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/IdentifierTable.h"
31#include "clang/Basic/LLVM.h"
32#include "clang/Basic/LangOptions.h"
33#include "clang/Basic/Module.h"
34#include "clang/Basic/SourceLocation.h"
35#include "clang/Basic/SourceManager.h"
36#include "clang/Basic/TargetInfo.h"
37#include "clang/Lex/CodeCompletionHandler.h"
38#include "clang/Lex/DependencyDirectivesScanner.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/NoTrivialPPDirectiveTracer.h"
48#include "clang/Lex/Pragma.h"
49#include "clang/Lex/PreprocessingRecord.h"
50#include "clang/Lex/PreprocessorLexer.h"
51#include "clang/Lex/PreprocessorOptions.h"
52#include "clang/Lex/ScratchBuffer.h"
53#include "clang/Lex/Token.h"
54#include "clang/Lex/TokenLexer.h"
55#include "llvm/ADT/APInt.h"
56#include "llvm/ADT/ArrayRef.h"
57#include "llvm/ADT/DenseMap.h"
58#include "llvm/ADT/STLExtras.h"
59#include "llvm/ADT/ScopeExit.h"
60#include "llvm/ADT/SmallVector.h"
61#include "llvm/ADT/StringRef.h"
62#include "llvm/Support/Capacity.h"
63#include "llvm/Support/ErrorHandling.h"
64#include "llvm/Support/FormatVariadic.h"
65#include "llvm/Support/MemoryBuffer.h"
66#include "llvm/Support/MemoryBufferRef.h"
67#include "llvm/Support/SaveAndRestore.h"
68#include "llvm/Support/raw_ostream.h"
69#include <algorithm>
70#include <cassert>
71#include <memory>
72#include <optional>
73#include <string>
74#include <utility>
75#include <vector>
76
77using namespace clang;
78
79/// Minimum distance between two check points, in tokens.
80static constexpr unsigned CheckPointStepSize = 1024;
81
82LLVM_INSTANTIATE_REGISTRY(PragmaHandlerRegistry)
83
84ExternalPreprocessorSource::~ExternalPreprocessorSource() = default;
85
86Preprocessor::Preprocessor(const PreprocessorOptions &PPOpts,
87 DiagnosticsEngine &diags, const LangOptions &opts,
88 SourceManager &SM, HeaderSearch &Headers,
89 ModuleLoader &TheModuleLoader,
90 IdentifierInfoLookup *IILookup, bool OwnsHeaders,
91 TranslationUnitKind TUKind)
92 : PPOpts(PPOpts), Diags(&diags), LangOpts(opts),
93 FileMgr(Headers.getFileMgr()), SourceMgr(SM),
94 ScratchBuf(new ScratchBuffer(SourceMgr)), HeaderInfo(Headers),
95 TheModuleLoader(TheModuleLoader), ExternalSource(nullptr),
96 // As the language options may have not been loaded yet (when
97 // deserializing an ASTUnit), adding keywords to the identifier table is
98 // deferred to Preprocessor::Initialize().
99 Identifiers(IILookup), PragmaHandlers(new PragmaNamespace(StringRef())),
100 TUKind(TUKind), SkipMainFilePreamble(0, true),
101 CurSubmoduleState(&NullSubmoduleState) {
102 OwnsHeaderSearch = OwnsHeaders;
103
104 // Default to discarding comments.
105 KeepComments = false;
106 KeepMacroComments = false;
107 SuppressIncludeNotFoundError = false;
108
109 // Macro expansion is enabled.
110 DisableMacroExpansion = false;
111 MacroExpansionInDirectivesOverride = false;
112 InMacroArgs = false;
113 ArgMacro = nullptr;
114 InMacroArgPreExpansion = false;
115 NumCachedTokenLexers = 0;
116 PragmasEnabled = true;
117 ParsingIfOrElifDirective = false;
118 PreprocessedOutput = false;
119
120 // We haven't read anything from the external source.
121 ReadMacrosFromExternalSource = false;
122
123 LastExportKeyword.startToken();
124
125 BuiltinInfo = std::make_unique<Builtin::Context>();
126
127 // "Poison" __VA_ARGS__, __VA_OPT__ which can only appear in the expansion of
128 // a macro. They get unpoisoned where it is allowed.
129 (Ident__VA_ARGS__ = getIdentifierInfo(Name: "__VA_ARGS__"))->setIsPoisoned();
130 SetPoisonReason(II: Ident__VA_ARGS__,DiagID: diag::ext_pp_bad_vaargs_use);
131 (Ident__VA_OPT__ = getIdentifierInfo(Name: "__VA_OPT__"))->setIsPoisoned();
132 SetPoisonReason(II: Ident__VA_OPT__,DiagID: diag::ext_pp_bad_vaopt_use);
133
134 // Initialize the pragma handlers.
135 RegisterBuiltinPragmas();
136
137 // Initialize builtin macros like __LINE__ and friends.
138 RegisterBuiltinMacros();
139
140 if(LangOpts.Borland) {
141 Ident__exception_info = getIdentifierInfo(Name: "_exception_info");
142 Ident___exception_info = getIdentifierInfo(Name: "__exception_info");
143 Ident_GetExceptionInfo = getIdentifierInfo(Name: "GetExceptionInformation");
144 Ident__exception_code = getIdentifierInfo(Name: "_exception_code");
145 Ident___exception_code = getIdentifierInfo(Name: "__exception_code");
146 Ident_GetExceptionCode = getIdentifierInfo(Name: "GetExceptionCode");
147 Ident__abnormal_termination = getIdentifierInfo(Name: "_abnormal_termination");
148 Ident___abnormal_termination = getIdentifierInfo(Name: "__abnormal_termination");
149 Ident_AbnormalTermination = getIdentifierInfo(Name: "AbnormalTermination");
150 } else {
151 Ident__exception_info = Ident__exception_code = nullptr;
152 Ident__abnormal_termination = Ident___exception_info = nullptr;
153 Ident___exception_code = Ident___abnormal_termination = nullptr;
154 Ident_GetExceptionInfo = Ident_GetExceptionCode = nullptr;
155 Ident_AbnormalTermination = nullptr;
156 }
157
158 // Default incremental processing to -fincremental-extensions, clients can
159 // override with `enableIncrementalProcessing` if desired.
160 IncrementalProcessing = LangOpts.IncrementalExtensions;
161
162 // If using a PCH where a #pragma hdrstop is expected, start skipping tokens.
163 if (usingPCHWithPragmaHdrStop())
164 SkippingUntilPragmaHdrStop = true;
165
166 // If using a PCH with a through header, start skipping tokens.
167 if (!this->PPOpts.PCHThroughHeader.empty() &&
168 !this->PPOpts.ImplicitPCHInclude.empty())
169 SkippingUntilPCHThroughHeader = true;
170
171 if (this->PPOpts.GeneratePreamble)
172 PreambleConditionalStack.startRecording();
173
174 MaxTokens = LangOpts.MaxTokens;
175}
176
177Preprocessor::~Preprocessor() {
178 assert(!isBacktrackEnabled() && "EnableBacktrack/Backtrack imbalance!");
179
180 IncludeMacroStack.clear();
181
182 // Free any cached macro expanders.
183 // This populates MacroArgCache, so all TokenLexers need to be destroyed
184 // before the code below that frees up the MacroArgCache list.
185 std::fill(first: TokenLexerCache, last: TokenLexerCache + NumCachedTokenLexers, value: nullptr);
186 CurTokenLexer.reset();
187
188 // Free any cached MacroArgs.
189 for (MacroArgs *ArgList = MacroArgCache; ArgList;)
190 ArgList = ArgList->deallocate();
191
192 // Delete the header search info, if we own it.
193 if (OwnsHeaderSearch)
194 delete &HeaderInfo;
195}
196
197void Preprocessor::Initialize(const TargetInfo &Target,
198 const TargetInfo *AuxTarget) {
199 assert((!this->Target || this->Target == &Target) &&
200 "Invalid override of target information");
201 this->Target = &Target;
202
203 assert((!this->AuxTarget || this->AuxTarget == AuxTarget) &&
204 "Invalid override of aux target information.");
205 this->AuxTarget = AuxTarget;
206
207 // Initialize information about built-ins.
208 BuiltinInfo->InitializeTarget(Target, AuxTarget);
209 HeaderInfo.setTarget(Target);
210
211 // Populate the identifier table with info about keywords for the current language.
212 Identifiers.AddKeywords(LangOpts);
213
214 // Initialize the __FTL_EVAL_METHOD__ macro to the TargetInfo.
215 setTUFPEvalMethod(getTargetInfo().getFPEvalMethod());
216
217 if (getLangOpts().getFPEvalMethod() == LangOptions::FEM_UnsetOnCommandLine)
218 // Use setting from TargetInfo.
219 setCurrentFPEvalMethod(PragmaLoc: SourceLocation(), Val: Target.getFPEvalMethod());
220 else
221 // Set initial value of __FLT_EVAL_METHOD__ from the command line.
222 setCurrentFPEvalMethod(PragmaLoc: SourceLocation(), Val: getLangOpts().getFPEvalMethod());
223}
224
225void Preprocessor::InitializeForModelFile() {
226 NumEnteredSourceFiles = 0;
227
228 // Reset pragmas
229 PragmaHandlersBackup = std::move(PragmaHandlers);
230 PragmaHandlers = std::make_unique<PragmaNamespace>(args: StringRef());
231 RegisterBuiltinPragmas();
232
233 // Reset PredefinesFileID
234 PredefinesFileID = FileID();
235}
236
237void Preprocessor::FinalizeForModelFile() {
238 NumEnteredSourceFiles = 1;
239
240 PragmaHandlers = std::move(PragmaHandlersBackup);
241}
242
243void Preprocessor::DumpToken(const Token &Tok, bool DumpFlags) const {
244 std::string TokenStr;
245 llvm::raw_string_ostream OS(TokenStr);
246
247 // The alignment of 16 is chosen to comfortably fit most identifiers.
248 OS << llvm::formatv(Fmt: "{0,-16} ", Vals: tok::getTokenName(Kind: Tok.getKind()));
249
250 // Annotation tokens are just markers that don't have a spelling -- they
251 // indicate where something expanded.
252 if (!Tok.isAnnotation()) {
253 OS << "'";
254 // Escape string to prevent token spelling from spanning multiple lines.
255 OS.write_escaped(Str: getSpelling(Tok));
256 OS << "'";
257 }
258
259 // The alignment of 48 (32 characters for the spelling + the 16 for
260 // the identifier name) fits most variable names, keywords and annotations.
261 llvm::errs() << llvm::formatv(Fmt: "{0,-48} ", Vals&: OS.str());
262
263 if (!DumpFlags) return;
264
265 auto Loc = Tok.getLocation();
266 llvm::errs() << "Loc=<";
267 DumpLocation(Loc);
268 llvm::errs() << ">";
269
270 // If the token points directly to a file location (i.e. not a macro
271 // expansion), then add additional padding so that trailing markers
272 // align, provided the line/column numbers are reasonably sized.
