Warning: That file was not part of the compilation database. It may have many parsing errors.

1//===- Preprocessor.h - C Language Family Preprocessor ----------*- C++ -*-===//
2//
3// The LLVM Compiler Infrastructure
4//
5// This file is distributed under the University of Illinois Open Source
6// License. See LICENSE.TXT for details.
7//
8//===----------------------------------------------------------------------===//
9//
10/// \file
11/// Defines the clang::Preprocessor interface.
12//
13//===----------------------------------------------------------------------===//
14
15#ifndef LLVM_CLANG_LEX_PREPROCESSOR_H
16#define LLVM_CLANG_LEX_PREPROCESSOR_H
17
18#include "clang/Basic/Builtins.h"
19#include "clang/Basic/Diagnostic.h"
20#include "clang/Basic/IdentifierTable.h"
21#include "clang/Basic/LLVM.h"
22#include "clang/Basic/LangOptions.h"
23#include "clang/Basic/Module.h"
24#include "clang/Basic/SourceLocation.h"
25#include "clang/Basic/SourceManager.h"
26#include "clang/Basic/TokenKinds.h"
27#include "clang/Lex/Lexer.h"
28#include "clang/Lex/MacroInfo.h"
29#include "clang/Lex/ModuleLoader.h"
30#include "clang/Lex/ModuleMap.h"
31#include "clang/Lex/PPCallbacks.h"
32#include "clang/Lex/PTHLexer.h"
33#include "clang/Lex/Token.h"
34#include "clang/Lex/TokenLexer.h"
35#include "llvm/ADT/ArrayRef.h"
36#include "llvm/ADT/DenseMap.h"
37#include "llvm/ADT/FoldingSet.h"
38#include "llvm/ADT/None.h"
39#include "llvm/ADT/Optional.h"
40#include "llvm/ADT/PointerUnion.h"
41#include "llvm/ADT/STLExtras.h"
42#include "llvm/ADT/SmallPtrSet.h"
43#include "llvm/ADT/SmallVector.h"
44#include "llvm/ADT/StringRef.h"
45#include "llvm/ADT/TinyPtrVector.h"
46#include "llvm/ADT/iterator_range.h"
47#include "llvm/Support/Allocator.h"
48#include "llvm/Support/Casting.h"
49#include "llvm/Support/Registry.h"
50#include <cassert>
51#include <cstddef>
52#include <cstdint>
53#include <memory>
54#include <map>
55#include <string>
56#include <utility>
57#include <vector>
58
59namespace llvm {
60
61template<unsigned InternalLen> class SmallString;
62
63} // namespace llvm
64
65namespace clang {
66
67class CodeCompletionHandler;
68class CommentHandler;
69class DirectoryEntry;
70class DirectoryLookup;
71class ExternalPreprocessorSource;
72class FileEntry;
73class FileManager;
74class HeaderSearch;
75class MacroArgs;
76class MemoryBufferCache;
77class PragmaHandler;
78class PragmaNamespace;
79class PreprocessingRecord;
80class PreprocessorLexer;
81class PreprocessorOptions;
82class PTHManager;
83class ScratchBuffer;
84class TargetInfo;
85
86/// Stores token information for comparing actual tokens with
87/// predefined values. Only handles simple tokens and identifiers.
88class TokenValue {
89 tok::TokenKind Kind;
90 IdentifierInfo *II;
91
92public:
93 TokenValue(tok::TokenKind Kind) : Kind(Kind), II(nullptr) {
94 assert(Kind != tok::raw_identifier && "Raw identifiers are not supported.");
95 assert(Kind != tok::identifier &&
96 "Identifiers should be created by TokenValue(IdentifierInfo *)");
97 assert(!tok::isLiteral(Kind) && "Literals are not supported.");
98 assert(!tok::isAnnotation(Kind) && "Annotations are not supported.");
99 }
100
101 TokenValue(IdentifierInfo *II) : Kind(tok::identifier), II(II) {}
102
103 bool operator==(const Token &Tok) const {
104 return Tok.getKind() == Kind &&
105 (!II || II == Tok.getIdentifierInfo());
106 }
107};
108
109/// Context in which macro name is used.
110enum MacroUse {
111 // other than #define or #undef
112 MU_Other = 0,
113
114 // macro name specified in #define
115 MU_Define = 1,
116
117 // macro name specified in #undef
118 MU_Undef = 2
119};
120
121/// Engages in a tight little dance with the lexer to efficiently
122/// preprocess tokens.
123///
124/// Lexers know only about tokens within a single source file, and don't
125/// know anything about preprocessor-level issues like the \#include stack,
126/// token expansion, etc.
127class Preprocessor {
128 friend class VAOptDefinitionContext;
129 friend class VariadicMacroScopeGuard;
130
131 std::shared_ptr<PreprocessorOptions> PPOpts;
132 DiagnosticsEngine *Diags;
133 LangOptions &LangOpts;
134 const TargetInfo *Target = nullptr;
135 const TargetInfo *AuxTarget = nullptr;
136 FileManager &FileMgr;
137 SourceManager &SourceMgr;
138 MemoryBufferCache &PCMCache;
139 std::unique_ptr<ScratchBuffer> ScratchBuf;
140 HeaderSearch &HeaderInfo;
141 ModuleLoader &TheModuleLoader;
142
143 /// External source of macros.
