsHeaders,TranslationUnitKind TUKind)79 Preprocessor::Preprocessor(std::shared_ptr<PreprocessorOptions> PPOpts,
80 DiagnosticsEngine &diags, LangOptions &opts,
81 SourceManager &SM, HeaderSearch &Headers,
82 ModuleLoader &TheModuleLoader,
83 IdentifierInfoLookup *IILookup, bool OwnsHeaders,
84 TranslationUnitKind TUKind)
85 : PPOpts(std::move(PPOpts)), Diags(&diags), LangOpts(opts),
86 FileMgr(Headers.getFileMgr()), SourceMgr(SM),
87 ScratchBuf(new ScratchBuffer(SourceMgr)), HeaderInfo(Headers),
88 TheModuleLoader(TheModuleLoader), ExternalSource(nullptr),
89 // As the language options may have not been loaded yet (when
90 // deserializing an ASTUnit), adding keywords to the identifier table is
91 // deferred to Preprocessor::Initialize().
92 Identifiers(IILookup), PragmaHandlers(new PragmaNamespace(StringRef())),
93 TUKind(TUKind), SkipMainFilePreamble(0, true),
94 CurSubmoduleState(&NullSubmoduleState) {
95 OwnsHeaderSearch = OwnsHeaders;
96
97 // Default to discarding comments.
98 KeepComments = false;
99 KeepMacroComments = false;
100 SuppressIncludeNotFoundError = false;
101
102 // Macro expansion is enabled.
103 DisableMacroExpansion = false;
104 MacroExpansionInDirectivesOverride = false;
105 InMacroArgs = false;
106 ArgMacro = nullptr;
107 InMacroArgPreExpansion = false;
108 NumCachedTokenLexers = 0;
109 PragmasEnabled = true;
110 ParsingIfOrElifDirective = false;
111 PreprocessedOutput = false;
112
113 // We haven't read anything from the external source.
114 ReadMacrosFromExternalSource = false;
115
116 BuiltinInfo = std::make_unique<Builtin::Context>();
117
118 // "Poison" __VA_ARGS__, __VA_OPT__ which can only appear in the expansion of
119 // a macro. They get unpoisoned where it is allowed.
120 (Ident__VA_ARGS__ = getIdentifierInfo("__VA_ARGS__"))->setIsPoisoned();
121 SetPoisonReason(Ident__VA_ARGS__,diag::ext_pp_bad_vaargs_use);
122 (Ident__VA_OPT__ = getIdentifierInfo("__VA_OPT__"))->setIsPoisoned();
123 SetPoisonReason(Ident__VA_OPT__,diag::ext_pp_bad_vaopt_use);
124
125 // Initialize the pragma handlers.
126 RegisterBuiltinPragmas();
127
128 // Initialize builtin macros like __LINE__ and friends.
129 RegisterBuiltinMacros();
130
131 if(LangOpts.Borland) {
132 Ident__exception_info = getIdentifierInfo("_exception_info");
133 Ident___exception_info = getIdentifierInfo("__exception_info");
134 Ident_GetExceptionInfo = getIdentifierInfo("GetExceptionInformation");
135 Ident__exception_code = getIdentifierInfo("_exception_code");
136 Ident___exception_code = getIdentifierInfo("__exception_code");
137 Ident_GetExceptionCode = getIdentifierInfo("GetExceptionCode");
138 Ident__abnormal_termination = getIdentifierInfo("_abnormal_termination");
139 Ident___abnormal_termination = getIdentifierInfo("__abnormal_termination");
140 Ident_AbnormalTermination = getIdentifierInfo("AbnormalTermination");
141 } else {
142 Ident__exception_info = Ident__exception_code = nullptr;
143 Ident__abnormal_termination = Ident___exception_info = nullptr;
144 Ident___exception_code = Ident___abnormal_termination = nullptr;
145 Ident_GetExceptionInfo = Ident_GetExceptionCode = nullptr;
146 Ident_AbnormalTermination = nullptr;
147 }
148
149 // If using a PCH where a #pragma hdrstop is expected, start skipping tokens.
150 if (usingPCHWithPragmaHdrStop())
151 SkippingUntilPragmaHdrStop = true;
152
153 // If using a PCH with a through header, start skipping tokens.
154 if (!this->PPOpts->PCHThroughHeader.empty() &&
155 !this->PPOpts->ImplicitPCHInclude.empty())
156 SkippingUntilPCHThroughHeader = true;
157
158 if (this->PPOpts->GeneratePreamble)
159 PreambleConditionalStack.startRecording();
160
161 ExcludedConditionalDirectiveSkipMappings =
162 this->PPOpts->ExcludedConditionalDirectiveSkipMappings;
163 if (ExcludedConditionalDirectiveSkipMappings)
164 ExcludedConditionalDirectiveSkipMappings->clear();
165
166 MaxTokens = LangOpts.MaxTokens;
167 }
168
~Preprocessor()169 Preprocessor::~Preprocessor() {
170 assert(BacktrackPositions.empty() && "EnableBacktrack/Backtrack imbalance!");
171
172 IncludeMacroStack.clear();
173
174 // Destroy any macro definitions.
175 while (MacroInfoChain *I = MIChainHead) {
176 MIChainHead = I->Next;
177 I->~MacroInfoChain();
178 }
179
180 // Free any cached macro expanders.
181 // This populates MacroArgCache, so all TokenLexers need to be destroyed
182 // before the code below that frees up the MacroArgCache list.
183 std::fill(TokenLexerCache, TokenLexerCache + NumCachedTokenLexers, nullptr);
184 CurTokenLexer.reset();
185
186 // Free any cached MacroArgs.
187 for (MacroArgs *ArgList = MacroArgCache; ArgList;)
188 ArgList = ArgList->deallocate();
189
190 // Delete the header search info, if we own it.
191 if (OwnsHeaderSearch)
192 delete &HeaderInfo;
193 }
194
Initialize(const TargetInfo & Target,const TargetInfo * AuxTarget)195 void Preprocessor::Initialize(const TargetInfo &Target,
196 const TargetInfo *AuxTarget) {
197 assert((!this->Target || this->Target == &Target) &&
198 "Invalid override of target information");
199 this->Target = &Target;
200
201 assert((!this->AuxTarget || this->AuxTarget == AuxTarget) &&
202 "Invalid override of aux target information.");
203 this->AuxTarget = AuxTarget;
204
205 // Initialize information about built-ins.
206 BuiltinInfo->InitializeTarget(Target, AuxTarget);
207 HeaderInfo.setTarget(Target);
208
209 // Populate the identifier table with info about keywords for the current language.