273 //
274 // Otherwise, if it's a macro expansion, don't bother with alignment,
275 // as the line will include multiple locations and be very long.
276 //
277 // NOTE: To keep this stateless, it doesn't account for filename
278 // length, so when a header starts markers will be temporarily misaligned.
279 if (Loc.isFileID()) {
280 PresumedLoc PLoc = SourceMgr.getPresumedLoc(Loc);
281
282 if (!PLoc.isInvalid()) {
283 int LineWidth = llvm::utostr(X: PLoc.getLine()).size();
284 int ColumnWidth = llvm::utostr(X: PLoc.getColumn()).size();
285
286 // Reserve space for lines up to 9999 and columns up to 99,
287 // which is 4 + 2 = 6 characters in total.
288 const int ReservedSpace = 6;
289
290 int LeftSpace = ReservedSpace - LineWidth - ColumnWidth;
291 int Padding = std::max<int>(a: 0, b: LeftSpace);
292
293 llvm::errs().indent(NumSpaces: Padding);
294 }
295 }
296
297 if (Tok.isAtStartOfLine())
298 llvm::errs() << " [StartOfLine]";
299 if (Tok.hasLeadingSpace())
300 llvm::errs() << " [LeadingSpace]";
301 if (Tok.isExpandDisabled())
302 llvm::errs() << " [ExpandDisabled]";
303 if (Tok.needsCleaning()) {
304 const char *Start = SourceMgr.getCharacterData(SL: Tok.getLocation());
305 llvm::errs() << " [UnClean='" << StringRef(Start, Tok.getLength()) << "']";
306 }
307}
308
309void Preprocessor::DumpLocation(SourceLocation Loc) const {
310 Loc.print(OS&: llvm::errs(), SM: SourceMgr);
311}
312
313void Preprocessor::DumpMacro(const MacroInfo &MI) const {
314 llvm::errs() << "MACRO: ";
315 for (unsigned i = 0, e = MI.getNumTokens(); i != e; ++i) {
316 DumpToken(Tok: MI.getReplacementToken(Tok: i));
317 llvm::errs() << " ";
318 }
319 llvm::errs() << "\n";
320}
321
322void Preprocessor::PrintStats() {
323 llvm::errs() << "\n*** Preprocessor Stats:\n";
324 llvm::errs() << NumDirectives << " directives found:\n";
325 llvm::errs() << " " << NumDefined << " #define.\n";
326 llvm::errs() << " " << NumUndefined << " #undef.\n";
327 llvm::errs() << " #include/#include_next/#import:\n";
328 llvm::errs() << " " << NumEnteredSourceFiles << " source files entered.\n";
329 llvm::errs() << " " << MaxIncludeStackDepth << " max include stack depth\n";
330 llvm::errs() << " " << NumIf << " #if/#ifndef/#ifdef.\n";
331 llvm::errs() << " " << NumElse << " #else/#elif/#elifdef/#elifndef.\n";
332 llvm::errs() << " " << NumEndif << " #endif.\n";
333 llvm::errs() << " " << NumPragma << " #pragma.\n";
334 llvm::errs() << NumSkipped << " #if/#ifndef#ifdef regions skipped\n";
335
336 llvm::errs() << NumMacroExpanded << "/" << NumFnMacroExpanded << "/"
337 << NumBuiltinMacroExpanded << " obj/fn/builtin macros expanded, "
338 << NumFastMacroExpanded << " on the fast path.\n";
339 llvm::errs() << (NumFastTokenPaste+NumTokenPaste)
340 << " token paste (##) operations performed, "
341 << NumFastTokenPaste << " on the fast path.\n";
342
343 llvm::errs() << "\nPreprocessor Memory: " << getTotalMemory() << "B total";
344
345 llvm::errs() << "\n BumpPtr: " << BP.getTotalMemory();
346 llvm::errs() << "\n Macro Expanded Tokens: "
347 << llvm::capacity_in_bytes(X: MacroExpandedTokens);
348 llvm::errs() << "\n Predefines Buffer: " << Predefines.capacity();
349 // FIXME: List information for all submodules.
350 llvm::errs() << "\n Macros: "
351 << llvm::capacity_in_bytes(X: CurSubmoduleState->Macros);
352 llvm::errs() << "\n #pragma push_macro Info: "
353 << llvm::capacity_in_bytes(X: PragmaPushMacroInfo);
354 llvm::errs() << "\n Poison Reasons: "
355 << llvm::capacity_in_bytes(X: PoisonReasons);
356 llvm::errs() << "\n Comment Handlers: "
357 << llvm::capacity_in_bytes(x: CommentHandlers) << "\n";
358}
359
360Preprocessor::macro_iterator
361Preprocessor::macro_begin(bool IncludeExternalMacros) const {
362 if (IncludeExternalMacros && ExternalSource &&
363 !ReadMacrosFromExternalSource) {
364 ReadMacrosFromExternalSource = true;
365 ExternalSource->ReadDefinedMacros();
366 }
367
368 // Make sure we cover all macros in visible modules.
369 for (const ModuleMacro &Macro : ModuleMacros)
370 CurSubmoduleState->Macros.try_emplace(Key: Macro.II);
371
372 return CurSubmoduleState->Macros.begin();
373}
374
375size_t Preprocessor::getTotalMemory() const {
376 return BP.getTotalMemory()
377 + llvm::capacity_in_bytes(X: MacroExpandedTokens)
378 + Predefines.capacity() /* Predefines buffer. */
379 // FIXME: Include sizes from all submodules, and include MacroInfo sizes,
380 // and ModuleMacros.
381 + llvm::capacity_in_bytes(X: CurSubmoduleState->Macros)
382 + llvm::capacity_in_bytes(X: PragmaPushMacroInfo)
383 + llvm::capacity_in_bytes(X: PoisonReasons)
384 + llvm::capacity_in_bytes(x: CommentHandlers);
385}
386
387Preprocessor::macro_iterator
388Preprocessor::macro_end(bool IncludeExternalMacros) const {
389 if (IncludeExternalMacros && ExternalSource &&
390 !ReadMacrosFromExternalSource) {
391 ReadMacrosFromExternalSource = true;
392 ExternalSource->ReadDefinedMacros();
393 }
394
395 return CurSubmoduleState->Macros.end();
396}
397
398/// Compares macro tokens with a specified token value sequence.
399static bool MacroDefinitionEquals(const MacroInfo *MI,
400 ArrayRef<TokenValue> Tokens) {
401 return Tokens.size() == MI->getNumTokens() &&
402 std::equal(first1: Tokens.begin(), last1: Tokens.end(), first2: MI->tokens_begin());
403}
404
405StringRef Preprocessor::getLastMacroWithSpelling(
406 SourceLocation Loc,
407 ArrayRef<TokenValue> Tokens) const {
408 SourceLocation BestLocation;
409 StringRef BestSpelling;
410 for (Preprocessor::macro_iterator I = macro_begin(), E = macro_end();
411 I != E; ++I) {
412 const MacroDirective::DefInfo
413 Def = I->second.findDirectiveAtLoc(Loc, SourceMgr);
414 if (!Def || !Def.getMacroInfo())
415 continue;
416 if (!Def.getMacroInfo()->isObjectLike())
417 continue;
418 if (!MacroDefinitionEquals(MI: Def.getMacroInfo(), Tokens))
419 continue;
420 SourceLocation Location = Def.getLocation();
421 // Choose the macro defined latest.
422 if (BestLocation.isInvalid() ||
423 (Location.isValid() &&
424 SourceMgr.isBeforeInTranslationUnit(LHS: BestLocation, RHS: Location))) {
425 BestLocation = Location;
426 BestSpelling = I->first->getName();
427 }
428 }
429 return BestSpelling;
430}
431
432void Preprocessor::recomputeCurLexerKind() {
433 if (CurLexer)
434 CurLexerCallback = CurLexer->isDependencyDirectivesLexer()
435 ? CLK_DependencyDirectivesLexer
436 : CLK_Lexer;
437 else if (CurTokenLexer)
438 CurLexerCallback = CLK_TokenLexer;
439 else
440 CurLexerCallback = CLK_CachingLexer;
441}
442
443bool Preprocessor::SetCodeCompletionPoint(FileEntryRef File,
444 unsigned CompleteLine,
445 unsigned CompleteColumn) {
446 assert(CompleteLine && CompleteColumn && "Starts from 1:1");
447 assert(!CodeCompletionFile && "Already set");
448
449 // Load the actual file's contents.
450 std::optional<llvm::MemoryBufferRef> Buffer =
451 SourceMgr.getMemoryBufferForFileOrNone(File);
452 if (!Buffer)
453 return true;
454
455 // Find the byte position of the truncation point.
456 const char *Position = Buffer->getBufferStart();
457 for (unsigned Line = 1; Line < CompleteLine; ++Line) {
458 for (; *Position; ++Position) {
459 if (*Position != '\r' && *Position != '\n')
460 continue;
461
462 // Eat \r\n or \n\r as a single line.
463 if ((Position[1] == '\r' || Position[1] == '\n') &&
464 Position[0] != Position[1])
465 ++Position;
466 ++Position;
467 break;
468 }
469 }
470
471 Position += CompleteColumn - 1;
472
473 // If pointing inside the preamble, adjust the position at the beginning of
474 // the file after the preamble.
475 if (SkipMainFilePreamble.first &&
476 SourceMgr.getFileEntryForID(FID: SourceMgr.getMainFileID()) == File) {
477 if (Position - Buffer->getBufferStart() < SkipMainFilePreamble.first)
478 Position = Buffer->getBufferStart() + SkipMainFilePreamble.first;
479 }
480
481 if (Position > Buffer->getBufferEnd())
482 Position = Buffer->getBufferEnd();
483
484 CodeCompletionFile = File;
485 CodeCompletionOffset = Position - Buffer->getBufferStart();
486
487 auto NewBuffer = llvm::WritableMemoryBuffer::getNewUninitMemBuffer(
488 Size: Buffer->getBufferSize() + 1, BufferName: Buffer->getBufferIdentifier());
489 char *NewBuf = NewBuffer->getBufferStart();
490 char *NewPos = std::copy(first: Buffer->getBufferStart(), last: Position, result: NewBuf);
491 *NewPos = '\0';
492 std::copy(first: Position, last: Buffer->getBufferEnd(), result: NewPos+1);
493 SourceMgr.overrideFileContents(SourceFile: File, Buffer: std::move(NewBuffer));
494
495 return false;
496}
497
498void Preprocessor::CodeCompleteIncludedFile(llvm::StringRef Dir,
499 bool IsAngled) {
500 setCodeCompletionReached();
501 if (CodeComplete)
502 CodeComplete->CodeCompleteIncludedFile(Dir, IsAngled);
503}
504
505void Preprocessor::CodeCompleteNaturalLanguage() {
506 setCodeCompletionReached();
507 if (CodeComplete)
508 CodeComplete->CodeCompleteNaturalLanguage();
509}
510
511/// getSpelling - This method is used to get the spelling of a token into a
512/// SmallVector. Note that the returned StringRef may not point to the
513/// supplied buffer if a copy can be avoided.