144 ExternalPreprocessorSource *ExternalSource;
145
146 /// An optional PTHManager object used for getting tokens from
147 /// a token cache rather than lexing the original source file.
148 std::unique_ptr<PTHManager> PTH;
149
150 /// A BumpPtrAllocator object used to quickly allocate and release
151 /// objects internal to the Preprocessor.
152 llvm::BumpPtrAllocator BP;
153
154 /// Identifiers for builtin macros and other builtins.
155 IdentifierInfo *Ident__LINE__, *Ident__FILE__; // __LINE__, __FILE__
156 IdentifierInfo *Ident__DATE__, *Ident__TIME__; // __DATE__, __TIME__
157 IdentifierInfo *Ident__INCLUDE_LEVEL__; // __INCLUDE_LEVEL__
158 IdentifierInfo *Ident__BASE_FILE__; // __BASE_FILE__
159 IdentifierInfo *Ident__TIMESTAMP__; // __TIMESTAMP__
160 IdentifierInfo *Ident__COUNTER__; // __COUNTER__
161 IdentifierInfo *Ident_Pragma, *Ident__pragma; // _Pragma, __pragma
162 IdentifierInfo *Ident__identifier; // __identifier
163 IdentifierInfo *Ident__VA_ARGS__; // __VA_ARGS__
164 IdentifierInfo *Ident__VA_OPT__; // __VA_OPT__
165 IdentifierInfo *Ident__has_feature; // __has_feature
166 IdentifierInfo *Ident__has_extension; // __has_extension
167 IdentifierInfo *Ident__has_builtin; // __has_builtin
168 IdentifierInfo *Ident__has_attribute; // __has_attribute
169 IdentifierInfo *Ident__has_include; // __has_include
170 IdentifierInfo *Ident__has_include_next; // __has_include_next
171 IdentifierInfo *Ident__has_warning; // __has_warning
172 IdentifierInfo *Ident__is_identifier; // __is_identifier
173 IdentifierInfo *Ident__building_module; // __building_module
174 IdentifierInfo *Ident__MODULE__; // __MODULE__
175 IdentifierInfo *Ident__has_cpp_attribute; // __has_cpp_attribute
176 IdentifierInfo *Ident__has_c_attribute; // __has_c_attribute
177 IdentifierInfo *Ident__has_declspec; // __has_declspec_attribute
178 IdentifierInfo *Ident__is_target_arch; // __is_target_arch
179 IdentifierInfo *Ident__is_target_vendor; // __is_target_vendor
180 IdentifierInfo *Ident__is_target_os; // __is_target_os
181 IdentifierInfo *Ident__is_target_environment; // __is_target_environment
182
183 SourceLocation DATELoc, TIMELoc;
184
185 // Next __COUNTER__ value, starts at 0.
186 unsigned CounterValue = 0;
187
188 enum {
189 /// Maximum depth of \#includes.
190 MaxAllowedIncludeStackDepth = 200
191 };
192
193 // State that is set before the preprocessor begins.
194 bool KeepComments : 1;
195 bool KeepMacroComments : 1;
196 bool SuppressIncludeNotFoundError : 1;
197
198 // State that changes while the preprocessor runs:
199 bool InMacroArgs : 1; // True if parsing fn macro invocation args.
200
201 /// Whether the preprocessor owns the header search object.
202 bool OwnsHeaderSearch : 1;
203
204 /// True if macro expansion is disabled.
205 bool DisableMacroExpansion : 1;
206
207 /// Temporarily disables DisableMacroExpansion (i.e. enables expansion)
208 /// when parsing preprocessor directives.
209 bool MacroExpansionInDirectivesOverride : 1;
210
211 class ResetMacroExpansionHelper;
212
213 /// Whether we have already loaded macros from the external source.
214 mutable bool ReadMacrosFromExternalSource : 1;
215
216 /// True if pragmas are enabled.
217 bool PragmasEnabled : 1;
218
219 /// True if the current build action is a preprocessing action.
220 bool PreprocessedOutput : 1;
221
222 /// True if we are currently preprocessing a #if or #elif directive
223 bool ParsingIfOrElifDirective;
224
225 /// True if we are pre-expanding macro arguments.
226 bool InMacroArgPreExpansion;
227
228 /// Mapping/lookup information for all identifiers in
229 /// the program, including program keywords.
230 mutable IdentifierTable Identifiers;
231
232 /// This table contains all the selectors in the program.
233 ///
234 /// Unlike IdentifierTable above, this table *isn't* populated by the
235 /// preprocessor. It is declared/expanded here because its role/lifetime is
236 /// conceptually similar to the IdentifierTable. In addition, the current
237 /// control flow (in clang::ParseAST()), make it convenient to put here.
238 ///
239 /// FIXME: Make sure the lifetime of Identifiers/Selectors *isn't* tied to
240 /// the lifetime of the preprocessor.
241 SelectorTable Selectors;
242
243 /// Information about builtins.
244 Builtin::Context BuiltinInfo;
245
246 /// Tracks all of the pragmas that the client registered
247 /// with this preprocessor.
248 std::unique_ptr<PragmaNamespace> PragmaHandlers;
249
250 /// Pragma handlers of the original source is stored here during the
251 /// parsing of a model file.