210 Identifiers.AddKeywords(LangOpts);
211 }
212
InitializeForModelFile()213 void Preprocessor::InitializeForModelFile() {
214 NumEnteredSourceFiles = 0;
215
216 // Reset pragmas
217 PragmaHandlersBackup = std::move(PragmaHandlers);
218 PragmaHandlers = std::make_unique<PragmaNamespace>(StringRef());
219 RegisterBuiltinPragmas();
220
221 // Reset PredefinesFileID
222 PredefinesFileID = FileID();
223 }
224
FinalizeForModelFile()225 void Preprocessor::FinalizeForModelFile() {
226 NumEnteredSourceFiles = 1;
227
228 PragmaHandlers = std::move(PragmaHandlersBackup);
229 }
230
DumpToken(const Token & Tok,bool DumpFlags) const231 void Preprocessor::DumpToken(const Token &Tok, bool DumpFlags) const {
232 llvm::errs() << tok::getTokenName(Tok.getKind()) << " '"
233 << getSpelling(Tok) << "'";
234
235 if (!DumpFlags) return;
236
237 llvm::errs() << "\t";
238 if (Tok.isAtStartOfLine())
239 llvm::errs() << " [StartOfLine]";
240 if (Tok.hasLeadingSpace())
241 llvm::errs() << " [LeadingSpace]";
242 if (Tok.isExpandDisabled())
243 llvm::errs() << " [ExpandDisabled]";
244 if (Tok.needsCleaning()) {
245 const char *Start = SourceMgr.getCharacterData(Tok.getLocation());
246 llvm::errs() << " [UnClean='" << StringRef(Start, Tok.getLength())
247 << "']";
248 }
249
250 llvm::errs() << "\tLoc=<";
251 DumpLocation(Tok.getLocation());
252 llvm::errs() << ">";
253 }
254
DumpLocation(SourceLocation Loc) const255 void Preprocessor::DumpLocation(SourceLocation Loc) const {
256 Loc.print(llvm::errs(), SourceMgr);
257 }
258
DumpMacro(const MacroInfo & MI) const259 void Preprocessor::DumpMacro(const MacroInfo &MI) const {
260 llvm::errs() << "MACRO: ";
261 for (unsigned i = 0, e = MI.getNumTokens(); i != e; ++i) {
262 DumpToken(MI.getReplacementToken(i));
263 llvm::errs() << " ";
264 }
265 llvm::errs() << "\n";
266 }
267
PrintStats()268 void Preprocessor::PrintStats() {
269 llvm::errs() << "\n*** Preprocessor Stats:\n";
270 llvm::errs() << NumDirectives << " directives found:\n";
271 llvm::errs() << " " << NumDefined << " #define.\n";
272 llvm::errs() << " " << NumUndefined << " #undef.\n";
273 llvm::errs() << " #include/#include_next/#import:\n";
274 llvm::errs() << " " << NumEnteredSourceFiles << " source files entered.\n";
275 llvm::errs() << " " << MaxIncludeStackDepth << " max include stack depth\n";
276 llvm::errs() << " " << NumIf << " #if/#ifndef/#ifdef.\n";
277 llvm::errs() << " " << NumElse << " #else/#elif/#elifdef/#elifndef.\n";
278 llvm::errs() << " " << NumEndif << " #endif.\n";
279 llvm::errs() << " " << NumPragma << " #pragma.\n";
280 llvm::errs() << NumSkipped << " #if/#ifndef#ifdef regions skipped\n";
281
282 llvm::errs() << NumMacroExpanded << "/" << NumFnMacroExpanded << "/"
283 << NumBuiltinMacroExpanded << " obj/fn/builtin macros expanded, "
284 << NumFastMacroExpanded << " on the fast path.\n";
285 llvm::errs() << (NumFastTokenPaste+NumTokenPaste)
286 << " token paste (##) operations performed, "
287 << NumFastTokenPaste << " on the fast path.\n";
288
289 llvm::errs() << "\nPreprocessor Memory: " << getTotalMemory() << "B total";
290
291 llvm::errs() << "\n BumpPtr: " << BP.getTotalMemory();
292 llvm::errs() << "\n Macro Expanded Tokens: "
293 << llvm::capacity_in_bytes(MacroExpandedTokens);
294 llvm::errs() << "\n Predefines Buffer: " << Predefines.capacity();
295 // FIXME: List information for all submodules.
296 llvm::errs() << "\n Macros: "
297 << llvm::capacity_in_bytes(CurSubmoduleState->Macros);
298 llvm::errs() << "\n #pragma push_macro Info: "
299 << llvm::capacity_in_bytes(PragmaPushMacroInfo);
300 llvm::errs() << "\n Poison Reasons: "
301 << llvm::capacity_in_bytes(PoisonReasons);
302 llvm::errs() << "\n Comment Handlers: "
303 << llvm::capacity_in_bytes(CommentHandlers) << "\n";
304 }
305
306 Preprocessor::macro_iterator
macro_begin(bool IncludeExternalMacros) const307 Preprocessor::macro_begin(bool IncludeExternalMacros) const {
308 if (IncludeExternalMacros && ExternalSource &&
309 !ReadMacrosFromExternalSource) {
310 ReadMacrosFromExternalSource = true;
311 ExternalSource->ReadDefinedMacros();
312 }
313
314 // Make sure we cover all macros in visible modules.
315 for (const ModuleMacro &Macro : ModuleMacros)
316 CurSubmoduleState->Macros.insert(std::make_pair(Macro.II, MacroState()));
317
318 return CurSubmoduleState->Macros.begin();
319 }
320
getTotalMemory() const321 size_t Preprocessor::getTotalMemory() const {
322 return BP.getTotalMemory()
323 + llvm::capacity_in_bytes(MacroExpandedTokens)
324 + Predefines.capacity() /* Predefines buffer. */
325 // FIXME: Include sizes from all submodules, and include MacroInfo sizes,
326 // and ModuleMacros.
327 + llvm::capacity_in_bytes(CurSubmoduleState->Macros)
328 + llvm::capacity_in_bytes(PragmaPushMacroInfo)
329 + llvm::capacity_in_bytes(PoisonReasons)
330 + llvm::capacity_in_bytes(CommentHandlers);
331 }
332
333 Preprocessor::macro_iterator
macro_end(bool IncludeExternalMacros) const334 Preprocessor::macro_end(bool IncludeExternalMacros) const {
335 if (IncludeExternalMacros && ExternalSource &&
336 !ReadMacrosFromExternalSource) {
337 ReadMacrosFromExternalSource = true;
338 ExternalSource->ReadDefinedMacros();
339 }
340
341 return CurSubmoduleState->Macros.end();
342 }
343
344 /// Compares macro tokens with a specified token value sequence.
MacroDefinitionEquals(const MacroInfo * MI,ArrayRef<TokenValue> Tokens)345 static bool MacroDefinitionEquals(const MacroInfo *MI,
346 ArrayRef<TokenValue> Tokens) {
347 return Tokens.size() == MI->getNumTokens() &&
348 std::equal(Tokens.begin(), Tokens.end(), MI->tokens_begin());
349 }
350
getLastMacroWithSpelling(SourceLocation Loc,ArrayRef<TokenValue> Tokens) const351 StringRef Preprocessor::getLastMacroWithSpelling(
352 SourceLocation Loc,
353 ArrayRef<TokenValue> Tokens) const {
354 SourceLocation BestLocation;
355 StringRef BestSpelling;
356 for (Preprocessor::macro_iterator I = macro_begin(), E = macro_end();
357 I != E; ++I) {
358 const MacroDirective::DefInfo
359 Def = I->second.findDirectiveAtLoc(Loc, SourceMgr);
360 if (!Def || !Def.getMacroInfo())
361 continue;
362 if (!Def.getMacroInfo()->isObjectLike())
363 continue;
364 if (!MacroDefinitionEquals(Def.getMacroInfo(), Tokens))
365 continue;
366 SourceLocation Location = Def.getLocation();
367 // Choose the macro defined latest.
368 if (BestLocation.isInvalid() ||
369 (Location.isValid() &&
370 SourceMgr.isBeforeInTranslationUnit(BestLocation, Location))) {
371 BestLocation = Location;
372 BestSpelling = I->first->getName();
373 }
374 }
375 return BestSpelling;
376 }
377
recomputeCurLexerKind()378 void Preprocessor::recomputeCurLexerKind() {
379 if (CurLexer)
380 CurLexerKind = CLK_Lexer;
381 else if (CurTokenLexer)
382 CurLexerKind = CLK_TokenLexer;
383 else
384 CurLexerKind = CLK_CachingLexer;
385 }
386
SetCodeCompletionPoint(const FileEntry * File,unsigned CompleteLine,unsigned CompleteColumn)387 bool Preprocessor::SetCodeCompletionPoint(const FileEntry *File,
388 unsigned CompleteLine,
389 unsigned CompleteColumn) {
390 assert(File);
391 assert(CompleteLine && CompleteColumn && "Starts from 1:1");
392 assert(!CodeCompletionFile && "Already set");
393
394 // Load the actual file's contents.