514StringRef Preprocessor::getSpelling(const Token &Tok,
515 SmallVectorImpl<char> &Buffer,
516 bool *Invalid) const {
517 // NOTE: this has to be checked *before* testing for an IdentifierInfo.
518 if (Tok.isNot(K: tok::raw_identifier) && !Tok.hasUCN()) {
519 // Try the fast path.
520 if (const IdentifierInfo *II = Tok.getIdentifierInfo())
521 return II->getName();
522 }
523
524 // Resize the buffer if we need to copy into it.
525 if (Tok.needsCleaning())
526 Buffer.resize(N: Tok.getLength());
527
528 const char *Ptr = Buffer.data();
529 unsigned Len = getSpelling(Tok, Buffer&: Ptr, Invalid);
530 return StringRef(Ptr, Len);
531}
532
533/// CreateString - Plop the specified string into a scratch buffer and return a
534/// location for it. If specified, the source location provides a source
535/// location for the token.
536void Preprocessor::CreateString(StringRef Str, Token &Tok,
537 SourceLocation ExpansionLocStart,
538 SourceLocation ExpansionLocEnd) {
539 Tok.setLength(Str.size());
540
541 const char *DestPtr;
542 SourceLocation Loc = ScratchBuf->getToken(Buf: Str.data(), Len: Str.size(), DestPtr);
543
544 if (ExpansionLocStart.isValid())
545 Loc = SourceMgr.createExpansionLoc(SpellingLoc: Loc, ExpansionLocStart,
546 ExpansionLocEnd, Length: Str.size());
547 Tok.setLocation(Loc);
548
549 // If this is a raw identifier or a literal token, set the pointer data.
550 if (Tok.is(K: tok::raw_identifier))
551 Tok.setRawIdentifierData(DestPtr);
552 else if (Tok.isLiteral())
553 Tok.setLiteralData(DestPtr);
554}
555
556SourceLocation Preprocessor::SplitToken(SourceLocation Loc, unsigned Length) {
557 auto &SM = getSourceManager();
558 SourceLocation SpellingLoc = SM.getSpellingLoc(Loc);
559 FileIDAndOffset LocInfo = SM.getDecomposedLoc(Loc: SpellingLoc);
560 bool Invalid = false;
561 StringRef Buffer = SM.getBufferData(FID: LocInfo.first, Invalid: &Invalid);
562 if (Invalid)
563 return SourceLocation();
564
565 // FIXME: We could consider re-using spelling for tokens we see repeatedly.
566 const char *DestPtr;
567 SourceLocation Spelling =
568 ScratchBuf->getToken(Buf: Buffer.data() + LocInfo.second, Len: Length, DestPtr);
569 return SM.createTokenSplitLoc(SpellingLoc: Spelling, TokenStart: Loc, TokenEnd: Loc.getLocWithOffset(Offset: Length));
570}
571
572Module *Preprocessor::getCurrentModule() {
573 if (!getLangOpts().isCompilingModule())
574 return nullptr;
575
576 return getHeaderSearchInfo().lookupModule(ModuleName: getLangOpts().CurrentModule);
577}
578
579Module *Preprocessor::getCurrentModuleImplementation() {
580 if (!getLangOpts().isCompilingModuleImplementation())
581 return nullptr;
582
583 return getHeaderSearchInfo().lookupModule(ModuleName: getLangOpts().ModuleName);
584}
585
586//===----------------------------------------------------------------------===//
587// Preprocessor Initialization Methods
588//===----------------------------------------------------------------------===//
589
590/// EnterMainSourceFile - Enter the specified FileID as the main source file,
591/// which implicitly adds the builtin defines etc.
592void Preprocessor::EnterMainSourceFile() {
593 // We do not allow the preprocessor to reenter the main file. Doing so will
594 // cause FileID's to accumulate information from both runs (e.g. #line
595 // information) and predefined macros aren't guaranteed to be set properly.
596 assert(NumEnteredSourceFiles == 0 && "Cannot reenter the main file!");
597 FileID MainFileID = SourceMgr.getMainFileID();
598
599 // If MainFileID is loaded it means we loaded an AST file, no need to enter
600 // a main file.
601 if (!SourceMgr.isLoadedFileID(FID: MainFileID)) {
602 // Enter the main file source buffer.
603 EnterSourceFile(FID: MainFileID, Dir: nullptr, Loc: SourceLocation());
604
605 // If we've been asked to skip bytes in the main file (e.g., as part of a
606 // precompiled preamble), do so now.
607 if (SkipMainFilePreamble.first > 0)
608 CurLexer->SetByteOffset(Offset: SkipMainFilePreamble.first,
609 StartOfLine: SkipMainFilePreamble.second);
610
611 // Tell the header info that the main file was entered. If the file is later
612 // #imported, it won't be re-entered.
613 if (OptionalFileEntryRef FE = SourceMgr.getFileEntryRefForID(FID: MainFileID))
614 markIncluded(File: *FE);
615
616 // Record the first PP token in the main file. This is used to generate
617 // better diagnostics for C++ modules.
618 //
619 // // This is a comment.
620 // #define FOO int // note: add 'module;' to the start of the file
621 // ^ FirstPPToken // to introduce a global module fragment.
622 //
623 // export module M; // error: module declaration must occur
624 // // at the start of the translation unit.
625 if (getLangOpts().CPlusPlusModules) {
626 std::optional<StringRef> Input =
627 getSourceManager().getBufferDataOrNone(FID: MainFileID);
628 if (!isPreprocessedModuleFile() && Input)
629 MainFileIsPreprocessedModuleFile =
630 clang::isPreprocessedModuleFile(Source: *Input);
631 auto Tracer = std::make_unique<NoTrivialPPDirectiveTracer>(args&: *this);
632 DirTracer = Tracer.get();
633 addPPCallbacks(C: std::move(Tracer));
634 std::optional<Token> FirstPPTok = CurLexer->peekNextPPToken();
635 if (FirstPPTok)
636 FirstPPTokenLoc = FirstPPTok->getLocation();
637 }
638 }
639
640 // Preprocess Predefines to populate the initial preprocessor state.
641 std::unique_ptr<llvm::MemoryBuffer> SB =
642 llvm::MemoryBuffer::getMemBufferCopy(InputData: Predefines, BufferName: "<built-in>");
643 assert(SB && "Cannot create predefined source buffer");
644 FileID FID = SourceMgr.createFileID(Buffer: std::move(SB));
645 assert(FID.isValid() && "Could not create FileID for predefines?");
646 setPredefinesFileID(FID);
647
648 // Start parsing the predefines.
649 EnterSourceFile(FID, Dir: nullptr, Loc: SourceLocation());
650
651 if (!PPOpts.PCHThroughHeader.empty()) {
652 // Lookup and save the FileID for the through header. If it isn't found
653 // in the search path, it's a fatal error.
654 OptionalFileEntryRef File = LookupFile(
655 FilenameLoc: SourceLocation(), Filename: PPOpts.PCHThroughHeader,
656 /*isAngled=*/false, /*FromDir=*/nullptr, /*FromFile=*/nullptr,
657 /*CurDir=*/nullptr, /*SearchPath=*/nullptr, /*RelativePath=*/nullptr,
658 /*SuggestedModule=*/nullptr, /*IsMapped=*/nullptr,
659 /*IsFrameworkFound=*/nullptr);
660 if (!File) {
661 Diag(Loc: SourceLocation(), DiagID: diag::err_pp_through_header_not_found)
662 << PPOpts.PCHThroughHeader;
663 return;
664 }
665 setPCHThroughHeaderFileID(
666 SourceMgr.createFileID(SourceFile: *File, IncludePos: SourceLocation(), FileCharacter: SrcMgr::C_User));
667 }
668
669 // Skip tokens from the Predefines and if needed the main file.
670 if ((usingPCHWithThroughHeader() && SkippingUntilPCHThroughHeader) ||
671 (usingPCHWithPragmaHdrStop() && SkippingUntilPragmaHdrStop))
672 SkipTokensWhileUsingPCH();
673}
674
675void Preprocessor::setPCHThroughHeaderFileID(FileID FID) {
676 assert(PCHThroughHeaderFileID.isInvalid() &&
677 "PCHThroughHeaderFileID already set!");
678 PCHThroughHeaderFileID = FID;
679}
680
681bool Preprocessor::isPCHThroughHeader(const FileEntry *FE) {
682 assert(PCHThroughHeaderFileID.isValid() &&
683 "Invalid PCH through header FileID");
684 return FE == SourceMgr.getFileEntryForID(FID: PCHThroughHeaderFileID);
685}
686
687bool Preprocessor::creatingPCHWithThroughHeader() {
688 return TUKind == TU_Prefix && !PPOpts.PCHThroughHeader.empty() &&
689 PCHThroughHeaderFileID.isValid();
690}
691
692bool Preprocessor::usingPCHWithThroughHeader() {
693 return TUKind != TU_Prefix && !PPOpts.PCHThroughHeader.empty() &&
694 PCHThroughHeaderFileID.isValid();
695}
696
697bool Preprocessor::creatingPCHWithPragmaHdrStop() {
698 return TUKind == TU_Prefix && PPOpts.PCHWithHdrStop;
699}
700
701bool Preprocessor::usingPCHWithPragmaHdrStop() {
702 return TUKind != TU_Prefix && PPOpts.PCHWithHdrStop;
703}
704
705/// Skip tokens until after the #include of the through header or
706/// until after a #pragma hdrstop is seen. Tokens in the predefines file
707/// and the main file may be skipped. If the end of the predefines file
708/// is reached, skipping continues into the main file. If the end of the
709/// main file is reached, it's a fatal error.
710void Preprocessor::SkipTokensWhileUsingPCH() {
711 bool ReachedMainFileEOF = false;
712 bool UsingPCHThroughHeader = SkippingUntilPCHThroughHeader;
713 bool UsingPragmaHdrStop = SkippingUntilPragmaHdrStop;
714 Token Tok;
715 while (true) {
716 bool InPredefines =
717 (CurLexer && CurLexer->getFileID() == getPredefinesFileID());
718 CurLexerCallback(*this, Tok);
719 if (Tok.is(K: tok::eof) && !InPredefines) {
720 ReachedMainFileEOF = true;
721 break;
722 }
723 if (UsingPCHThroughHeader && !SkippingUntilPCHThroughHeader)
724 break;
725 if (UsingPragmaHdrStop && !SkippingUntilPragmaHdrStop)
726 break;
727 }
728 if (ReachedMainFileEOF) {
729 if (UsingPCHThroughHeader)
730 Diag(Loc: SourceLocation(), DiagID: diag::err_pp_through_header_not_seen)
731 << PPOpts.PCHThroughHeader << 1;
732 else if (!PPOpts.PCHWithHdrStopCreate)
733 Diag(Loc: SourceLocation(), DiagID: diag::err_pp_pragma_hdrstop_not_seen);
734 }
735}
736
737void Preprocessor::replayPreambleConditionalStack() {
738 // Restore the conditional stack from the preamble, if there is one.