252 std::unique_ptr<PragmaNamespace> PragmaHandlersBackup;
253
254 /// Tracks all of the comment handlers that the client registered
255 /// with this preprocessor.
256 std::vector<CommentHandler *> CommentHandlers;
257
258 /// True if we want to ignore EOF token and continue later on (thus
259 /// avoid tearing the Lexer and etc. down).
260 bool IncrementalProcessing = false;
261
262 /// The kind of translation unit we are processing.
263 TranslationUnitKind TUKind;
264
265 /// The code-completion handler.
266 CodeCompletionHandler *CodeComplete = nullptr;
267
268 /// The file that we're performing code-completion for, if any.
269 const FileEntry *CodeCompletionFile = nullptr;
270
271 /// The offset in file for the code-completion point.
272 unsigned CodeCompletionOffset = 0;
273
274 /// The location for the code-completion point. This gets instantiated
275 /// when the CodeCompletionFile gets \#include'ed for preprocessing.
276 SourceLocation CodeCompletionLoc;
277
278 /// The start location for the file of the code-completion point.
279 ///
280 /// This gets instantiated when the CodeCompletionFile gets \#include'ed
281 /// for preprocessing.
282 SourceLocation CodeCompletionFileLoc;
283
284 /// The source location of the \c import contextual keyword we just
285 /// lexed, if any.
286 SourceLocation ModuleImportLoc;
287
288 /// The module import path that we're currently processing.
289 SmallVector<std::pair<IdentifierInfo *, SourceLocation>, 2> ModuleImportPath;
290
291 /// Whether the last token we lexed was an '@'.
292 bool LastTokenWasAt = false;
293
294 /// Whether the module import expects an identifier next. Otherwise,
295 /// it expects a '.' or ';'.
296 bool ModuleImportExpectsIdentifier = false;
297
298 /// The source location of the currently-active
299 /// \#pragma clang arc_cf_code_audited begin.
300 SourceLocation PragmaARCCFCodeAuditedLoc;
301
302 /// The source location of the currently-active
303 /// \#pragma clang assume_nonnull begin.
304 SourceLocation PragmaAssumeNonNullLoc;
305
306 /// True if we hit the code-completion point.
307 bool CodeCompletionReached = false;
308
309 /// The code completion token containing the information
310 /// on the stem that is to be code completed.
311 IdentifierInfo *CodeCompletionII = nullptr;
312
313 /// The directory that the main file should be considered to occupy,
314 /// if it does not correspond to a real file (as happens when building a
315 /// module).
316 const DirectoryEntry *MainFileDir = nullptr;
317
318 /// The number of bytes that we will initially skip when entering the
319 /// main file, along with a flag that indicates whether skipping this number
320 /// of bytes will place the lexer at the start of a line.
321 ///
322 /// This is used when loading a precompiled preamble.
323 std::pair<int, bool> SkipMainFilePreamble;
324
325public:
326 struct PreambleSkipInfo {
327 SourceLocation HashTokenLoc;
328 SourceLocation IfTokenLoc;
329 bool FoundNonSkipPortion;
330 bool FoundElse;
331 SourceLocation ElseLoc;
332
333 PreambleSkipInfo(SourceLocation HashTokenLoc, SourceLocation IfTokenLoc,
334 bool FoundNonSkipPortion, bool FoundElse,
335 SourceLocation ElseLoc)
336 : HashTokenLoc(HashTokenLoc), IfTokenLoc(IfTokenLoc),
337 FoundNonSkipPortion(FoundNonSkipPortion), FoundElse(FoundElse),
338 ElseLoc(ElseLoc) {}
339 };
340
341private:
342 friend class ASTReader;
343 friend class MacroArgs;
344
345 class PreambleConditionalStackStore {
346 enum State {
347 Off = 0,
348 Recording = 1,
349 Replaying = 2,
350 };
351
352 public:
353 PreambleConditionalStackStore() = default;
354
355 void startRecording() { ConditionalStackState = Recording; }
356 void startReplaying() { ConditionalStackState = Replaying; }
357 bool isRecording() const { return ConditionalStackState == Recording; }
358 bool isReplaying() const { return ConditionalStackState == Replaying; }
359
360 ArrayRef<PPConditionalInfo> getStack() const {
361 return ConditionalStack;
362 }
363
364 void doneReplaying() {
365 ConditionalStack.clear();
366 ConditionalStackState = Off;
367 }
368
369 void setStack(ArrayRef<PPConditionalInfo> s) {
370 if (!isRecording() && !isReplaying())
371 return;
372 ConditionalStack.clear();
373 ConditionalStack.append(s.begin(), s.end());
374 }
375
376 bool hasRecordedPreamble() const { return !ConditionalStack.empty(); }
377
378 bool reachedEOFWhileSkipping() const { return SkipInfo.hasValue(); }
379
380 void clearSkipInfo() { SkipInfo.reset(); }
381
382 llvm::Optional<PreambleSkipInfo> SkipInfo;
383
384 private:
385 SmallVector<PPConditionalInfo, 4> ConditionalStack;
386 State ConditionalStackState = Off;
387 } PreambleConditionalStack;
388
389 /// The current top of the stack that we're lexing from if
390 /// not expanding a macro and we are lexing directly from source code.
391 ///
392 /// Only one of CurLexer, CurPTHLexer, or CurTokenLexer will be non-null.