395 Optional<llvm::MemoryBufferRef> Buffer =
396 SourceMgr.getMemoryBufferForFileOrNone(File);
397 if (!Buffer)
398 return true;
399
400 // Find the byte position of the truncation point.
401 const char *Position = Buffer->getBufferStart();
402 for (unsigned Line = 1; Line < CompleteLine; ++Line) {
403 for (; *Position; ++Position) {
404 if (*Position != '\r' && *Position != '\n')
405 continue;
406
407 // Eat \r\n or \n\r as a single line.
408 if ((Position[1] == '\r' || Position[1] == '\n') &&
409 Position[0] != Position[1])
410 ++Position;
411 ++Position;
412 break;
413 }
414 }
415
416 Position += CompleteColumn - 1;
417
418 // If pointing inside the preamble, adjust the position at the beginning of
419 // the file after the preamble.
420 if (SkipMainFilePreamble.first &&
421 SourceMgr.getFileEntryForID(SourceMgr.getMainFileID()) == File) {
422 if (Position - Buffer->getBufferStart() < SkipMainFilePreamble.first)
423 Position = Buffer->getBufferStart() + SkipMainFilePreamble.first;
424 }
425
426 if (Position > Buffer->getBufferEnd())
427 Position = Buffer->getBufferEnd();
428
429 CodeCompletionFile = File;
430 CodeCompletionOffset = Position - Buffer->getBufferStart();
431
432 auto NewBuffer = llvm::WritableMemoryBuffer::getNewUninitMemBuffer(
433 Buffer->getBufferSize() + 1, Buffer->getBufferIdentifier());
434 char *NewBuf = NewBuffer->getBufferStart();
435 char *NewPos = std::copy(Buffer->getBufferStart(), Position, NewBuf);
436 *NewPos = '\0';
437 std::copy(Position, Buffer->getBufferEnd(), NewPos+1);
438 SourceMgr.overrideFileContents(File, std::move(NewBuffer));
439
440 return false;
441 }
442
CodeCompleteIncludedFile(llvm::StringRef Dir,bool IsAngled)443 void Preprocessor::CodeCompleteIncludedFile(llvm::StringRef Dir,
444 bool IsAngled) {
445 setCodeCompletionReached();
446 if (CodeComplete)
447 CodeComplete->CodeCompleteIncludedFile(Dir, IsAngled);
448 }
449
CodeCompleteNaturalLanguage()450 void Preprocessor::CodeCompleteNaturalLanguage() {
451 setCodeCompletionReached();
452 if (CodeComplete)
453 CodeComplete->CodeCompleteNaturalLanguage();
454 }
455
456 /// getSpelling - This method is used to get the spelling of a token into a
457 /// SmallVector. Note that the returned StringRef may not point to the
458 /// supplied buffer if a copy can be avoided.
getSpelling(const Token & Tok,SmallVectorImpl<char> & Buffer,bool * Invalid) const459 StringRef Preprocessor::getSpelling(const Token &Tok,
460 SmallVectorImpl<char> &Buffer,
461 bool *Invalid) const {
462 // NOTE: this has to be checked *before* testing for an IdentifierInfo.
463 if (Tok.isNot(tok::raw_identifier) && !Tok.hasUCN()) {
464 // Try the fast path.
465 if (const IdentifierInfo *II = Tok.getIdentifierInfo())
466 return II->getName();
467 }
468
469 // Resize the buffer if we need to copy into it.
470 if (Tok.needsCleaning())
471 Buffer.resize(Tok.getLength());
472
473 const char *Ptr = Buffer.data();
474 unsigned Len = getSpelling(Tok, Ptr, Invalid);
475 return StringRef(Ptr, Len);
476 }
477
478 /// CreateString - Plop the specified string into a scratch buffer and return a
479 /// location for it. If specified, the source location provides a source
480 /// location for the token.
CreateString(StringRef Str,Token & Tok,SourceLocation ExpansionLocStart,SourceLocation ExpansionLocEnd)481 void Preprocessor::CreateString(StringRef Str, Token &Tok,
482 SourceLocation ExpansionLocStart,
483 SourceLocation ExpansionLocEnd) {
484 Tok.setLength(Str.size());
485
486 const char *DestPtr;
487 SourceLocation Loc = ScratchBuf->getToken(Str.data(), Str.size(), DestPtr);
488
489 if (ExpansionLocStart.isValid())
490 Loc = SourceMgr.createExpansionLoc(Loc, ExpansionLocStart,
491 ExpansionLocEnd, Str.size());
492 Tok.setLocation(Loc);
493
494 // If this is a raw identifier or a literal token, set the pointer data.
495 if (Tok.is(tok::raw_identifier))
496 Tok.setRawIdentifierData(DestPtr);
497 else if (Tok.isLiteral())
498 Tok.setLiteralData(DestPtr);
499 }
500
SplitToken(SourceLocation Loc,unsigned Length)501 SourceLocation Preprocessor::SplitToken(SourceLocation Loc, unsigned Length) {
502 auto &SM = getSourceManager();
503 SourceLocation SpellingLoc = SM.getSpellingLoc(Loc);
504 std::pair<FileID, unsigned> LocInfo = SM.getDecomposedLoc(SpellingLoc);
505 bool Invalid = false;
506 StringRef Buffer = SM.getBufferData(LocInfo.first, &Invalid);
507 if (Invalid)
508 return SourceLocation();
509
510 // FIXME: We could consider re-using spelling for tokens we see repeatedly.
511 const char *DestPtr;
512 SourceLocation Spelling =
513 ScratchBuf->getToken(Buffer.data() + LocInfo.second, Length, DestPtr);
514 return SM.createTokenSplitLoc(Spelling, Loc, Loc.getLocWithOffset(Length));
515 }
516
getCurrentModule()517 Module *Preprocessor::getCurrentModule() {
518 if (!getLangOpts().isCompilingModule())
519 return nullptr;
520
521 return getHeaderSearchInfo().lookupModule(getLangOpts().CurrentModule);
522 }
523
524 //===----------------------------------------------------------------------===//
525 // Preprocessor Initialization Methods
526 //===----------------------------------------------------------------------===//
527
528 /// EnterMainSourceFile - Enter the specified FileID as the main source file,
529 /// which implicitly adds the builtin defines etc.
EnterMainSourceFile()530 void Preprocessor::EnterMainSourceFile() {
531 // We do not allow the preprocessor to reenter the main file. Doing so will
532 // cause FileID's to accumulate information from both runs (e.g. #line
533 // information) and predefined macros aren't guaranteed to be set properly.
534 assert(NumEnteredSourceFiles == 0 && "Cannot reenter the main file!");
535 FileID MainFileID = SourceMgr.getMainFileID();
536
537 // If MainFileID is loaded it means we loaded an AST file, no need to enter
538 // a main file.
539 if (!SourceMgr.isLoadedFileID(MainFileID)) {
540 // Enter the main file source buffer.
541 EnterSourceFile(MainFileID, nullptr, SourceLocation());
542
543 // If we've been asked to skip bytes in the main file (e.g., as part of a
544 // precompiled preamble), do so now.
545 if (SkipMainFilePreamble.first > 0)
546 CurLexer->SetByteOffset(SkipMainFilePreamble.first,
547 SkipMainFilePreamble.second);
548
549 // Tell the header info that the main file was entered. If the file is later
550 // #imported, it won't be re-entered.
551 if (const FileEntry *FE = SourceMgr.getFileEntryForID(MainFileID))
552 HeaderInfo.IncrementIncludeCount(FE);
553 }
554
555 // Preprocess Predefines to populate the initial preprocessor state.
556 std::unique_ptr<llvm::MemoryBuffer> SB =
557 llvm::MemoryBuffer::getMemBufferCopy(Predefines, "<built-in>");
558 assert(SB && "Cannot create predefined source buffer");
559 FileID FID = SourceMgr.createFileID(std::move(SB));
560 assert(FID.isValid() && "Could not create FileID for predefines?");
561 setPredefinesFileID(FID);
562
563 // Start parsing the predefines.