739 if (PreambleConditionalStack.isReplaying()) {
740 assert(CurPPLexer &&
741 "CurPPLexer is null when calling replayPreambleConditionalStack.");
742 CurPPLexer->setConditionalLevels(PreambleConditionalStack.getStack());
743 PreambleConditionalStack.doneReplaying();
744 if (PreambleConditionalStack.reachedEOFWhileSkipping())
745 SkipExcludedConditionalBlock(
746 HashTokenLoc: PreambleConditionalStack.SkipInfo->HashTokenLoc,
747 IfTokenLoc: PreambleConditionalStack.SkipInfo->IfTokenLoc,
748 FoundNonSkipPortion: PreambleConditionalStack.SkipInfo->FoundNonSkipPortion,
749 FoundElse: PreambleConditionalStack.SkipInfo->FoundElse,
750 ElseLoc: PreambleConditionalStack.SkipInfo->ElseLoc);
751 }
752}
753
754void Preprocessor::EndSourceFile() {
755 // Notify the client that we reached the end of the source file.
756 if (Callbacks)
757 Callbacks->EndOfMainFile();
758}
759
760//===----------------------------------------------------------------------===//
761// Lexer Event Handling.
762//===----------------------------------------------------------------------===//
763
764/// LookUpIdentifierInfo - Given a tok::raw_identifier token, look up the
765/// identifier information for the token and install it into the token,
766/// updating the token kind accordingly.
767IdentifierInfo *Preprocessor::LookUpIdentifierInfo(Token &Identifier) const {
768 assert(!Identifier.getRawIdentifier().empty() && "No raw identifier data!");
769
770 // Look up this token, see if it is a macro, or if it is a language keyword.
771 IdentifierInfo *II;
772 if (!Identifier.needsCleaning() && !Identifier.hasUCN()) {
773 // No cleaning needed, just use the characters from the lexed buffer.
774 II = getIdentifierInfo(Name: Identifier.getRawIdentifier());
775 } else {
776 // Cleaning needed, alloca a buffer, clean into it, then use the buffer.
777 SmallString<64> IdentifierBuffer;
778 StringRef CleanedStr = getSpelling(Tok: Identifier, Buffer&: IdentifierBuffer);
779
780 if (Identifier.hasUCN()) {
781 SmallString<64> UCNIdentifierBuffer;
782 expandUCNs(Buf&: UCNIdentifierBuffer, Input: CleanedStr);
783 II = getIdentifierInfo(Name: UCNIdentifierBuffer);
784 } else {
785 II = getIdentifierInfo(Name: CleanedStr);
786 }
787 }
788
789 // Update the token info (identifier info and appropriate token kind).
790 // FIXME: the raw_identifier may contain leading whitespace which is removed
791 // from the cleaned identifier token. The SourceLocation should be updated to
792 // refer to the non-whitespace character. For instance, the text "\\\nB" (a
793 // line continuation before 'B') is parsed as a single tok::raw_identifier and
794 // is cleaned to tok::identifier "B". After cleaning the token's length is
795 // still 3 and the SourceLocation refers to the location of the backslash.
796 Identifier.setIdentifierInfo(II);
797 Identifier.setKind(II->getTokenID());
798
799 return II;
800}
801
802void Preprocessor::SetPoisonReason(IdentifierInfo *II, unsigned DiagID) {
803 PoisonReasons[II] = DiagID;
804}
805
806void Preprocessor::PoisonSEHIdentifiers(bool Poison) {
807 assert(Ident__exception_code && Ident__exception_info);
808 assert(Ident___exception_code && Ident___exception_info);
809 Ident__exception_code->setIsPoisoned(Poison);
810 Ident___exception_code->setIsPoisoned(Poison);
811 Ident_GetExceptionCode->setIsPoisoned(Poison);
812 Ident__exception_info->setIsPoisoned(Poison);
813 Ident___exception_info->setIsPoisoned(Poison);
814 Ident_GetExceptionInfo->setIsPoisoned(Poison);
815 Ident__abnormal_termination->setIsPoisoned(Poison);
816 Ident___abnormal_termination->setIsPoisoned(Poison);
817 Ident_AbnormalTermination->setIsPoisoned(Poison);
818}
819
820void Preprocessor::HandlePoisonedIdentifier(Token & Identifier) {
821 assert(Identifier.getIdentifierInfo() &&
822 "Can't handle identifiers without identifier info!");
823 llvm::DenseMap<IdentifierInfo*,unsigned>::const_iterator it =
824 PoisonReasons.find(Val: Identifier.getIdentifierInfo());
825 if(it == PoisonReasons.end())
826 Diag(Tok: Identifier, DiagID: diag::err_pp_used_poisoned_id);
827 else
828 Diag(Tok: Identifier,DiagID: it->second) << Identifier.getIdentifierInfo();
829}
830
831void Preprocessor::updateOutOfDateIdentifier(const IdentifierInfo &II) const {
832 assert(II.isOutOfDate() && "not out of date");
833 assert(getExternalSource() &&
834 "getExternalSource() should not return nullptr");
835 getExternalSource()->updateOutOfDateIdentifier(II);
836}
837
838/// HandleIdentifier - This callback is invoked when the lexer reads an
839/// identifier. This callback looks up the identifier in the map and/or
840/// potentially macro expands it or turns it into a named token (like 'for').
841///
842/// Note that callers of this method are guarded by checking the
843/// IdentifierInfo's 'isHandleIdentifierCase' bit. If this method changes, the
844/// IdentifierInfo methods that compute these properties will need to change to
845/// match.
846bool Preprocessor::HandleIdentifier(Token &Identifier) {
847 assert(Identifier.getIdentifierInfo() &&
848 "Can't handle identifiers without identifier info!");
849
850 IdentifierInfo &II = *Identifier.getIdentifierInfo();
851
852 // If the information about this identifier is out of date, update it from
853 // the external source.
854 // We have to treat __VA_ARGS__ in a special way, since it gets
855 // serialized with isPoisoned = true, but our preprocessor may have
856 // unpoisoned it if we're defining a C99 macro.
857 if (II.isOutOfDate()) {
858 bool CurrentIsPoisoned = false;
859 const bool IsSpecialVariadicMacro =
860 &II == Ident__VA_ARGS__ || &II == Ident__VA_OPT__;
861 if (IsSpecialVariadicMacro)
862 CurrentIsPoisoned = II.isPoisoned();
863
864 updateOutOfDateIdentifier(II);
865 Identifier.setKind(II.getTokenID());
866
867 if (IsSpecialVariadicMacro)
868 II.setIsPoisoned(CurrentIsPoisoned);
869 }
870
871 // If this identifier was poisoned, and if it was not produced from a macro
872 // expansion, emit an error.
873 if (II.isPoisoned() && CurPPLexer) {
874 HandlePoisonedIdentifier(Identifier);
875 }
876
877 // If this is a macro to be expanded, do it.
878 if (const MacroDefinition MD = getMacroDefinition(II: &II)) {
879 const auto *MI = MD.getMacroInfo();
880 assert(MI && "macro definition with no macro info?");
881 if (!DisableMacroExpansion) {
882 if (!Identifier.isExpandDisabled() && MI->isEnabled()) {
883 // C99 6.10.3p10: If the preprocessing token immediately after the
884 // macro name isn't a '(', this macro should not be expanded.
885 if (!MI->isFunctionLike() || isNextPPTokenOneOf(Ks: tok::l_paren))
886 return HandleMacroExpandedIdentifier(Identifier, MD);
887 } else {
888 // C99 6.10.3.4p2 says that a disabled macro may never again be
889 // expanded, even if it's in a context where it could be expanded in the
890 // future.
891 Identifier.setFlag(Token::DisableExpand);
892 if (MI->isObjectLike() || isNextPPTokenOneOf(Ks: tok::l_paren))
893 Diag(Tok: Identifier, DiagID: diag::pp_disabled_macro_expansion);
894 }
895 }
896 }
897
898 // If this identifier is a keyword in a newer Standard or proposed Standard,
899 // produce a warning. Don't warn if we're not considering macro expansion,
900 // since this identifier might be the name of a macro.
901 // FIXME: This warning is disabled in cases where it shouldn't be, like
902 // "#define constexpr constexpr", "int constexpr;"
903 if (II.isFutureCompatKeyword() && !DisableMacroExpansion) {
904 Diag(Tok: Identifier, DiagID: getIdentifierTable().getFutureCompatDiagKind(II, LangOpts: getLangOpts()))
905 << II.getName();
906 // Don't diagnose this keyword again in this translation unit.
907 II.setIsFutureCompatKeyword(false);
908 }
909
910 // If this identifier would be a keyword in C++, diagnose as a compatibility
911 // issue.
912 if (II.IsKeywordInCPlusPlus() && !DisableMacroExpansion)
913 Diag(Tok: Identifier, DiagID: diag::warn_pp_identifier_is_cpp_keyword) << &II;
914
915 // If this is an extension token, diagnose its use.
916 // We avoid diagnosing tokens that originate from macro definitions.
917 // FIXME: This warning is disabled in cases where it shouldn't be,
918 // like "#define TY typeof", "TY(1) x".
919 if (II.isExtensionToken() && !DisableMacroExpansion)
920 Diag(Tok: Identifier, DiagID: diag::ext_token_used);
921
922 // Handle module contextual keywords.
923 if (getLangOpts().CPlusPlusModules && CurLexer &&
924 !CurLexer->isLexingRawMode() && !CurLexer->isPragmaLexer() &&
925 !CurLexer->ParsingPreprocessorDirective &&
926 Identifier.isModuleContextualKeyword() &&
927 HandleModuleContextualKeyword(Result&: Identifier)) {
928 HandleDirective(Result&: Identifier);
929 // With a fatal failure in the module loader, we abort parsing.
930 return hadModuleLoaderFatalFailure();
931 }
932
933 // If this is the 'import' contextual keyword following an '@', note
934 // that the next token indicates a module name.
935 //
936 // Note that we do not treat 'import' as a contextual
937 // keyword when we're in a caching lexer, because caching lexers only get
938 // used in contexts where import declarations are disallowed.
939 //
940 // Likewise if this is the standard C++ import keyword.