393 std::unique_ptr<Lexer> CurLexer;
394
395 /// The current top of stack that we're lexing from if
396 /// not expanding from a macro and we are lexing from a PTH cache.
397 ///
398 /// Only one of CurLexer, CurPTHLexer, or CurTokenLexer will be non-null.
399 std::unique_ptr<PTHLexer> CurPTHLexer;
400
401 /// The current top of the stack what we're lexing from
402 /// if not expanding a macro.
403 ///
404 /// This is an alias for either CurLexer or CurPTHLexer.
405 PreprocessorLexer *CurPPLexer = nullptr;
406
407 /// Used to find the current FileEntry, if CurLexer is non-null
408 /// and if applicable.
409 ///
410 /// This allows us to implement \#include_next and find directory-specific
411 /// properties.
412 const DirectoryLookup *CurDirLookup = nullptr;
413
414 /// The current macro we are expanding, if we are expanding a macro.
415 ///
416 /// One of CurLexer and CurTokenLexer must be null.
417 std::unique_ptr<TokenLexer> CurTokenLexer;
418
419 /// The kind of lexer we're currently working with.
420 enum CurLexerKind {
421 CLK_Lexer,
422 CLK_PTHLexer,
423 CLK_TokenLexer,
424 CLK_CachingLexer,
425 CLK_LexAfterModuleImport
426 } CurLexerKind = CLK_Lexer;
427
428 /// If the current lexer is for a submodule that is being built, this
429 /// is that submodule.
430 Module *CurLexerSubmodule = nullptr;
431
432 /// Keeps track of the stack of files currently
433 /// \#included, and macros currently being expanded from, not counting
434 /// CurLexer/CurTokenLexer.
435 struct IncludeStackInfo {
436 enum CurLexerKind CurLexerKind;
437 Module *TheSubmodule;
438 std::unique_ptr<Lexer> TheLexer;
439 std::unique_ptr<PTHLexer> ThePTHLexer;
440 PreprocessorLexer *ThePPLexer;
441 std::unique_ptr<TokenLexer> TheTokenLexer;
442 const DirectoryLookup *TheDirLookup;
443
444 // The following constructors are completely useless copies of the default
445 // versions, only needed to pacify MSVC.
446 IncludeStackInfo(enum CurLexerKind CurLexerKind, Module *TheSubmodule,
447 std::unique_ptr<Lexer> &&TheLexer,
448 std::unique_ptr<PTHLexer> &&ThePTHLexer,
449 PreprocessorLexer *ThePPLexer,
450 std::unique_ptr<TokenLexer> &&TheTokenLexer,
451 const DirectoryLookup *TheDirLookup)
452 : CurLexerKind(std::move(CurLexerKind)),
453 TheSubmodule(std::move(TheSubmodule)), TheLexer(std::move(TheLexer)),
454 ThePTHLexer(std::move(ThePTHLexer)),
455 ThePPLexer(std::move(ThePPLexer)),
456 TheTokenLexer(std::move(TheTokenLexer)),
457 TheDirLookup(std::move(TheDirLookup)) {}
458 };
459 std::vector<IncludeStackInfo> IncludeMacroStack;
460
461 /// Actions invoked when some preprocessor activity is
462 /// encountered (e.g. a file is \#included, etc).
463 std::unique_ptr<PPCallbacks> Callbacks;
464
465 struct MacroExpandsInfo {
466 Token Tok;
467 MacroDefinition MD;
468 SourceRange Range;
469
470 MacroExpandsInfo(Token Tok, MacroDefinition MD, SourceRange Range)
471 : Tok(Tok), MD(MD), Range(Range) {}
472 };
473 SmallVector<MacroExpandsInfo, 2> DelayedMacroExpandsCallbacks;
474
475 /// Information about a name that has been used to define a module macro.
476 struct ModuleMacroInfo {
477 /// The most recent macro directive for this identifier.
478 MacroDirective *MD;
479
480 /// The active module macros for this identifier.
481 llvm::TinyPtrVector<ModuleMacro *> ActiveModuleMacros;
482
483 /// The generation number at which we last updated ActiveModuleMacros.
484 /// \see Preprocessor::VisibleModules.
485 unsigned ActiveModuleMacrosGeneration = 0;
486
487 /// Whether this macro name is ambiguous.
488 bool IsAmbiguous = false;
489
490 /// The module macros that are overridden by this macro.
491 llvm::TinyPtrVector<ModuleMacro *> OverriddenMacros;
492
493 ModuleMacroInfo(MacroDirective *MD) : MD(MD) {}
494 };
495
496 /// The state of a macro for an identifier.
497 class MacroState {
498 mutable llvm::PointerUnion<MacroDirective *, ModuleMacroInfo *> State;
499
500 ModuleMacroInfo *getModuleInfo(Preprocessor &PP,
501 const IdentifierInfo *II) const {
502 if (II->isOutOfDate())
503 PP.updateOutOfDateIdentifier(const_cast<IdentifierInfo&>(*II));
504 // FIXME: Find a spare bit on IdentifierInfo and store a
505 // HasModuleMacros flag.