564 EnterSourceFile(FID, nullptr, SourceLocation());
565
566 if (!PPOpts->PCHThroughHeader.empty()) {
567 // Lookup and save the FileID for the through header. If it isn't found
568 // in the search path, it's a fatal error.
569 const DirectoryLookup *CurDir;
570 Optional<FileEntryRef> File = LookupFile(
571 SourceLocation(), PPOpts->PCHThroughHeader,
572 /*isAngled=*/false, /*FromDir=*/nullptr, /*FromFile=*/nullptr, CurDir,
573 /*SearchPath=*/nullptr, /*RelativePath=*/nullptr,
574 /*SuggestedModule=*/nullptr, /*IsMapped=*/nullptr,
575 /*IsFrameworkFound=*/nullptr);
576 if (!File) {
577 Diag(SourceLocation(), diag::err_pp_through_header_not_found)
578 << PPOpts->PCHThroughHeader;
579 return;
580 }
581 setPCHThroughHeaderFileID(
582 SourceMgr.createFileID(*File, SourceLocation(), SrcMgr::C_User));
583 }
584
585 // Skip tokens from the Predefines and if needed the main file.
586 if ((usingPCHWithThroughHeader() && SkippingUntilPCHThroughHeader) ||
587 (usingPCHWithPragmaHdrStop() && SkippingUntilPragmaHdrStop))
588 SkipTokensWhileUsingPCH();
589 }
590
setPCHThroughHeaderFileID(FileID FID)591 void Preprocessor::setPCHThroughHeaderFileID(FileID FID) {
592 assert(PCHThroughHeaderFileID.isInvalid() &&
593 "PCHThroughHeaderFileID already set!");
594 PCHThroughHeaderFileID = FID;
595 }
596
isPCHThroughHeader(const FileEntry * FE)597 bool Preprocessor::isPCHThroughHeader(const FileEntry *FE) {
598 assert(PCHThroughHeaderFileID.isValid() &&
599 "Invalid PCH through header FileID");
600 return FE == SourceMgr.getFileEntryForID(PCHThroughHeaderFileID);
601 }
602
creatingPCHWithThroughHeader()603 bool Preprocessor::creatingPCHWithThroughHeader() {
604 return TUKind == TU_Prefix && !PPOpts->PCHThroughHeader.empty() &&
605 PCHThroughHeaderFileID.isValid();
606 }
607
usingPCHWithThroughHeader()608 bool Preprocessor::usingPCHWithThroughHeader() {
609 return TUKind != TU_Prefix && !PPOpts->PCHThroughHeader.empty() &&
610 PCHThroughHeaderFileID.isValid();
611 }
612
creatingPCHWithPragmaHdrStop()613 bool Preprocessor::creatingPCHWithPragmaHdrStop() {
614 return TUKind == TU_Prefix && PPOpts->PCHWithHdrStop;
615 }
616
usingPCHWithPragmaHdrStop()617 bool Preprocessor::usingPCHWithPragmaHdrStop() {
618 return TUKind != TU_Prefix && PPOpts->PCHWithHdrStop;
619 }
620
621 /// Skip tokens until after the #include of the through header or
622 /// until after a #pragma hdrstop is seen. Tokens in the predefines file
623 /// and the main file may be skipped. If the end of the predefines file
624 /// is reached, skipping continues into the main file. If the end of the
625 /// main file is reached, it's a fatal error.
SkipTokensWhileUsingPCH()626 void Preprocessor::SkipTokensWhileUsingPCH() {
627 bool ReachedMainFileEOF = false;
628 bool UsingPCHThroughHeader = SkippingUntilPCHThroughHeader;
629 bool UsingPragmaHdrStop = SkippingUntilPragmaHdrStop;
630 Token Tok;
631 while (true) {
632 bool InPredefines =
633 (CurLexer && CurLexer->getFileID() == getPredefinesFileID());
634 switch (CurLexerKind) {
635 case CLK_Lexer:
636 CurLexer->Lex(Tok);
637 break;
638 case CLK_TokenLexer:
639 CurTokenLexer->Lex(Tok);
640 break;
641 case CLK_CachingLexer:
642 CachingLex(Tok);
643 break;
644 case CLK_LexAfterModuleImport:
645 LexAfterModuleImport(Tok);
646 break;
647 }
648 if (Tok.is(tok::eof) && !InPredefines) {
649 ReachedMainFileEOF = true;
650 break;
651 }
652 if (UsingPCHThroughHeader && !SkippingUntilPCHThroughHeader)
653 break;
654 if (UsingPragmaHdrStop && !SkippingUntilPragmaHdrStop)
655 break;
656 }
657 if (ReachedMainFileEOF) {
658 if (UsingPCHThroughHeader)
659 Diag(SourceLocation(), diag::err_pp_through_header_not_seen)
660 << PPOpts->PCHThroughHeader << 1;
661 else if (!PPOpts->PCHWithHdrStopCreate)
662 Diag(SourceLocation(), diag::err_pp_pragma_hdrstop_not_seen);
663 }
664 }
665
replayPreambleConditionalStack()666 void Preprocessor::replayPreambleConditionalStack() {
667 // Restore the conditional stack from the preamble, if there is one.
668 if (PreambleConditionalStack.isReplaying()) {
669 assert(CurPPLexer &&
670 "CurPPLexer is null when calling replayPreambleConditionalStack.");
671 CurPPLexer->setConditionalLevels(PreambleConditionalStack.getStack());
672 PreambleConditionalStack.doneReplaying();
673 if (PreambleConditionalStack.reachedEOFWhileSkipping())
674 SkipExcludedConditionalBlock(
675 PreambleConditionalStack.SkipInfo->HashTokenLoc,
676 PreambleConditionalStack.SkipInfo->IfTokenLoc,
677 PreambleConditionalStack.SkipInfo->FoundNonSkipPortion,
678 PreambleConditionalStack.SkipInfo->FoundElse,
679 PreambleConditionalStack.SkipInfo->ElseLoc);
680 }
681 }
682
EndSourceFile()683 void Preprocessor::EndSourceFile() {
684 // Notify the client that we reached the end of the source file.
685 if (Callbacks)
686 Callbacks->EndOfMainFile();
687 }
688
689 //===----------------------------------------------------------------------===//
690 // Lexer Event Handling.
691 //===----------------------------------------------------------------------===//
692
693 /// LookUpIdentifierInfo - Given a tok::raw_identifier token, look up the
694 /// identifier information for the token and install it into the token,
695 /// updating the token kind accordingly.
LookUpIdentifierInfo(Token & Identifier) const696 IdentifierInfo *Preprocessor::LookUpIdentifierInfo(Token &Identifier) const {
697 assert(!Identifier.getRawIdentifier().empty() && "No raw identifier data!");
698
699 // Look up this token, see if it is a macro, or if it is a language keyword.
700 IdentifierInfo *II;
701 if (!Identifier.needsCleaning() && !Identifier.hasUCN()) {
702 // No cleaning needed, just use the characters from the lexed buffer.
703 II = getIdentifierInfo(Identifier.getRawIdentifier());
704 } else {
705 // Cleaning needed, alloca a buffer, clean into it, then use the buffer.
706 SmallString<64> IdentifierBuffer;
707 StringRef CleanedStr = getSpelling(Identifier, IdentifierBuffer);
708
709 if (Identifier.hasUCN()) {
710 SmallString<64> UCNIdentifierBuffer;
711 expandUCNs(UCNIdentifierBuffer, CleanedStr);
712 II = getIdentifierInfo(UCNIdentifierBuffer);
713 } else {
714 II = getIdentifierInfo(CleanedStr);
715 }
716 }
717
718 // Update the token info (identifier info and appropriate token kind).