941 if (((LastTokenWasAt && II.isImportKeyword()) ||
942 Identifier.is(K: tok::kw_import)) &&
943 !InMacroArgs &&
944 (!DisableMacroExpansion || MacroExpansionInDirectivesOverride) &&
945 CurLexerCallback != CLK_CachingLexer) {
946 ModuleImportLoc = Identifier.getLocation();
947 IsAtImport = true;
948 CurLexerCallback = CLK_LexAfterModuleImport;
949 }
950 return true;
951}
952
953void Preprocessor::Lex(Token &Result) {
954 ++LexLevel;
955
956 // We loop here until a lex function returns a token; this avoids recursion.
957 while (!CurLexerCallback(*this, Result))
958 ;
959
960 if (Result.is(K: tok::unknown) && TheModuleLoader.HadFatalFailure)
961 return;
962
963 if (Result.is(K: tok::code_completion) && Result.getIdentifierInfo()) {
964 // Remember the identifier before code completion token.
965 setCodeCompletionIdentifierInfo(Result.getIdentifierInfo());
966 setCodeCompletionTokenRange(Start: Result.getLocation(), End: Result.getEndLoc());
967 // Set IdenfitierInfo to null to avoid confusing code that handles both
968 // identifiers and completion tokens.
969 Result.setIdentifierInfo(nullptr);
970 }
971
972 // Update StdCXXImportSeqState to track our position within a C++20 import-seq
973 // if this token is being produced as a result of phase 4 of translation.
974 // Update TrackGMFState to decide if we are currently in a Global Module
975 // Fragment. GMF state updates should precede StdCXXImportSeq ones, since GMF state
976 // depends on the prevailing StdCXXImportSeq state in two cases.
977 if (getLangOpts().CPlusPlusModules && LexLevel == 1 &&
978 !Result.getFlag(Flag: Token::IsReinjected)) {
979 switch (Result.getKind()) {
980 case tok::l_paren: case tok::l_square: case tok::l_brace:
981 StdCXXImportSeqState.handleOpenBracket();
982 break;
983 case tok::r_paren: case tok::r_square:
984 StdCXXImportSeqState.handleCloseBracket();
985 break;
986 case tok::r_brace:
987 StdCXXImportSeqState.handleCloseBrace();
988 break;
989#define PRAGMA_ANNOTATION(X) case tok::annot_##X:
990// For `#pragma ...` mimic ';'.
991#include "clang/Basic/TokenKinds.def"
992#undef PRAGMA_ANNOTATION
993 // This token is injected to represent the translation of '#include "a.h"'
994 // into "import a.h;". Mimic the notional ';'.
995 case tok::annot_module_include:
996 case tok::annot_repl_input_end:
997 case tok::semi:
998 TrackGMFState.handleSemi();
999 StdCXXImportSeqState.handleSemi();
1000 ModuleDeclState.handleSemi();
1001 break;
1002 case tok::header_name:
1003 case tok::annot_header_unit:
1004 StdCXXImportSeqState.handleHeaderName();
1005 break;
1006 case tok::kw_export:
1007 if (hasSeenNoTrivialPPDirective())
1008 Result.setFlag(Token::HasSeenNoTrivialPPDirective);
1009 TrackGMFState.handleExport();
1010 StdCXXImportSeqState.handleExport();
1011 ModuleDeclState.handleExport();
1012 break;
1013 case tok::colon:
1014 ModuleDeclState.handleColon();
1015 break;
1016 case tok::kw_import:
1017 if (StdCXXImportSeqState.atTopLevel()) {
1018 TrackGMFState.handleImport(AfterTopLevelTokenSeq: StdCXXImportSeqState.afterTopLevelSeq());
1019 StdCXXImportSeqState.handleImport();
1020 }
1021 break;
1022 case tok::kw_module:
1023 if (StdCXXImportSeqState.atTopLevel()) {
1024 if (hasSeenNoTrivialPPDirective())
1025 Result.setFlag(Token::HasSeenNoTrivialPPDirective);
1026 TrackGMFState.handleModule(AfterTopLevelTokenSeq: StdCXXImportSeqState.afterTopLevelSeq());
1027 ModuleDeclState.handleModule();
1028 }
1029 break;
1030 case tok::annot_module_name:
1031 ModuleDeclState.handleModuleName(
1032 NameLoc: static_cast<ModuleNameLoc *>(Result.getAnnotationValue()));
1033 if (ModuleDeclState.isModuleCandidate())
1034 break;
1035 [[fallthrough]];
1036 default:
1037 TrackGMFState.handleMisc();
1038 StdCXXImportSeqState.handleMisc();
1039 ModuleDeclState.handleMisc();
1040 break;
1041 }
1042 }
1043
1044 if (CurLexer && ++CheckPointCounter == CheckPointStepSize) {
1045 CheckPoints[CurLexer->getFileID()].push_back(Elt: CurLexer->BufferPtr);
1046 CheckPointCounter = 0;
1047 }
1048
1049 LastTokenWasAt = Result.is(K: tok::at);
1050 if (Result.isNot(K: tok::kw_export))
1051 LastExportKeyword.startToken();
1052
1053 --LexLevel;
1054
1055 // Destroy any lexers that were deferred while we were in nested Lex calls.
1056 // This must happen after decrementing LexLevel but before any other
1057 // processing that might re-enter Lex.
1058 if (LexLevel == 0 && !PendingDestroyLexers.empty())
1059 PendingDestroyLexers.clear();
1060
1061 if ((LexLevel == 0 || PreprocessToken) &&
1062 !Result.getFlag(Flag: Token::IsReinjected)) {
1063 if (LexLevel == 0)
1064 ++TokenCount;
1065 if (OnToken)
1066 OnToken(Result);
1067 }
1068}
1069
1070void Preprocessor::LexTokensUntilEOF(std::vector<Token> *Tokens) {
1071 while (1) {
1072 Token Tok;
1073 Lex(Result&: Tok);
1074 if (Tok.isOneOf(Ks: tok::unknown, Ks: tok::eof, Ks: tok::eod,
1075 Ks: tok::annot_repl_input_end))
1076 break;
1077 if (Tokens != nullptr)
1078 Tokens->push_back(x: Tok);
1079 }
1080}
1081
1082/// Lex a header-name token (including one formed from header-name-tokens if
1083/// \p AllowMacroExpansion is \c true).
1084///
1085/// \param FilenameTok Filled in with the next token. On success, this will
1086/// be either a header_name token. On failure, it will be whatever other
1087/// token was found instead.
1088/// \param AllowMacroExpansion If \c true, allow the header name to be formed
1089/// by macro expansion (concatenating tokens as necessary if the first
1090/// token is a '<').
1091/// \return \c true if we reached EOD or EOF while looking for a > token in
1092/// a concatenated header name and diagnosed it. \c false otherwise.
1093bool Preprocessor::LexHeaderName(Token &FilenameTok, bool AllowMacroExpansion) {
1094 // Lex using header-name tokenization rules if tokens are being lexed from
1095 // a file. Just grab a token normally if we're in a macro expansion.
1096 if (CurPPLexer) {
1097 // Avoid nested header-name lexing when macro expansion recurses
1098 // __has_include(__has_include))
1099 if (CurPPLexer->ParsingFilename)
1100 LexUnexpandedToken(Result&: FilenameTok);
1101 else
1102 CurPPLexer->LexIncludeFilename(FilenameTok);
1103 } else {
1104 Lex(Result&: FilenameTok);
1105 }
1106
1107 // This could be a <foo/bar.h> file coming from a macro expansion. In this
1108 // case, glue the tokens together into an angle_string_literal token.
1109 SmallString<128> FilenameBuffer;
1110 if (FilenameTok.is(K: tok::less) && AllowMacroExpansion) {
1111 bool StartOfLine = FilenameTok.isAtStartOfLine();
1112 bool LeadingSpace = FilenameTok.hasLeadingSpace();
1113 bool LeadingEmptyMacro = FilenameTok.hasLeadingEmptyMacro();
1114
1115 SourceLocation Start = FilenameTok.getLocation();
1116 SourceLocation End;
1117 FilenameBuffer.push_back(Elt: '<');
1118
1119 // Consume tokens until we find a '>'.
1120 // FIXME: A header-name could be formed starting or ending with an
1121 // alternative token. It's not clear whether that's ill-formed in all
1122 // cases.
1123 while (FilenameTok.isNot(K: tok::greater)) {
1124 Lex(Result&: FilenameTok);
1125 if (FilenameTok.isOneOf(Ks: tok::eod, Ks: tok::eof)) {
1126 Diag(Loc: FilenameTok.getLocation(), DiagID: diag::err_expected) << tok::greater;
1127 Diag(Loc: Start, DiagID: diag::note_matching) << tok::less;
1128 return true;
1129 }
1130
1131 End = FilenameTok.getLocation();
1132
1133 // FIXME: Provide code completion for #includes.
1134 if (FilenameTok.is(K: tok::code_completion)) {
1135 setCodeCompletionReached();
1136 Lex(Result&: FilenameTok);
1137 continue;
1138 }
1139
1140 // Append the spelling of this token to the buffer. If there was a space
1141 // before it, add it now.
1142 if (FilenameTok.hasLeadingSpace())
1143 FilenameBuffer.push_back(Elt: ' ');
1144
1145 // Get the spelling of the token, directly into FilenameBuffer if
1146 // possible.
1147 size_t PreAppendSize = FilenameBuffer.size();
1148 FilenameBuffer.resize(N: PreAppendSize + FilenameTok.getLength());
1149
1150 const char *BufPtr = &FilenameBuffer[PreAppendSize];
1151 unsigned ActualLen = getSpelling(Tok: FilenameTok, Buffer&: BufPtr);
1152
1153 // If the token was spelled somewhere else, copy it into FilenameBuffer.
1154 if (BufPtr != &FilenameBuffer[PreAppendSize])
1155 memcpy(dest: &FilenameBuffer[PreAppendSize], src: BufPtr, n: ActualLen);
1156
1157 // Resize FilenameBuffer to the correct size.
1158 if (FilenameTok.getLength() != ActualLen)
1159 FilenameBuffer.resize(N: PreAppendSize + ActualLen);
1160 }
1161
1162 FilenameTok.startToken();
1163 FilenameTok.setKind(tok::header_name);
1164 FilenameTok.setFlagValue(Flag: Token::StartOfLine, Val: StartOfLine);
1165 FilenameTok.setFlagValue(Flag: Token::LeadingSpace, Val: LeadingSpace);
1166 FilenameTok.setFlagValue(Flag: Token::LeadingEmptyMacro, Val: LeadingEmptyMacro);
1167 CreateString(Str: FilenameBuffer, Tok&: FilenameTok, ExpansionLocStart: Start, ExpansionLocEnd: End);
1168 } else if (FilenameTok.is(K: tok::string_literal) && AllowMacroExpansion) {
1169 // Convert a string-literal token of the form " h-char-sequence "
1170 // (produced by macro expansion) into a header-name token.