506 if (!II->hasMacroDefinition() ||
507 (!PP.getLangOpts().Modules &&
508 !PP.getLangOpts().ModulesLocalVisibility) ||
509 !PP.CurSubmoduleState->VisibleModules.getGeneration())
510 return nullptr;
511
512 auto *Info = State.dyn_cast<ModuleMacroInfo*>();
513 if (!Info) {
514 Info = new (PP.getPreprocessorAllocator())
515 ModuleMacroInfo(State.get<MacroDirective *>());
516 State = Info;
517 }
518
519 if (PP.CurSubmoduleState->VisibleModules.getGeneration() !=
520 Info->ActiveModuleMacrosGeneration)
521 PP.updateModuleMacroInfo(II, *Info);
522 return Info;
523 }
524
525 public:
526 MacroState() : MacroState(nullptr) {}
527 MacroState(MacroDirective *MD) : State(MD) {}
528
529 MacroState(MacroState &&O) noexcept : State(O.State) {
530 O.State = (MacroDirective *)nullptr;
531 }
532
533 MacroState &operator=(MacroState &&O) noexcept {
534 auto S = O.State;
535 O.State = (MacroDirective *)nullptr;
536 State = S;
537 return *this;
538 }
539
540 ~MacroState() {
541 if (auto *Info = State.dyn_cast<ModuleMacroInfo*>())
542 Info->~ModuleMacroInfo();
543 }
544
545 MacroDirective *getLatest() const {
546 if (auto *Info = State.dyn_cast<ModuleMacroInfo*>())
547 return Info->MD;
548 return State.get<MacroDirective*>();
549 }
550
551 void setLatest(MacroDirective *MD) {
552 if (auto *Info = State.dyn_cast<ModuleMacroInfo*>())
553 Info->MD = MD;
554 else
555 State = MD;
556 }
557
558 bool isAmbiguous(Preprocessor &PP, const IdentifierInfo *II) const {
559 auto *Info = getModuleInfo(PP, II);
560 return Info ? Info->IsAmbiguous : false;
561 }
562
563 ArrayRef<ModuleMacro *>
564 getActiveModuleMacros(Preprocessor &PP, const IdentifierInfo *II) const {
565 if (auto *Info = getModuleInfo(PP, II))
566 return Info->ActiveModuleMacros;
567 return None;
568 }
569
570 MacroDirective::DefInfo findDirectiveAtLoc(SourceLocation Loc,
571 SourceManager &SourceMgr) const {
572 // FIXME: Incorporate module macros into the result of this.
573 if (auto *Latest = getLatest())
574 return Latest->findDirectiveAtLoc(Loc, SourceMgr);
575 return {};
576 }
577
578 void overrideActiveModuleMacros(Preprocessor &PP, IdentifierInfo *II) {
579 if (auto *Info = getModuleInfo(PP, II)) {
580 Info->OverriddenMacros.insert(Info->OverriddenMacros.end(),
581 Info->ActiveModuleMacros.begin(),
582 Info->ActiveModuleMacros.end());
583 Info->ActiveModuleMacros.clear();
584 Info->IsAmbiguous = false;
585 }
586 }
587
588 ArrayRef<ModuleMacro*> getOverriddenMacros() const {
589 if (auto *Info = State.dyn_cast<ModuleMacroInfo*>())
590 return Info->OverriddenMacros;
591 return None;
592 }
593
594 void setOverriddenMacros(Preprocessor &PP,
595 ArrayRef<ModuleMacro *> Overrides) {
596 auto *Info = State.dyn_cast<ModuleMacroInfo*>();
597 if (!Info) {
598 if (Overrides.empty())
599 return;
600 Info = new (PP.getPreprocessorAllocator())
601 ModuleMacroInfo(State.get<MacroDirective *>());
602 State = Info;
603 }
604 Info->OverriddenMacros.clear();
605 Info->OverriddenMacros.insert(Info->OverriddenMacros.end(),
606 Overrides.begin(), Overrides.end());
607 Info->ActiveModuleMacrosGeneration = 0;
608 }
609 };
610
611 /// For each IdentifierInfo that was associated with a macro, we
612 /// keep a mapping to the history of all macro definitions and #undefs in
613 /// the reverse order (the latest one is in the head of the list).
614 ///
615 /// This mapping lives within the \p CurSubmoduleState.
616 using MacroMap = llvm::DenseMap<const IdentifierInfo *, MacroState>;
617
618 struct SubmoduleState;
619
620 /// Information about a submodule that we're currently building.
621 struct BuildingSubmoduleInfo {
622 /// The module that we are building.
623 Module *M;
624
625 /// The location at which the module was included.
626 SourceLocation ImportLoc;
627
628 /// Whether we entered this submodule via a pragma.
629 bool IsPragma;
630
631 /// The previous SubmoduleState.
632 SubmoduleState *OuterSubmoduleState;
633
634 /// The number of pending module macro names when we started building this.
635 unsigned OuterPendingModuleMacroNames;
636
637 BuildingSubmoduleInfo(Module *M, SourceLocation ImportLoc, bool IsPragma,
638 SubmoduleState *OuterSubmoduleState,
639 unsigned OuterPendingModuleMacroNames)
640 : M(M), ImportLoc(ImportLoc), IsPragma(IsPragma),
641 OuterSubmoduleState(OuterSubmoduleState),
642 OuterPendingModuleMacroNames(OuterPendingModuleMacroNames) {}
643 };
644 SmallVector<BuildingSubmoduleInfo, 8> BuildingSubmoduleStack;
645
646 /// Information about a submodule's preprocessor state.