719 Identifier.setIdentifierInfo(II);
720 if (getLangOpts().MSVCCompat && II->isCPlusPlusOperatorKeyword() &&
721 getSourceManager().isInSystemHeader(Identifier.getLocation()))
722 Identifier.setKind(tok::identifier);
723 else
724 Identifier.setKind(II->getTokenID());
725
726 return II;
727 }
728
SetPoisonReason(IdentifierInfo * II,unsigned DiagID)729 void Preprocessor::SetPoisonReason(IdentifierInfo *II, unsigned DiagID) {
730 PoisonReasons[II] = DiagID;
731 }
732
PoisonSEHIdentifiers(bool Poison)733 void Preprocessor::PoisonSEHIdentifiers(bool Poison) {
734 assert(Ident__exception_code && Ident__exception_info);
735 assert(Ident___exception_code && Ident___exception_info);
736 Ident__exception_code->setIsPoisoned(Poison);
737 Ident___exception_code->setIsPoisoned(Poison);
738 Ident_GetExceptionCode->setIsPoisoned(Poison);
739 Ident__exception_info->setIsPoisoned(Poison);
740 Ident___exception_info->setIsPoisoned(Poison);
741 Ident_GetExceptionInfo->setIsPoisoned(Poison);
742 Ident__abnormal_termination->setIsPoisoned(Poison);
743 Ident___abnormal_termination->setIsPoisoned(Poison);
744 Ident_AbnormalTermination->setIsPoisoned(Poison);
745 }
746
HandlePoisonedIdentifier(Token & Identifier)747 void Preprocessor::HandlePoisonedIdentifier(Token & Identifier) {
748 assert(Identifier.getIdentifierInfo() &&
749 "Can't handle identifiers without identifier info!");
750 llvm::DenseMap<IdentifierInfo*,unsigned>::const_iterator it =
751 PoisonReasons.find(Identifier.getIdentifierInfo());
752 if(it == PoisonReasons.end())
753 Diag(Identifier, diag::err_pp_used_poisoned_id);
754 else
755 Diag(Identifier,it->second) << Identifier.getIdentifierInfo();
756 }
757
758 /// Returns a diagnostic message kind for reporting a future keyword as
759 /// appropriate for the identifier and specified language.
getFutureCompatDiagKind(const IdentifierInfo & II,const LangOptions & LangOpts)760 static diag::kind getFutureCompatDiagKind(const IdentifierInfo &II,
761 const LangOptions &LangOpts) {
762 assert(II.isFutureCompatKeyword() && "diagnostic should not be needed");
763
764 if (LangOpts.CPlusPlus)
765 return llvm::StringSwitch<diag::kind>(II.getName())
766 #define CXX11_KEYWORD(NAME, FLAGS) \
767 .Case(#NAME, diag::warn_cxx11_keyword)
768 #define CXX20_KEYWORD(NAME, FLAGS) \
769 .Case(#NAME, diag::warn_cxx20_keyword)
770 #include "clang/Basic/TokenKinds.def"
771 // char8_t is not modeled as a CXX20_KEYWORD because it's not
772 // unconditionally enabled in C++20 mode. (It can be disabled
773 // by -fno-char8_t.)
774 .Case("char8_t", diag::warn_cxx20_keyword)
775 ;
776
777 llvm_unreachable(
778 "Keyword not known to come from a newer Standard or proposed Standard");
779 }
780
updateOutOfDateIdentifier(IdentifierInfo & II) const781 void Preprocessor::updateOutOfDateIdentifier(IdentifierInfo &II) const {
782 assert(II.isOutOfDate() && "not out of date");
783 getExternalSource()->updateOutOfDateIdentifier(II);
784 }
785
786 /// HandleIdentifier - This callback is invoked when the lexer reads an
787 /// identifier. This callback looks up the identifier in the map and/or
788 /// potentially macro expands it or turns it into a named token (like 'for').
789 ///
790 /// Note that callers of this method are guarded by checking the
791 /// IdentifierInfo's 'isHandleIdentifierCase' bit. If this method changes, the
792 /// IdentifierInfo methods that compute these properties will need to change to
793 /// match.
HandleIdentifier(Token & Identifier)794 bool Preprocessor::HandleIdentifier(Token &Identifier) {
795 assert(Identifier.getIdentifierInfo() &&
796 "Can't handle identifiers without identifier info!");
797
798 IdentifierInfo &II = *Identifier.getIdentifierInfo();
799
800 // If the information about this identifier is out of date, update it from
801 // the external source.
802 // We have to treat __VA_ARGS__ in a special way, since it gets
803 // serialized with isPoisoned = true, but our preprocessor may have
804 // unpoisoned it if we're defining a C99 macro.
805 if (II.isOutOfDate()) {
806 bool CurrentIsPoisoned = false;
807 const bool IsSpecialVariadicMacro =
808 &II == Ident__VA_ARGS__ || &II == Ident__VA_OPT__;
809 if (IsSpecialVariadicMacro)
810 CurrentIsPoisoned = II.isPoisoned();
811
812 updateOutOfDateIdentifier(II);
813 Identifier.setKind(II.getTokenID());
814
815 if (IsSpecialVariadicMacro)
816 II.setIsPoisoned(CurrentIsPoisoned);
817 }
818
819 // If this identifier was poisoned, and if it was not produced from a macro
820 // expansion, emit an error.
821 if (II.isPoisoned() && CurPPLexer) {
822 HandlePoisonedIdentifier(Identifier);
823 }
824
825 // If this is a macro to be expanded, do it.
826 if (MacroDefinition MD = getMacroDefinition(&II)) {
827 auto *MI = MD.getMacroInfo();
828 assert(MI && "macro definition with no macro info?");
829 if (!DisableMacroExpansion) {
830 if (!Identifier.isExpandDisabled() && MI->isEnabled()) {
831 // C99 6.10.3p10: If the preprocessing token immediately after the
832 // macro name isn't a '(', this macro should not be expanded.
833 if (!MI->isFunctionLike() || isNextPPTokenLParen())
834 return HandleMacroExpandedIdentifier(Identifier, MD);
835 } else {
836 // C99 6.10.3.4p2 says that a disabled macro may never again be
837 // expanded, even if it's in a context where it could be expanded in the
838 // future.
839 Identifier.setFlag(Token::DisableExpand);
840 if (MI->isObjectLike() || isNextPPTokenLParen())
841 Diag(Identifier, diag::pp_disabled_macro_expansion);
842 }
843 }
844 }
845
846 // If this identifier is a keyword in a newer Standard or proposed Standard,
847 // produce a warning. Don't warn if we're not considering macro expansion,
848 // since this identifier might be the name of a macro.
849 // FIXME: This warning is disabled in cases where it shouldn't be, like
850 // "#define constexpr constexpr", "int constexpr;"
851 if (II.isFutureCompatKeyword() && !DisableMacroExpansion) {
852 Diag(Identifier, getFutureCompatDiagKind(II, getLangOpts()))
853 << II.getName();
854 // Don't diagnose this keyword again in this translation unit.
855 II.setIsFutureCompatKeyword(false);
856 }
857
858 // If this is an extension token, diagnose its use.
859 // We avoid diagnosing tokens that originate from macro definitions.
860 // FIXME: This warning is disabled in cases where it shouldn't be,
861 // like "#define TY typeof", "TY(1) x".
862 if (II.isExtensionToken() && !DisableMacroExpansion)
863 Diag(Identifier, diag::ext_token_used);
864
865 // If this is the 'import' contextual keyword following an '@', note
866 // that the next token indicates a module name.
867 //
868 // Note that we do not treat 'import' as a contextual
869 // keyword when we're in a caching lexer, because caching lexers only get
870 // used in contexts where import declarations are disallowed.
871 //
872 // Likewise if this is the C++ Modules TS import keyword.