1171 //
1172 // The rules for header-names don't quite match the rules for
1173 // string-literals, but all the places where they differ result in
1174 // undefined behavior, so we can and do treat them the same.
1175 //
1176 // A string-literal with a prefix or suffix is not translated into a
1177 // header-name. This could theoretically be observable via the C++20
1178 // context-sensitive header-name formation rules.
1179 StringRef Str = getSpelling(Tok: FilenameTok, Buffer&: FilenameBuffer);
1180 if (Str.size() >= 2 && Str.front() == '"' && Str.back() == '"')
1181 FilenameTok.setKind(tok::header_name);
1182 }
1183
1184 return false;
1185}
1186
1187std::optional<Token> Preprocessor::peekNextPPToken() const {
1188 // Do some quick tests for rejection cases.
1189 std::optional<Token> Val;
1190 if (CurLexer)
1191 Val = CurLexer->peekNextPPToken();
1192 else
1193 Val = CurTokenLexer->peekNextPPToken();
1194
1195 if (!Val) {
1196 // We have run off the end. If it's a source file we don't
1197 // examine enclosing ones (C99 5.1.1.2p4). Otherwise walk up the
1198 // macro stack.
1199 if (CurPPLexer)
1200 return std::nullopt;
1201 for (const IncludeStackInfo &Entry : llvm::reverse(C: IncludeMacroStack)) {
1202 if (Entry.TheLexer)
1203 Val = Entry.TheLexer->peekNextPPToken();
1204 else
1205 Val = Entry.TheTokenLexer->peekNextPPToken();
1206
1207 if (Val)
1208 break;
1209
1210 // Ran off the end of a source file?
1211 if (Entry.ThePPLexer)
1212 return std::nullopt;
1213 }
1214 }
1215
1216 // Okay, we found the token and return. Otherwise we found the end of the
1217 // translation unit.
1218 return Val;
1219}
1220
1221// We represent the primary and partition names as 'Paths' which are sections
1222// of the hierarchical access path for a clang module. However for C++20
1223// the periods in a name are just another character, and we will need to
1224// flatten them into a string.
1225std::string ModuleLoader::getFlatNameFromPath(ModuleIdPath Path) {
1226 std::string Name;
1227 if (Path.empty())
1228 return Name;
1229
1230 for (auto &Piece : Path) {
1231 assert(Piece.getIdentifierInfo() && Piece.getLoc().isValid());
1232 if (!Name.empty())
1233 Name += ".";
1234 Name += Piece.getIdentifierInfo()->getName();
1235 }
1236 return Name;
1237}
1238
1239ModuleNameLoc *ModuleNameLoc::Create(Preprocessor &PP, ModuleIdPath Path) {
1240 assert(!Path.empty() && "expect at least one identifier in a module name");
1241 void *Mem = PP.getPreprocessorAllocator().Allocate(
1242 Size: totalSizeToAlloc<IdentifierLoc>(Counts: Path.size()), Alignment: alignof(ModuleNameLoc));
1243 return new (Mem) ModuleNameLoc(Path);
1244}
1245
1246bool Preprocessor::LexModuleNameContinue(Token &Tok, SourceLocation UseLoc,
1247 SmallVectorImpl<Token> &Suffix,
1248 SmallVectorImpl<IdentifierLoc> &Path,
1249 bool AllowMacroExpansion,
1250 bool IsPartition) {
1251 auto ConsumeToken = [&]() {
1252 if (AllowMacroExpansion)
1253 Lex(Result&: Tok);
1254 else
1255 LexUnexpandedToken(Result&: Tok);
1256 Suffix.push_back(Elt: Tok);
1257 };
1258
1259 while (true) {
1260 if (Tok.isNot(K: tok::identifier)) {
1261 if (Tok.is(K: tok::code_completion)) {
1262 CurLexer->cutOffLexing();
1263 CodeComplete->CodeCompleteModuleImport(ImportLoc: UseLoc, Path);
1264 return true;
1265 }
1266
1267 Diag(Tok, DiagID: diag::err_pp_module_expected_ident) << Path.empty();
1268 return true;
1269 }
1270
1271 // [cpp.pre]/p2:
1272 // No identifier in the pp-module-name or pp-module-partition shall
1273 // currently be defined as an object-like macro.
1274 if (MacroInfo *MI = getMacroInfo(II: Tok.getIdentifierInfo());
1275 MI && MI->isObjectLike() && getLangOpts().CPlusPlus20 &&
1276 !AllowMacroExpansion) {
1277 Diag(Tok, DiagID: diag::err_pp_module_name_is_macro)
1278 << IsPartition << Tok.getIdentifierInfo();
1279 Diag(Loc: MI->getDefinitionLoc(), DiagID: diag::note_macro_here)
1280 << Tok.getIdentifierInfo();
1281 }
1282
1283 // Record this part of the module path.
1284 Path.emplace_back(Args: Tok.getLocation(), Args: Tok.getIdentifierInfo());
1285 ConsumeToken();
1286
1287 if (Tok.isNot(K: tok::period))
1288 return false;
1289
1290 ConsumeToken();
1291 }
1292}
1293
1294bool Preprocessor::HandleModuleName(StringRef DirType, SourceLocation UseLoc,
1295 Token &Tok,
1296 SmallVectorImpl<IdentifierLoc> &Path,
1297 SmallVectorImpl<Token> &DirToks,
1298 bool AllowMacroExpansion,
1299 bool IsPartition) {
1300 bool LeadingSpace = Tok.hasLeadingSpace();
1301 unsigned NumToksInDirective = DirToks.size();
1302 if (LexModuleNameContinue(Tok, UseLoc, Suffix&: DirToks, Path, AllowMacroExpansion,
1303 IsPartition)) {
1304 if (Tok.isNot(K: tok::eod))
1305 CheckEndOfDirective(DirType,
1306 /*EnableMacros=*/false, ExtraToks: &DirToks);
1307 EnterModuleSuffixTokenStream(Toks: DirToks);
1308 return true;
1309 }
1310
1311 // Clean the module-name tokens and replace these tokens with
1312 // annot_module_name.
1313 DirToks.resize(N: NumToksInDirective);
1314 ModuleNameLoc *NameLoc = ModuleNameLoc::Create(PP&: *this, Path);
1315 DirToks.emplace_back();
1316 DirToks.back().setKind(tok::annot_module_name);
1317 DirToks.back().setAnnotationRange(NameLoc->getRange());
1318 DirToks.back().setAnnotationValue(static_cast<void *>(NameLoc));
1319 DirToks.back().setFlagValue(Flag: Token::LeadingSpace, Val: LeadingSpace);
1320 DirToks.push_back(Elt: Tok);
1321 return false;
1322}
1323
1324/// [cpp.pre]/p2:
1325/// A preprocessing directive consists of a sequence of preprocessing tokens
1326/// that satisfies the following constraints: At the start of translation phase
1327/// 4, the first preprocessing token in the sequence, referred to as a
1328/// directive-introducing token, begins with the first character in the source
1329/// file (optionally after whitespace containing no new-line characters) or
1330/// follows whitespace containing at least one new-line character, and is:
1331/// - a # preprocessing token, or
1332/// - an import preprocessing token immediately followed on the same logical
1333/// source line by a header-name, <, identifier, or : preprocessing token, or
1334/// - a module preprocessing token immediately followed on the same logical
1335/// source line by an identifier, :, or ; preprocessing token, or
1336/// - an export preprocessing token immediately followed on the same logical
1337/// source line by one of the two preceding forms.
1338///
1339///
1340/// At the start of phase 4 an import or module token is treated as starting a
1341/// directive and are converted to their respective keywords iff:
1342/// - After skipping horizontal whitespace are
1343/// - at the start of a logical line, or
1344/// - preceded by an 'export' at the start of the logical line.
1345/// - Are followed by an identifier pp token (before macro expansion), or
1346/// - <, ", or : (but not ::) pp tokens for 'import', or
1347/// - ; for 'module'
1348/// Otherwise the token is treated as an identifier.
1349bool Preprocessor::HandleModuleContextualKeyword(Token &Result) {
1350 if (!getLangOpts().CPlusPlusModules || !Result.isModuleContextualKeyword())
1351 return false;
1352
1353 if (Result.is(K: tok::kw_export)) {
1354 LastExportKeyword = Result;
1355 return false;
1356 }
1357
1358 /// Trait 'module' and 'import' as a identifier when the main file is a
1359 /// preprocessed module file. We only allow '__preprocessed_module' and
1360 /// '__preprocessed_import' in this context.
1361 IdentifierInfo *II = Result.getIdentifierInfo();
1362 if (isPreprocessedModuleFile() &&
1363 (II->isStr(Str: tok::getKeywordSpelling(Kind: tok::kw_import)) ||
1364 II->isStr(Str: tok::getKeywordSpelling(Kind: tok::kw_module))))
1365 return false;
1366
1367 if (LastExportKeyword.is(K: tok::kw_export)) {
1368 // The export keyword was not at the start of line, it's not a
1369 // directive-introducing token.
1370 if (!LastExportKeyword.isAtPhysicalStartOfLine())
1371 return false;
1372 // [cpp.pre]/1.4
1373 // export // not a preprocessing directive
1374 // import foo; // preprocessing directive (ill-formed at phase7)
1375 if (Result.isAtPhysicalStartOfLine())
1376 return false;
1377 } else if (!Result.isAtPhysicalStartOfLine())
1378 return false;
1379
1380 llvm::SaveAndRestore<bool> SavedParsingPreprocessorDirective(
1381 CurPPLexer->ParsingPreprocessorDirective, true);
1382
1383 // The next token may be an angled string literal after import keyword.
1384 llvm::SaveAndRestore<bool> SavedParsingFilemame(
1385 CurPPLexer->ParsingFilename,
1386 Result.getIdentifierInfo()->isImportKeyword());
1387
1388 std::optional<Token> NextTok =
1389 CurLexer ? CurLexer->peekNextPPToken() : CurTokenLexer->peekNextPPToken();
1390 if (!NextTok)
1391 return false;
1392
1393 if (NextTok->is(K: tok::raw_identifier))
1394 LookUpIdentifierInfo(Identifier&: *NextTok);
1395
1396 if (Result.getIdentifierInfo()->isImportKeyword()) {
1397 if (NextTok->isOneOf(Ks: tok::identifier, Ks: tok::less, Ks: tok::colon,
1398 Ks: tok::header_name)) {
1399 Result.setKind(tok::kw_import);
1400 ModuleImportLoc = Result.getLocation();
1401 IsAtImport = false;
1402 return true;
1403 }
1404 }
1405
1406 if (Result.getIdentifierInfo()->isModuleKeyword() &&
1407 NextTok->isOneOf(Ks: tok::identifier, Ks: tok::colon, Ks: tok::semi)) {
1408 Result.setKind(tok::kw_module);
1409 ModuleDeclLoc = Result.getLocation();
1410 return true;
1411 }
1412
1413 // Ok, it's an identifier.