647 struct SubmoduleState {
648 /// The macros for the submodule.
649 MacroMap Macros;
650
651 /// The set of modules that are visible within the submodule.
652 VisibleModuleSet VisibleModules;
653
654 // FIXME: CounterValue?
655 // FIXME: PragmaPushMacroInfo?
656 };
657 std::map<Module *, SubmoduleState> Submodules;
658
659 /// The preprocessor state for preprocessing outside of any submodule.
660 SubmoduleState NullSubmoduleState;
661
662 /// The current submodule state. Will be \p NullSubmoduleState if we're not
663 /// in a submodule.
664 SubmoduleState *CurSubmoduleState;
665
666 /// The set of known macros exported from modules.
667 llvm::FoldingSet<ModuleMacro> ModuleMacros;
668
669 /// The names of potential module macros that we've not yet processed.
670 llvm::SmallVector<const IdentifierInfo *, 32> PendingModuleMacroNames;
671
672 /// The list of module macros, for each identifier, that are not overridden by
673 /// any other module macro.
674 llvm::DenseMap<const IdentifierInfo *, llvm::TinyPtrVector<ModuleMacro *>>
675 LeafModuleMacros;
676
677 /// Macros that we want to warn because they are not used at the end
678 /// of the translation unit.
679 ///
680 /// We store just their SourceLocations instead of
681 /// something like MacroInfo*. The benefit of this is that when we are
682 /// deserializing from PCH, we don't need to deserialize identifier & macros
683 /// just so that we can report that they are unused, we just warn using
684 /// the SourceLocations of this set (that will be filled by the ASTReader).
685 /// We are using SmallPtrSet instead of a vector for faster removal.
686 using WarnUnusedMacroLocsTy = llvm::SmallPtrSet<SourceLocation, 32>;
687 WarnUnusedMacroLocsTy WarnUnusedMacroLocs;
688
689 /// A "freelist" of MacroArg objects that can be
690 /// reused for quick allocation.
691 MacroArgs *MacroArgCache = nullptr;
692
693 /// For each IdentifierInfo used in a \#pragma push_macro directive,
694 /// we keep a MacroInfo stack used to restore the previous macro value.
695 llvm::DenseMap<IdentifierInfo *, std::vector<MacroInfo *>>
696 PragmaPushMacroInfo;
697
698 // Various statistics we track for performance analysis.
699 unsigned NumDirectives = 0;
700 unsigned NumDefined = 0;
701 unsigned NumUndefined = 0;
702 unsigned NumPragma = 0;
703 unsigned NumIf = 0;
704 unsigned NumElse = 0;
705 unsigned NumEndif = 0;
706 unsigned NumEnteredSourceFiles = 0;
707 unsigned MaxIncludeStackDepth = 0;
708 unsigned NumMacroExpanded = 0;
709 unsigned NumFnMacroExpanded = 0;
710 unsigned NumBuiltinMacroExpanded = 0;
711 unsigned NumFastMacroExpanded = 0;
712 unsigned NumTokenPaste = 0;
713 unsigned NumFastTokenPaste = 0;
714 unsigned NumSkipped = 0;
715
716 /// The predefined macros that preprocessor should use from the
717 /// command line etc.
718 std::string Predefines;
719
720 /// The file ID for the preprocessor predefines.
721 FileID PredefinesFileID;
722
723 /// The file ID for the PCH through header.
724 FileID PCHThroughHeaderFileID;
725
726 /// Whether tokens are being skipped until the through header is seen.
727 bool SkippingUntilPCHThroughHeader = false;
728
729 /// \{
730 /// Cache of macro expanders to reduce malloc traffic.
731 enum { TokenLexerCacheSize = 8 };
732 unsigned NumCachedTokenLexers;
733 std::unique_ptr<TokenLexer> TokenLexerCache[TokenLexerCacheSize];
734 /// \}
735
736 /// Keeps macro expanded tokens for TokenLexers.
737 //
738 /// Works like a stack; a TokenLexer adds the macro expanded tokens that is
739 /// going to lex in the cache and when it finishes the tokens are removed
740 /// from the end of the cache.
741 SmallVector<Token, 16> MacroExpandedTokens;
742 std::vector<std::pair<TokenLexer *, size_t>> MacroExpandingLexersStack;
743
744 /// A record of the macro definitions and expansions that
745 /// occurred during preprocessing.
746 ///
747 /// This is an optional side structure that can be enabled with
748 /// \c createPreprocessingRecord() prior to preprocessing.
749 PreprocessingRecord *Record = nullptr;
750
751 /// Cached tokens state.
752 using CachedTokensTy = SmallVector<Token, 1>;
753
754 /// Cached tokens are stored here when we do backtracking or
755 /// lookahead. They are "lexed" by the CachingLex() method.
756 CachedTokensTy CachedTokens;
757
758 /// The position of the cached token that CachingLex() should
759 /// "lex" next.
760 ///
761 /// If it points beyond the CachedTokens vector, it means that a normal
762 /// Lex() should be invoked.
763 CachedTokensTy::size_type CachedLexPos = 0;
764
765 /// Stack of backtrack positions, allowing nested backtracks.