873 if (((LastTokenWasAt && II.isModulesImport()) ||
874 Identifier.is(tok::kw_import)) &&
875 !InMacroArgs && !DisableMacroExpansion &&
876 (getLangOpts().Modules || getLangOpts().DebuggerSupport) &&
877 CurLexerKind != CLK_CachingLexer) {
878 ModuleImportLoc = Identifier.getLocation();
879 ModuleImportPath.clear();
880 ModuleImportExpectsIdentifier = true;
881 CurLexerKind = CLK_LexAfterModuleImport;
882 }
883 return true;
884 }
885
Lex(Token & Result)886 void Preprocessor::Lex(Token &Result) {
887 ++LexLevel;
888
889 // We loop here until a lex function returns a token; this avoids recursion.
890 bool ReturnedToken;
891 do {
892 switch (CurLexerKind) {
893 case CLK_Lexer:
894 ReturnedToken = CurLexer->Lex(Result);
895 break;
896 case CLK_TokenLexer:
897 ReturnedToken = CurTokenLexer->Lex(Result);
898 break;
899 case CLK_CachingLexer:
900 CachingLex(Result);
901 ReturnedToken = true;
902 break;
903 case CLK_LexAfterModuleImport:
904 ReturnedToken = LexAfterModuleImport(Result);
905 break;
906 }
907 } while (!ReturnedToken);
908
909 if (Result.is(tok::unknown) && TheModuleLoader.HadFatalFailure)
910 return;
911
912 if (Result.is(tok::code_completion) && Result.getIdentifierInfo()) {
913 // Remember the identifier before code completion token.
914 setCodeCompletionIdentifierInfo(Result.getIdentifierInfo());
915 setCodeCompletionTokenRange(Result.getLocation(), Result.getEndLoc());
916 // Set IdenfitierInfo to null to avoid confusing code that handles both
917 // identifiers and completion tokens.
918 Result.setIdentifierInfo(nullptr);
919 }
920
921 // Update ImportSeqState to track our position within a C++20 import-seq
922 // if this token is being produced as a result of phase 4 of translation.
923 if (getLangOpts().CPlusPlusModules && LexLevel == 1 &&
924 !Result.getFlag(Token::IsReinjected)) {
925 switch (Result.getKind()) {
926 case tok::l_paren: case tok::l_square: case tok::l_brace:
927 ImportSeqState.handleOpenBracket();
928 break;
929 case tok::r_paren: case tok::r_square:
930 ImportSeqState.handleCloseBracket();
931 break;
932 case tok::r_brace:
933 ImportSeqState.handleCloseBrace();
934 break;
935 case tok::semi:
936 ImportSeqState.handleSemi();
937 break;
938 case tok::header_name:
939 case tok::annot_header_unit:
940 ImportSeqState.handleHeaderName();
941 break;
942 case tok::kw_export:
943 ImportSeqState.handleExport();
944 break;
945 case tok::identifier:
946 if (Result.getIdentifierInfo()->isModulesImport()) {
947 ImportSeqState.handleImport();
948 if (ImportSeqState.afterImportSeq()) {
949 ModuleImportLoc = Result.getLocation();
950 ModuleImportPath.clear();
951 ModuleImportExpectsIdentifier = true;
952 CurLexerKind = CLK_LexAfterModuleImport;
953 }
954 break;
955 }
956 LLVM_FALLTHROUGH;
957 default:
958 ImportSeqState.handleMisc();
959 break;
960 }
961 }
962
963 LastTokenWasAt = Result.is(tok::at);
964 --LexLevel;
965
966 if ((LexLevel == 0 || PreprocessToken) &&
967 !Result.getFlag(Token::IsReinjected)) {
968 if (LexLevel == 0)
969 ++TokenCount;
970 if (OnToken)
971 OnToken(Result);
972 }
973 }
974
975 /// Lex a header-name token (including one formed from header-name-tokens if
976 /// \p AllowConcatenation is \c true).
977 ///
978 /// \param FilenameTok Filled in with the next token. On success, this will
979 /// be either a header_name token. On failure, it will be whatever other
980 /// token was found instead.
981 /// \param AllowMacroExpansion If \c true, allow the header name to be formed
982 /// by macro expansion (concatenating tokens as necessary if the first
983 /// token is a '<').
984 /// \return \c true if we reached EOD or EOF while looking for a > token in
985 /// a concatenated header name and diagnosed it. \c false otherwise.
LexHeaderName(Token & FilenameTok,bool AllowMacroExpansion)986 bool Preprocessor::LexHeaderName(Token &FilenameTok, bool AllowMacroExpansion) {
987 // Lex using header-name tokenization rules if tokens are being lexed from
988 // a file. Just grab a token normally if we're in a macro expansion.
989 if (CurPPLexer)
990 CurPPLexer->LexIncludeFilename(FilenameTok);
991 else
992 Lex(FilenameTok);
993
994 // This could be a <foo/bar.h> file coming from a macro expansion. In this
995 // case, glue the tokens together into an angle_string_literal token.
996 SmallString<128> FilenameBuffer;
997 if (FilenameTok.is(tok::less) && AllowMacroExpansion) {
998 bool StartOfLine = FilenameTok.isAtStartOfLine();
999 bool LeadingSpace = FilenameTok.hasLeadingSpace();
1000 bool LeadingEmptyMacro = FilenameTok.hasLeadingEmptyMacro();
1001
1002 SourceLocation Start = FilenameTok.getLocation();
1003 SourceLocation End;
1004 FilenameBuffer.push_back('<');
1005
1006 // Consume tokens until we find a '>'.
1007 // FIXME: A header-name could be formed starting or ending with an
1008 // alternative token. It's not clear whether that's ill-formed in all
1009 // cases.
1010 while (FilenameTok.isNot(tok::greater)) {
1011 Lex(FilenameTok);
1012 if (FilenameTok.isOneOf(tok::eod, tok::eof)) {
1013 Diag(FilenameTok.getLocation(), diag::err_expected) << tok::greater;
1014 Diag(Start, diag::note_matching) << tok::less;
1015 return true;
1016 }
1017
1018 End = FilenameTok.getLocation();
1019
1020 // FIXME: Provide code completion for #includes.
1021 if (FilenameTok.is(tok::code_completion)) {
1022 setCodeCompletionReached();
1023 Lex(FilenameTok);
1024 continue;
1025 }
1026
1027 // Append the spelling of this token to the buffer. If there was a space
1028 // before it, add it now.
1029 if (FilenameTok.hasLeadingSpace())
1030 FilenameBuffer.push_back(' ');
1031
1032 // Get the spelling of the token, directly into FilenameBuffer if
1033 // possible.
1034 size_t PreAppendSize = FilenameBuffer.size();
1035 FilenameBuffer.resize(PreAppendSize + FilenameTok.getLength());
1036
1037 const char *BufPtr = &FilenameBuffer[PreAppendSize];
1038 unsigned ActualLen = getSpelling(FilenameTok, BufPtr);
1039
1040 // If the token was spelled somewhere else, copy it into FilenameBuffer.
1041 if (BufPtr != &FilenameBuffer[PreAppendSize])
1042 memcpy(&FilenameBuffer[PreAppendSize], BufPtr, ActualLen);
1043
1044 // Resize FilenameBuffer to the correct size.
1045 if (FilenameTok.getLength() != ActualLen)
1046 FilenameBuffer.resize(PreAppendSize + ActualLen);
1047 }
1048
1049 FilenameTok.startToken();
1050 FilenameTok.setKind(tok::header_name);
1051 FilenameTok.setFlagValue(Token::StartOfLine, StartOfLine);
1052 FilenameTok.setFlagValue(Token::LeadingSpace, LeadingSpace);
1053 FilenameTok.setFlagValue(Token::LeadingEmptyMacro, LeadingEmptyMacro);
1054 CreateString(FilenameBuffer, FilenameTok, Start, End);
1055 } else if (FilenameTok.is(tok::string_literal) && AllowMacroExpansion) {
1056 // Convert a string-literal token of the form " h-char-sequence "
1057 // (produced by macro expansion) into a header-name token.