1414 return false;
1415}
1416
1417bool Preprocessor::CollectPPImportSuffixAndEnterStream(
1418 SmallVectorImpl<Token> &Toks, bool StopUntilEOD) {
1419 CollectPPImportSuffix(Toks);
1420 EnterModuleSuffixTokenStream(Toks);
1421 return false;
1422}
1423
1424/// Collect the tokens of a C++20 pp-import-suffix.
1425void Preprocessor::CollectPPImportSuffix(SmallVectorImpl<Token> &Toks,
1426 bool StopUntilEOD) {
1427 while (true) {
1428 Toks.emplace_back();
1429 Lex(Result&: Toks.back());
1430
1431 switch (Toks.back().getKind()) {
1432 case tok::semi:
1433 if (!StopUntilEOD)
1434 return;
1435 [[fallthrough]];
1436 case tok::eod:
1437 case tok::eof:
1438 return;
1439 default:
1440 break;
1441 }
1442 }
1443}
1444
1445// Allocate a holding buffer for a sequence of tokens and introduce it into
1446// the token stream.
1447void Preprocessor::EnterModuleSuffixTokenStream(ArrayRef<Token> Toks) {
1448 if (Toks.empty())
1449 return;
1450 auto ToksCopy = std::make_unique<Token[]>(num: Toks.size());
1451 std::copy(first: Toks.begin(), last: Toks.end(), result: ToksCopy.get());
1452 EnterTokenStream(Toks: std::move(ToksCopy), NumToks: Toks.size(),
1453 /*DisableMacroExpansion*/ false, /*IsReinject*/ false);
1454 assert(CurTokenLexer && "Must have a TokenLexer");
1455 CurTokenLexer->setLexingCXXModuleDirective();
1456}
1457
1458/// Lex a token following the 'import' contextual keyword.
1459///
1460/// pp-import: [C++20]
1461/// import header-name pp-import-suffix[opt] ;
1462/// import header-name-tokens pp-import-suffix[opt] ;
1463/// [ObjC] @ import module-name ;
1464/// [Clang] import module-name ;
1465///
1466/// header-name-tokens:
1467/// string-literal
1468/// < [any sequence of preprocessing-tokens other than >] >
1469///
1470/// module-name:
1471/// module-name-qualifier[opt] identifier
1472///
1473/// module-name-qualifier
1474/// module-name-qualifier[opt] identifier .
1475///
1476/// We respond to a pp-import by importing macros from the named module.
1477bool Preprocessor::LexAfterModuleImport(Token &Result) {
1478 // Figure out what kind of lexer we actually have.
1479 recomputeCurLexerKind();
1480
1481 SmallVector<Token, 32> Suffix;
1482 SmallVector<IdentifierLoc, 3> Path;
1483 Lex(Result);
1484 if (LexModuleNameContinue(Tok&: Result, UseLoc: ModuleImportLoc, Suffix, Path,
1485 /*AllowMacroExpansion=*/true,
1486 /*IsPartition=*/false))
1487 return CollectPPImportSuffixAndEnterStream(Toks&: Suffix);
1488
1489 ModuleNameLoc *NameLoc = ModuleNameLoc::Create(PP&: *this, Path);
1490 Suffix.clear();
1491 Suffix.emplace_back();
1492 Suffix.back().setKind(tok::annot_module_name);
1493 Suffix.back().setAnnotationRange(NameLoc->getRange());
1494 Suffix.back().setAnnotationValue(static_cast<void *>(NameLoc));
1495 Suffix.push_back(Elt: Result);
1496
1497 // Consume the pp-import-suffix and expand any macros in it now, if we're not
1498 // at the semicolon already.
1499 SourceLocation SemiLoc = Result.getLocation();
1500 if (Suffix.back().isNot(K: tok::semi)) {
1501 if (Suffix.back().isNot(K: tok::eof))
1502 CollectPPImportSuffix(Toks&: Suffix);
1503 if (Suffix.back().isNot(K: tok::semi)) {
1504 // This is not an import after all.
1505 EnterModuleSuffixTokenStream(Toks: Suffix);
1506 return false;
1507 }
1508 SemiLoc = Suffix.back().getLocation();
1509 }
1510
1511 Module *Imported = nullptr;
1512 if (getLangOpts().Modules) {
1513 Imported = TheModuleLoader.loadModule(ImportLoc: ModuleImportLoc, Path, Visibility: Module::Hidden,
1514 /*IsInclusionDirective=*/false);
1515 if (Imported)
1516 makeModuleVisible(M: Imported, Loc: SemiLoc);
1517 }
1518
1519 if (Callbacks)
1520 Callbacks->moduleImport(ImportLoc: ModuleImportLoc, Path, Imported);
1521
1522 if (!Suffix.empty()) {
1523 EnterModuleSuffixTokenStream(Toks: Suffix);
1524 return false;
1525 }
1526 return true;
1527}
1528
1529void Preprocessor::makeModuleVisible(Module *M, SourceLocation Loc,
1530 bool IncludeExports) {
1531 CurSubmoduleState->VisibleModules.setVisible(
1532 M, Loc, IncludeExports, Vis: [](Module *) {},
1533 Cb: [&](ArrayRef<Module *> Path, Module *Conflict, StringRef Message) {
1534 // FIXME: Include the path in the diagnostic.
1535 // FIXME: Include the import location for the conflicting module.
1536 Diag(Loc: ModuleImportLoc, DiagID: diag::warn_module_conflict)
1537 << Path[0]->getFullModuleName()
1538 << Conflict->getFullModuleName()
1539 << Message;
1540 });
1541
1542 // Add this module to the imports list of the currently-built submodule.
1543 if (!BuildingSubmoduleStack.empty() && M != BuildingSubmoduleStack.back().M)
1544 BuildingSubmoduleStack.back().M->Imports.insert(X: M);
1545}
1546
1547bool Preprocessor::FinishLexStringLiteral(Token &Result, std::string &String,
1548 const char *DiagnosticTag,
1549 bool AllowMacroExpansion) {
1550 // We need at least one string literal.
1551 if (Result.isNot(K: tok::string_literal)) {
1552 Diag(Tok: Result, DiagID: diag::err_expected_string_literal)
1553 << /*Source='in...'*/0 << DiagnosticTag;
1554 return false;
1555 }
1556
1557 // Lex string literal tokens, optionally with macro expansion.
1558 SmallVector<Token, 4> StrToks;
1559 do {
1560 StrToks.push_back(Elt: Result);
1561
1562 if (Result.hasUDSuffix())
1563 Diag(Tok: Result, DiagID: diag::err_invalid_string_udl);
1564
1565 if (AllowMacroExpansion)
1566 Lex(Result);
1567 else
1568 LexUnexpandedToken(Result);
1569 } while (Result.is(K: tok::string_literal));
1570
1571 // Concatenate and parse the strings.
1572 StringLiteralParser Literal(StrToks, *this);
1573 assert(Literal.isOrdinary() && "Didn't allow wide strings in");
1574
1575 if (Literal.hadError)
1576 return false;
1577
1578 if (Literal.Pascal) {
1579 Diag(Loc: StrToks[0].getLocation(), DiagID: diag::err_expected_string_literal)
1580 << /*Source='in...'*/0 << DiagnosticTag;
1581 return false;
1582 }
1583
1584 String = std::string(Literal.GetString());
1585 return true;
1586}
1587
1588bool Preprocessor::parseSimpleIntegerLiteral(Token &Tok, uint64_t &Value) {
1589 assert(Tok.is(tok::numeric_constant));
1590 SmallString<8> IntegerBuffer;
1591 bool NumberInvalid = false;
1592 StringRef Spelling = getSpelling(Tok, Buffer&: IntegerBuffer, Invalid: &NumberInvalid);
1593 if (NumberInvalid)
1594 return false;
1595 NumericLiteralParser Literal(Spelling, Tok.getLocation(), getSourceManager(),
1596 getLangOpts(), getTargetInfo(),
1597 getDiagnostics());
1598 if (Literal.hadError || !Literal.isIntegerLiteral() || Literal.hasUDSuffix())
1599 return false;
1600 llvm::APInt APVal(64, 0);
1601 if (Literal.GetIntegerValue(Val&: APVal))
1602 return false;
1603 Lex(Result&: Tok);
1604 Value = APVal.getLimitedValue();
1605 return true;
1606}
1607
1608void Preprocessor::addCommentHandler(CommentHandler *Handler) {
1609 assert(Handler && "NULL comment handler");
1610 assert(!llvm::is_contained(CommentHandlers, Handler) &&
1611 "Comment handler already registered");
1612 CommentHandlers.push_back(x: Handler);
1613}
1614
1615void Preprocessor::removeCommentHandler(CommentHandler *Handler) {
1616 std::vector<CommentHandler *>::iterator Pos =
1617 llvm::find(Range&: CommentHandlers, Val: Handler);
1618 assert(Pos != CommentHandlers.end() && "Comment handler not registered");
1619 CommentHandlers.erase(position: Pos);
1620}
1621
1622bool Preprocessor::HandleComment(Token &result, SourceRange Comment) {
1623 bool AnyPendingTokens = false;
1624 for (CommentHandler *H : CommentHandlers) {
1625 if (H->HandleComment(PP&: *this, Comment))
1626 AnyPendingTokens = true;
1627 }
1628 if (!AnyPendingTokens || getCommentRetentionState())
1629 return false;
1630 Lex(Result&: result);
1631 return true;
1632}
1633
1634void Preprocessor::emitMacroDeprecationWarning(const Token &Identifier) const {
1635 const MacroAnnotations &A =
1636 getMacroAnnotations(II: Identifier.getIdentifierInfo());
1637 assert(A.DeprecationInfo &&
1638 "Macro deprecation warning without recorded annotation!");
1639 const MacroAnnotationInfo &Info = *A.DeprecationInfo;
1640 if (Info.Message.empty())
1641 Diag(Tok: Identifier, DiagID: diag::warn_pragma_deprecated_macro_use)
1642 << Identifier.getIdentifierInfo() << 0;
1643 else
1644 Diag(Tok: Identifier, DiagID: diag::warn_pragma_deprecated_macro_use)
1645 << Identifier.getIdentifierInfo() << 1 << Info.Message;
1646 Diag(Loc: Info.Location, DiagID: diag::note_pp_macro_annotation) << 0;
1647}
1648
1649void Preprocessor::emitRestrictExpansionWarning(const Token &Identifier) const {
1650 const MacroAnnotations &A =
1651 getMacroAnnotations(II: Identifier.getIdentifierInfo());
1652 assert(A.RestrictExpansionInfo &&
1653 "Macro restricted expansion warning without recorded annotation!");
1654 const MacroAnnotationInfo &Info = *A.RestrictExpansionInfo;
1655 if (Info.Message.empty())
1656 Diag(Tok: Identifier, DiagID: diag::warn_pragma_restrict_expansion_macro_use)
1657 << Identifier.getIdentifierInfo() << 0;
1658 else
1659 Diag(Tok: Identifier, DiagID: diag::warn_pragma_restrict_expansion_macro_use)
1660 << Identifier.getIdentifierInfo() << 1 << Info.Message;
1661 Diag(Loc: Info.Location, DiagID: diag::note_pp_macro_annotation) << 1;
1662}
1663
1664void Preprocessor::emitRestrictInfNaNWarning(const Token &Identifier,
1665 unsigned DiagSelection) const {
1666 Diag(Tok: Identifier, DiagID: diag::warn_fp_nan_inf_when_disabled) << DiagSelection << 1;
1667}
1668
1669void Preprocessor::emitFinalMacroWarning(const Token &Identifier,
1670 bool IsUndef) const {
1671 const MacroAnnotations &A =
1672 getMacroAnnotations(II: Identifier.getIdentifierInfo());
1673 assert(A.FinalAnnotationLoc &&
1674 "Final macro warning without recorded annotation!");
1675
1676 Diag(Tok: Identifier, DiagID: diag::warn_pragma_final_macro)
1677 << Identifier.getIdentifierInfo() << (IsUndef ? 0 : 1);
1678 Diag(Loc: *A.FinalAnnotationLoc, DiagID: diag::note_pp_macro_annotation) << 2;
1679}
1680
1681bool Preprocessor::isSafeBufferOptOut(const SourceManager &SourceMgr,
1682 const SourceLocation &Loc) const {
1683 // The lambda that tests if a `Loc` is in an opt-out region given one opt-out
1684 // region map:
1685 auto TestInMap = [&SourceMgr](const SafeBufferOptOutRegionsTy &Map,
1686 const SourceLocation &Loc) -> bool {
1687 // Try to find a region in `SafeBufferOptOutMap` where `Loc` is in:
1688 auto FirstRegionEndingAfterLoc = llvm::partition_point(
1689 Range: Map, P: [&SourceMgr,
1690 &Loc](const std::pair<SourceLocation, SourceLocation> &Region) {
1691 return SourceMgr.isBeforeInTranslationUnit(LHS: Region.second, RHS: Loc);
1692 });
1693
1694 if (FirstRegionEndingAfterLoc != Map.end()) {
1695 // To test if the start location of the found region precedes `Loc`:
1696 return SourceMgr.isBeforeInTranslationUnit(
1697 LHS: FirstRegionEndingAfterLoc->first, RHS: Loc);
1698 }
1699 // If we do not find a region whose end location passes `Loc`, we want to
1700 // check if the current region is still open:
1701 if (!Map.empty() && Map.back().first == Map.back().second)
1702 return SourceMgr.isBeforeInTranslationUnit(LHS: Map.back().first, RHS: Loc);
1703 return false;
1704 };
1705
1706 // What the following does:
1707 //
1708 // If `Loc` belongs to the local TU, we just look up `SafeBufferOptOutMap`.