766 ///
767 /// The EnableBacktrackAtThisPos() method pushes a position to
768 /// indicate where CachedLexPos should be set when the BackTrack() method is
769 /// invoked (at which point the last position is popped).
770 std::vector<CachedTokensTy::size_type> BacktrackPositions;
771
772 struct MacroInfoChain {
773 MacroInfo MI;
774 MacroInfoChain *Next;
775 };
776
777 /// MacroInfos are managed as a chain for easy disposal. This is the head
778 /// of that list.
779 MacroInfoChain *MIChainHead = nullptr;
780
781 void updateOutOfDateIdentifier(IdentifierInfo &II) const;
782
783public:
784 Preprocessor(std::shared_ptr<PreprocessorOptions> PPOpts,
785 DiagnosticsEngine &diags, LangOptions &opts, SourceManager &SM,
786 MemoryBufferCache &PCMCache,
787 HeaderSearch &Headers, ModuleLoader &TheModuleLoader,
788 IdentifierInfoLookup *IILookup = nullptr,
789 bool OwnsHeaderSearch = false,
790 TranslationUnitKind TUKind = TU_Complete);
791
792 ~Preprocessor();
793
794 /// Initialize the preprocessor using information about the target.
795 ///
796 /// \param Target is owned by the caller and must remain valid for the
797 /// lifetime of the preprocessor.
798 /// \param AuxTarget is owned by the caller and must remain valid for
799 /// the lifetime of the preprocessor.
800 void Initialize(const TargetInfo &Target,
801 const TargetInfo *AuxTarget = nullptr);
802
803 /// Initialize the preprocessor to parse a model file
804 ///
805 /// To parse model files the preprocessor of the original source is reused to
806 /// preserver the identifier table. However to avoid some duplicate
807 /// information in the preprocessor some cleanup is needed before it is used
808 /// to parse model files. This method does that cleanup.
809 void InitializeForModelFile();
810
811 /// Cleanup after model file parsing
812 void FinalizeForModelFile();
813
814 /// Retrieve the preprocessor options used to initialize this
815 /// preprocessor.
816 PreprocessorOptions &getPreprocessorOpts() const { return *PPOpts; }
817
818 DiagnosticsEngine &getDiagnostics() const { return *Diags; }
819 void setDiagnostics(DiagnosticsEngine &D) { Diags = &D; }
820
821 const LangOptions &getLangOpts() const { return LangOpts; }
822 const TargetInfo &getTargetInfo() const { return *Target; }
823 const TargetInfo *getAuxTargetInfo() const { return AuxTarget; }
824 FileManager &getFileManager() const { return FileMgr; }
825 SourceManager &getSourceManager() const { return SourceMgr; }
826 MemoryBufferCache &getPCMCache() const { return PCMCache; }
827 HeaderSearch &getHeaderSearchInfo() const { return HeaderInfo; }
828
829 IdentifierTable &getIdentifierTable() { return Identifiers; }
830 const IdentifierTable &getIdentifierTable() const { return Identifiers; }
831 SelectorTable &getSelectorTable() { return Selectors; }
832 Builtin::Context &getBuiltinInfo() { return BuiltinInfo; }
833 llvm::BumpPtrAllocator &getPreprocessorAllocator() { return BP; }
834
835 void setPTHManager(PTHManager* pm);
836
837 PTHManager *getPTHManager() { return PTH.get(); }
838
839 void setExternalSource(ExternalPreprocessorSource *Source) {
840 ExternalSource = Source;
841 }
842
843 ExternalPreprocessorSource *getExternalSource() const {
844 return ExternalSource;
845 }
846
847 /// Retrieve the module loader associated with this preprocessor.
848 ModuleLoader &getModuleLoader() const { return TheModuleLoader; }
849
850 bool hadModuleLoaderFatalFailure() const {
851 return TheModuleLoader.HadFatalFailure;
852 }
853
854 /// True if we are currently preprocessing a #if or #elif directive
855 bool isParsingIfOrElifDirective() const {
856 return ParsingIfOrElifDirective;
857 }
858
859 /// Control whether the preprocessor retains comments in output.
860 void SetCommentRetentionState(bool KeepComments, bool KeepMacroComments) {
861 this->KeepComments = KeepComments | KeepMacroComments;
862 this->KeepMacroComments = KeepMacroComments;
863 }
864
865 bool getCommentRetentionState() const { return KeepComments; }
866
867 void setPragmasEnabled(bool Enabled) { PragmasEnabled = Enabled; }
868 bool getPragmasEnabled() const { return PragmasEnabled; }
869
870 void SetSuppressIncludeNotFoundError(bool Suppress) {
871 SuppressIncludeNotFoundError = Suppress;
872 }
873
874 bool GetSuppressIncludeNotFoundError() {
875 return SuppressIncludeNotFoundError;
876 }
877
878 /// Sets whether the preprocessor is responsible for producing output or if
879 /// it is producing tokens to be consumed by Parse and Sema.
880 void setPreprocessedOutput(bool IsPreprocessedOutput) {
881 PreprocessedOutput = IsPreprocessedOutput;
882 }
883
884 /// Returns true if the preprocessor is responsible for generating output,
885 /// false if it is producing tokens to be consumed by Parse and Sema.
886 bool isPreprocessedOutput() const { return PreprocessedOutput; }
887
888 /// Return true if we are lexing directly from the specified lexer.