1058 //
1059 // The rules for header-names don't quite match the rules for
1060 // string-literals, but all the places where they differ result in
1061 // undefined behavior, so we can and do treat them the same.
1062 //
1063 // A string-literal with a prefix or suffix is not translated into a
1064 // header-name. This could theoretically be observable via the C++20
1065 // context-sensitive header-name formation rules.
1066 StringRef Str = getSpelling(FilenameTok, FilenameBuffer);
1067 if (Str.size() >= 2 && Str.front() == '"' && Str.back() == '"')
1068 FilenameTok.setKind(tok::header_name);
1069 }
1070
1071 return false;
1072 }
1073
1074 /// Collect the tokens of a C++20 pp-import-suffix.
CollectPpImportSuffix(SmallVectorImpl<Token> & Toks)1075 void Preprocessor::CollectPpImportSuffix(SmallVectorImpl<Token> &Toks) {
1076 // FIXME: For error recovery, consider recognizing attribute syntax here
1077 // and terminating / diagnosing a missing semicolon if we find anything
1078 // else? (Can we leave that to the parser?)
1079 unsigned BracketDepth = 0;
1080 while (true) {
1081 Toks.emplace_back();
1082 Lex(Toks.back());
1083
1084 switch (Toks.back().getKind()) {
1085 case tok::l_paren: case tok::l_square: case tok::l_brace:
1086 ++BracketDepth;
1087 break;
1088
1089 case tok::r_paren: case tok::r_square: case tok::r_brace:
1090 if (BracketDepth == 0)
1091 return;
1092 --BracketDepth;
1093 break;
1094
1095 case tok::semi:
1096 if (BracketDepth == 0)
1097 return;
1098 break;
1099
1100 case tok::eof:
1101 return;
1102
1103 default:
1104 break;
1105 }
1106 }
1107 }
1108
1109
1110 /// Lex a token following the 'import' contextual keyword.
1111 ///
1112 /// pp-import: [C++20]
1113 /// import header-name pp-import-suffix[opt] ;
1114 /// import header-name-tokens pp-import-suffix[opt] ;
1115 /// [ObjC] @ import module-name ;
1116 /// [Clang] import module-name ;
1117 ///
1118 /// header-name-tokens:
1119 /// string-literal
1120 /// < [any sequence of preprocessing-tokens other than >] >
1121 ///
1122 /// module-name:
1123 /// module-name-qualifier[opt] identifier
1124 ///
1125 /// module-name-qualifier
1126 /// module-name-qualifier[opt] identifier .
1127 ///
1128 /// We respond to a pp-import by importing macros from the named module.
LexAfterModuleImport(Token & Result)1129 bool Preprocessor::LexAfterModuleImport(Token &Result) {
1130 // Figure out what kind of lexer we actually have.
1131 recomputeCurLexerKind();
1132
1133 // Lex the next token. The header-name lexing rules are used at the start of
1134 // a pp-import.
1135 //
1136 // For now, we only support header-name imports in C++20 mode.
1137 // FIXME: Should we allow this in all language modes that support an import
1138 // declaration as an extension?
1139 if (ModuleImportPath.empty() && getLangOpts().CPlusPlusModules) {
1140 if (LexHeaderName(Result))
1141 return true;
1142 } else {
1143 Lex(Result);
1144 }
1145
1146 // Allocate a holding buffer for a sequence of tokens and introduce it into
1147 // the token stream.
1148 auto EnterTokens = [this](ArrayRef<Token> Toks) {
1149 auto ToksCopy = std::make_unique<Token[]>(Toks.size());
1150 std::copy(Toks.begin(), Toks.end(), ToksCopy.get());
1151 EnterTokenStream(std::move(ToksCopy), Toks.size(),
1152 /*DisableMacroExpansion*/ true, /*IsReinject*/ false);
1153 };
1154
1155 // Check for a header-name.
1156 SmallVector<Token, 32> Suffix;
1157 if (Result.is(tok::header_name)) {
1158 // Enter the header-name token into the token stream; a Lex action cannot
1159 // both return a token and cache tokens (doing so would corrupt the token
1160 // cache if the call to Lex comes from CachingLex / PeekAhead).
1161 Suffix.push_back(Result);
1162
1163 // Consume the pp-import-suffix and expand any macros in it now. We'll add
1164 // it back into the token stream later.
1165 CollectPpImportSuffix(Suffix);
1166 if (Suffix.back().isNot(tok::semi)) {
1167 // This is not a pp-import after all.
1168 EnterTokens(Suffix);
1169 return false;
1170 }
1171
1172 // C++2a [cpp.module]p1:
1173 // The ';' preprocessing-token terminating a pp-import shall not have
1174 // been produced by macro replacement.
1175 SourceLocation SemiLoc = Suffix.back().getLocation();
1176 if (SemiLoc.isMacroID())
1177 Diag(SemiLoc, diag::err_header_import_semi_in_macro);
1178
1179 // Reconstitute the import token.
1180 Token ImportTok;
1181 ImportTok.startToken();
1182 ImportTok.setKind(tok::kw_import);
1183 ImportTok.setLocation(ModuleImportLoc);
1184 ImportTok.setIdentifierInfo(getIdentifierInfo("import"));
1185 ImportTok.setLength(6);
1186
1187 auto Action = HandleHeaderIncludeOrImport(
1188 /*HashLoc*/ SourceLocation(), ImportTok, Suffix.front(), SemiLoc);
1189 switch (Action.Kind) {
1190 case ImportAction::None:
1191 break;
1192
1193 case ImportAction::ModuleBegin:
1194 // Let the parser know we're textually entering the module.
1195 Suffix.emplace_back();
1196 Suffix.back().startToken();
1197 Suffix.back().setKind(tok::annot_module_begin);
1198 Suffix.back().setLocation(SemiLoc);
1199 Suffix.back().setAnnotationEndLoc(SemiLoc);
1200 Suffix.back().setAnnotationValue(Action.ModuleForHeader);
1201 LLVM_FALLTHROUGH;
1202
1203 case ImportAction::ModuleImport:
1204 case ImportAction::SkippedModuleImport:
1205 // We chose to import (or textually enter) the file. Convert the
1206 // header-name token into a header unit annotation token.
1207 Suffix[0].setKind(tok::annot_header_unit);
1208 Suffix[0].setAnnotationEndLoc(Suffix[0].getLocation());
1209 Suffix[0].setAnnotationValue(Action.ModuleForHeader);
1210 // FIXME: Call the moduleImport callback?
1211 break;
1212 case ImportAction::Failure:
1213 assert(TheModuleLoader.HadFatalFailure &&
1214 "This should be an early exit only to a fatal error");
1215 Result.setKind(tok::eof);
1216 CurLexer->cutOffLexing();
1217 EnterTokens(Suffix);
1218 return true;
1219 }
1220
1221 EnterTokens(Suffix);
1222 return false;
1223 }
1224
1225 // The token sequence
1226 //
1227 // import identifier (. identifier)*
1228 //
1229 // indicates a module import directive. We already saw the 'import'
1230 // contextual keyword, so now we're looking for the identifiers.
1231 if (ModuleImportExpectsIdentifier && Result.getKind() == tok::identifier) {
1232 // We expected to see an identifier here, and we did; continue handling
1233 // identifiers.
1234 ModuleImportPath.push_back(std::make_pair(Result.getIdentifierInfo(),
1235 Result.getLocation()));
1236 ModuleImportExpectsIdentifier = false;
1237 CurLexerKind = CLK_LexAfterModuleImport;
1238 return true;
1239 }
1240
1241 // If we're expecting a '.' or a ';', and we got a '.', then wait until we
1242 // see the next identifier. (We can also see a '[[' that begins an
1243 // attribute-specifier-seq here under the C++ Modules TS.)