1709 // Otherwise, `Loc` is from a loaded AST. We look up the
1710 // `LoadedSafeBufferOptOutMap` first to get the opt-out region map of the
1711 // loaded AST where `Loc` is at. Then we find if `Loc` is in an opt-out
1712 // region w.r.t. the region map. If the region map is absent, it means there
1713 // is no opt-out pragma in that loaded AST.
1714 //
1715 // Opt-out pragmas in the local TU or a loaded AST is not visible to another
1716 // one of them. That means if you put the pragmas around a `#include
1717 // "module.h"`, where module.h is a module, it is not actually suppressing
1718 // warnings in module.h. This is fine because warnings in module.h will be
1719 // reported when module.h is compiled in isolation and nothing in module.h
1720 // will be analyzed ever again. So you will not see warnings from the file
1721 // that imports module.h anyway. And you can't even do the same thing for PCHs
1722 // because they can only be included from the command line.
1723
1724 if (SourceMgr.isLocalSourceLocation(Loc))
1725 return TestInMap(SafeBufferOptOutMap, Loc);
1726
1727 const SafeBufferOptOutRegionsTy *LoadedRegions =
1728 LoadedSafeBufferOptOutMap.lookupLoadedOptOutMap(Loc, SrcMgr: SourceMgr);
1729
1730 if (LoadedRegions)
1731 return TestInMap(*LoadedRegions, Loc);
1732 return false;
1733}
1734
1735bool Preprocessor::enterOrExitSafeBufferOptOutRegion(
1736 bool isEnter, const SourceLocation &Loc) {
1737 if (isEnter) {
1738 if (isPPInSafeBufferOptOutRegion())
1739 return true; // invalid enter action
1740 InSafeBufferOptOutRegion = true;
1741 CurrentSafeBufferOptOutStart = Loc;
1742
1743 // To set the start location of a new region:
1744
1745 if (!SafeBufferOptOutMap.empty()) {
1746 [[maybe_unused]] auto *PrevRegion = &SafeBufferOptOutMap.back();
1747 assert(PrevRegion->first != PrevRegion->second &&
1748 "Shall not begin a safe buffer opt-out region before closing the "
1749 "previous one.");
1750 }
1751 // If the start location equals to the end location, we call the region a
1752 // open region or a unclosed region (i.e., end location has not been set
1753 // yet).
1754 SafeBufferOptOutMap.emplace_back(Args: Loc, Args: Loc);
1755 } else {
1756 if (!isPPInSafeBufferOptOutRegion())
1757 return true; // invalid enter action
1758 InSafeBufferOptOutRegion = false;
1759
1760 // To set the end location of the current open region:
1761
1762 assert(!SafeBufferOptOutMap.empty() &&
1763 "Misordered safe buffer opt-out regions");
1764 auto *CurrRegion = &SafeBufferOptOutMap.back();
1765 assert(CurrRegion->first == CurrRegion->second &&
1766 "Set end location to a closed safe buffer opt-out region");
1767 CurrRegion->second = Loc;
1768 }
1769 return false;
1770}
1771
1772bool Preprocessor::isPPInSafeBufferOptOutRegion() {
1773 return InSafeBufferOptOutRegion;
1774}
1775bool Preprocessor::isPPInSafeBufferOptOutRegion(SourceLocation &StartLoc) {
1776 StartLoc = CurrentSafeBufferOptOutStart;
1777 return InSafeBufferOptOutRegion;
1778}
1779
1780SmallVector<SourceLocation, 64>
1781Preprocessor::serializeSafeBufferOptOutMap() const {
1782 assert(!InSafeBufferOptOutRegion &&
1783 "Attempt to serialize safe buffer opt-out regions before file being "
1784 "completely preprocessed");
1785
1786 SmallVector<SourceLocation, 64> SrcSeq;
1787
1788 for (const auto &[begin, end] : SafeBufferOptOutMap) {
1789 SrcSeq.push_back(Elt: begin);
1790 SrcSeq.push_back(Elt: end);
1791 }
1792 // Only `SafeBufferOptOutMap` gets serialized. No need to serialize
1793 // `LoadedSafeBufferOptOutMap` because if this TU loads a pch/module, every
1794 // pch/module in the pch-chain/module-DAG will be loaded one by one in order.
1795 // It means that for each loading pch/module m, it just needs to load m's own
1796 // `SafeBufferOptOutMap`.
1797 return SrcSeq;
1798}
1799
1800bool Preprocessor::setDeserializedSafeBufferOptOutMap(
1801 const SmallVectorImpl<SourceLocation> &SourceLocations) {
1802 if (SourceLocations.size() == 0)
1803 return false;
1804
1805 assert(SourceLocations.size() % 2 == 0 &&
1806 "ill-formed SourceLocation sequence");
1807
1808 auto It = SourceLocations.begin();
1809 SafeBufferOptOutRegionsTy &Regions =
1810 LoadedSafeBufferOptOutMap.findAndConsLoadedOptOutMap(Loc: *It, SrcMgr&: SourceMgr);
1811
1812 do {
1813 SourceLocation Begin = *It++;
1814 SourceLocation End = *It++;
1815
1816 Regions.emplace_back(Args&: Begin, Args&: End);
1817 } while (It != SourceLocations.end());
1818 return true;
1819}
1820
1821ModuleLoader::~ModuleLoader() = default;
1822
1823CommentHandler::~CommentHandler() = default;
1824
1825EmptylineHandler::~EmptylineHandler() = default;
1826
1827CodeCompletionHandler::~CodeCompletionHandler() = default;
1828
1829void Preprocessor::createPreprocessingRecord() {
1830 if (Record)
1831 return;
1832
1833 Record = new PreprocessingRecord(getSourceManager());
1834 addPPCallbacks(C: std::unique_ptr<PPCallbacks>(Record));
1835}
1836
1837const char *Preprocessor::getCheckPoint(FileID FID, const char *Start) const {
1838 if (auto It = CheckPoints.find(Val: FID); It != CheckPoints.end()) {
1839 const SmallVector<const char *> &FileCheckPoints = It->second;
1840 const char *Last = nullptr;
1841 // FIXME: Do better than a linear search.
1842 for (const char *P : FileCheckPoints) {
1843 if (P > Start)
1844 break;
1845 Last = P;
1846 }
1847 return Last;
1848 }
1849
1850 return nullptr;
1851}
1852
1853bool Preprocessor::hasSeenNoTrivialPPDirective() const {
1854 return DirTracer && DirTracer->hasSeenNoTrivialPPDirective();
1855}
1856
1857bool NoTrivialPPDirectiveTracer::hasSeenNoTrivialPPDirective() const {
1858 return SeenNoTrivialPPDirective;
1859}
1860
1861void NoTrivialPPDirectiveTracer::setSeenNoTrivialPPDirective() {
1862 if (InMainFile && !SeenNoTrivialPPDirective)
1863 SeenNoTrivialPPDirective = true;
1864}
1865
1866void NoTrivialPPDirectiveTracer::LexedFileChanged(
1867 FileID FID, LexedFileChangeReason Reason,
1868 SrcMgr::CharacteristicKind FileType, FileID PrevFID, SourceLocation Loc) {
1869 InMainFile = (FID == PP.getSourceManager().getMainFileID());
1870}
1871
1872void NoTrivialPPDirectiveTracer::MacroExpands(const Token &MacroNameTok,
1873 const MacroDefinition &MD,
1874 SourceRange Range,
1875 const MacroArgs *Args) {
1876 // FIXME: Does only enable builtin macro expansion make sense?
1877 if (!MD.getMacroInfo()->isBuiltinMacro())
1878 setSeenNoTrivialPPDirective();
1879}
1880