1244 if (!ModuleImportExpectsIdentifier && Result.getKind() == tok::period) {
1245 ModuleImportExpectsIdentifier = true;
1246 CurLexerKind = CLK_LexAfterModuleImport;
1247 return true;
1248 }
1249
1250 // If we didn't recognize a module name at all, this is not a (valid) import.
1251 if (ModuleImportPath.empty() || Result.is(tok::eof))
1252 return true;
1253
1254 // Consume the pp-import-suffix and expand any macros in it now, if we're not
1255 // at the semicolon already.
1256 SourceLocation SemiLoc = Result.getLocation();
1257 if (Result.isNot(tok::semi)) {
1258 Suffix.push_back(Result);
1259 CollectPpImportSuffix(Suffix);
1260 if (Suffix.back().isNot(tok::semi)) {
1261 // This is not an import after all.
1262 EnterTokens(Suffix);
1263 return false;
1264 }
1265 SemiLoc = Suffix.back().getLocation();
1266 }
1267
1268 // Under the Modules TS, the dot is just part of the module name, and not
1269 // a real hierarchy separator. Flatten such module names now.
1270 //
1271 // FIXME: Is this the right level to be performing this transformation?
1272 std::string FlatModuleName;
1273 if (getLangOpts().ModulesTS || getLangOpts().CPlusPlusModules) {
1274 for (auto &Piece : ModuleImportPath) {
1275 if (!FlatModuleName.empty())
1276 FlatModuleName += ".";
1277 FlatModuleName += Piece.first->getName();
1278 }
1279 SourceLocation FirstPathLoc = ModuleImportPath[0].second;
1280 ModuleImportPath.clear();
1281 ModuleImportPath.push_back(
1282 std::make_pair(getIdentifierInfo(FlatModuleName), FirstPathLoc));
1283 }
1284
1285 Module *Imported = nullptr;
1286 if (getLangOpts().Modules) {
1287 Imported = TheModuleLoader.loadModule(ModuleImportLoc,
1288 ModuleImportPath,
1289 Module::Hidden,
1290 /*IsInclusionDirective=*/false);
1291 if (Imported)
1292 makeModuleVisible(Imported, SemiLoc);
1293 }
1294 if (Callbacks)
1295 Callbacks->moduleImport(ModuleImportLoc, ModuleImportPath, Imported);
1296
1297 if (!Suffix.empty()) {
1298 EnterTokens(Suffix);
1299 return false;
1300 }
1301 return true;
1302 }
1303
makeModuleVisible(Module * M,SourceLocation Loc)1304 void Preprocessor::makeModuleVisible(Module *M, SourceLocation Loc) {
1305 CurSubmoduleState->VisibleModules.setVisible(
1306 M, Loc, [](Module *) {},
1307 [&](ArrayRef<Module *> Path, Module *Conflict, StringRef Message) {
1308 // FIXME: Include the path in the diagnostic.
1309 // FIXME: Include the import location for the conflicting module.
1310 Diag(ModuleImportLoc, diag::warn_module_conflict)
1311 << Path[0]->getFullModuleName()
1312 << Conflict->getFullModuleName()
1313 << Message;
1314 });
1315
1316 // Add this module to the imports list of the currently-built submodule.
1317 if (!BuildingSubmoduleStack.empty() && M != BuildingSubmoduleStack.back().M)
1318 BuildingSubmoduleStack.back().M->Imports.insert(M);
1319 }
1320
FinishLexStringLiteral(Token & Result,std::string & String,const char * DiagnosticTag,bool AllowMacroExpansion)1321 bool Preprocessor::FinishLexStringLiteral(Token &Result, std::string &String,
1322 const char *DiagnosticTag,
1323 bool AllowMacroExpansion) {
1324 // We need at least one string literal.
1325 if (Result.isNot(tok::string_literal)) {
1326 Diag(Result, diag::err_expected_string_literal)
1327 << /*Source='in...'*/0 << DiagnosticTag;
1328 return false;
1329 }
1330
1331 // Lex string literal tokens, optionally with macro expansion.
1332 SmallVector<Token, 4> StrToks;
1333 do {
1334 StrToks.push_back(Result);
1335
1336 if (Result.hasUDSuffix())
1337 Diag(Result, diag::err_invalid_string_udl);
1338
1339 if (AllowMacroExpansion)
1340 Lex(Result);
1341 else
1342 LexUnexpandedToken(Result);
1343 } while (Result.is(tok::string_literal));
1344
1345 // Concatenate and parse the strings.
1346 StringLiteralParser Literal(StrToks, *this);
1347 assert(Literal.isAscii() && "Didn't allow wide strings in");
1348
1349 if (Literal.hadError)
1350 return false;
1351
1352 if (Literal.Pascal) {
1353 Diag(StrToks[0].getLocation(), diag::err_expected_string_literal)
1354 << /*Source='in...'*/0 << DiagnosticTag;
1355 return false;
1356 }
1357
1358 String = std::string(Literal.GetString());
1359 return true;
1360 }
1361
parseSimpleIntegerLiteral(Token & Tok,uint64_t & Value)1362 bool Preprocessor::parseSimpleIntegerLiteral(Token &Tok, uint64_t &Value) {
1363 assert(Tok.is(tok::numeric_constant));
1364 SmallString<8> IntegerBuffer;
1365 bool NumberInvalid = false;
1366 StringRef Spelling = getSpelling(Tok, IntegerBuffer, &NumberInvalid);
1367 if (NumberInvalid)
1368 return false;
1369 NumericLiteralParser Literal(Spelling, Tok.getLocation(), getSourceManager(),
1370 getLangOpts(), getTargetInfo(),
1371 getDiagnostics());
1372 if (Literal.hadError || !Literal.isIntegerLiteral() || Literal.hasUDSuffix())
1373 return false;
1374 llvm::APInt APVal(64, 0);
1375 if (Literal.GetIntegerValue(APVal))
1376 return false;
1377 Lex(Tok);
1378 Value = APVal.getLimitedValue();
1379 return true;
1380 }
1381
addCommentHandler(CommentHandler * Handler)1382 void Preprocessor::addCommentHandler(CommentHandler *Handler) {
1383 assert(Handler && "NULL comment handler");
1384 assert(llvm::find(CommentHandlers, Handler) == CommentHandlers.end() &&
1385 "Comment handler already registered");
1386 CommentHandlers.push_back(Handler);
1387 }
1388
removeCommentHandler(CommentHandler * Handler)1389 void Preprocessor::removeCommentHandler(CommentHandler *Handler) {
1390 std::vector<CommentHandler *>::iterator Pos =
1391 llvm::find(CommentHandlers, Handler);
1392 assert(Pos != CommentHandlers.end() && "Comment handler not registered");
1393 CommentHandlers.erase(Pos);
1394 }
1395
HandleComment(Token & result,SourceRange Comment)1396 bool Preprocessor::HandleComment(Token &result, SourceRange Comment) {
1397 bool AnyPendingTokens = false;
1398 for (std::vector<CommentHandler *>::iterator H = CommentHandlers.begin(),
1399 HEnd = CommentHandlers.end();
1400 H != HEnd; ++H) {
1401 if ((*H)->HandleComment(*this, Comment))
1402 AnyPendingTokens = true;
1403 }
1404 if (!AnyPendingTokens || getCommentRetentionState())
1405 return false;
1406 Lex(result);
1407 return true;
1408 }
1409
1410 ModuleLoader::~ModuleLoader() = default;
1411
1412 CommentHandler::~CommentHandler() = default;
1413
1414 EmptylineHandler::~EmptylineHandler() = default;
1415
1416 CodeCompletionHandler::~CodeCompletionHandler() = default;
1417
createPreprocessingRecord()1418 void Preprocessor::createPreprocessingRecord() {
1419 if (Record)
1420 return;
1421
1422 Record = new PreprocessingRecord(getSourceManager());
1423 addPPCallbacks(std::unique_ptr<PPCallbacks>(Record));
1424 }
1425