889 bool isCurrentLexer(const PreprocessorLexer *L) const {
890 return CurPPLexer == L;
891 }
892
893 /// Return the current lexer being lexed from.
894 ///
895 /// Note that this ignores any potentially active macro expansions and _Pragma
896 /// expansions going on at the time.
897 PreprocessorLexer *getCurrentLexer() const { return CurPPLexer; }
898
899 /// Return the current file lexer being lexed from.
900 ///
901 /// Note that this ignores any potentially active macro expansions and _Pragma
902 /// expansions going on at the time.
903 PreprocessorLexer *getCurrentFileLexer() const;
904
905 /// Return the submodule owning the file being lexed. This may not be
906 /// the current module if we have changed modules since entering the file.
907 Module *getCurrentLexerSubmodule() const { return CurLexerSubmodule; }
908
909 /// Returns the FileID for the preprocessor predefines.
910 FileID getPredefinesFileID() const { return PredefinesFileID; }
911
912 /// \{
913 /// Accessors for preprocessor callbacks.
914 ///
915 /// Note that this class takes ownership of any PPCallbacks object given to
916 /// it.
917 PPCallbacks *getPPCallbacks() const { return Callbacks.get(); }
918 void addPPCallbacks(std::unique_ptr<PPCallbacks> C) {
919 if (Callbacks)
920 C = llvm::make_unique<PPChainedCallbacks>(std::move(C),
921 std::move(Callbacks));
922 Callbacks = std::move(C);
923 }
924 /// \}
925
926 bool isMacroDefined(StringRef Id) {
927 return isMacroDefined(&Identifiers.get(Id));
928 }
929 bool isMacroDefined(const IdentifierInfo *II) {
930 return II->hasMacroDefinition() &&
931 (!getLangOpts().Modules || (bool)getMacroDefinition(II));
932 }
933
934 /// Determine whether II is defined as a macro within the module M,
935 /// if that is a module that we've already preprocessed. Does not check for
936 /// macros imported into M.
937 bool isMacroDefinedInLocalModule(const IdentifierInfo *II, Module *M) {
938 if (!II->hasMacroDefinition())
939 return false;
940 auto I = Submodules.find(M);
941 if (I == Submodules.end())
942 return false;
943 auto J = I->second.Macros.find(II);
944 if (J == I->second.Macros.end())
945 return false;
946 auto *MD = J->second.getLatest();
947 return MD && MD->isDefined();
948 }
949
950 MacroDefinition getMacroDefinition(const IdentifierInfo *II) {
951 if (!II->hasMacroDefinition())
952 return {};
953
954 MacroState &S = CurSubmoduleState->Macros[II];
955 auto *MD = S.getLatest();
956 while (MD && isa<VisibilityMacroDirective>(MD))
957 MD = MD->getPrevious();
958 return MacroDefinition(dyn_cast_or_null<DefMacroDirective>(MD),
959 S.getActiveModuleMacros(*this, II),
960 S.isAmbiguous(*this, II));
961 }
962
963 MacroDefinition getMacroDefinitionAtLoc(const IdentifierInfo *II,
964 SourceLocation Loc) {
965 if (!II->hadMacroDefinition())
966 return {};
967
968 MacroState &S = CurSubmoduleState->Macros[II];
969 MacroDirective::DefInfo DI;
970 if (auto *MD = S.getLatest())
971 DI = MD->findDirectiveAtLoc(Loc, getSourceManager());
972 // FIXME: Compute the set of active module macros at the specified location.
973 return MacroDefinition(DI.getDirective(),
974 S.getActiveModuleMacros(*this, II),
975 S.isAmbiguous(*this, II));
976 }
977
978 /// Given an identifier, return its latest non-imported MacroDirective
979 /// if it is \#define'd and not \#undef'd, or null if it isn't \#define'd.
980 MacroDirective *getLocalMacroDirective(const IdentifierInfo *II) const {
981 if (!II->hasMacroDefinition())
982 return nullptr;
983
984 auto *MD = getLocalMacroDirectiveHistory(II);
985 if (!MD || MD->getDefinition().isUndefined())
986 return nullptr;
987
988 return MD;
989 }
990
991 const MacroInfo *getMacroInfo(const IdentifierInfo *II) const {
992 return const_cast<Preprocessor*>(this)->getMacroInfo(II);
993 }
994
995 MacroInfo *getMacroInfo(const IdentifierInfo *II) {
996 if (!II->hasMacroDefinition())
997 return nullptr;
998 if (auto MD = getMacroDefinition(II))
999 return MD.getMacroInfo();
1000 return nullptr;
1001 }
1002
1003 /// Given an identifier, return the latest non-imported macro
1004 /// directive for that identifier.
1005 ///
1006 /// One can iterate over all previous macro directives from the most recent
1007 /// one.
1008 MacroDirective *getLocalMacroDirectiveHistory(const IdentifierInfo *II) const;
1009
1010 /// Add a directive to the macro directive history for this identifier.
1011 void appendMacroDirective(IdentifierInfo *II, MacroDirective *MD);
1012 DefMacroDirective *appendDefMacroDirective(IdentifierInfo *II, MacroInfo *MI,
1013 SourceLocation Loc) {
1014 DefMacroDirective *MD = AllocateDefMacroDirective(MI, Loc);
1015 appendMacroDirective(II, MD);
1016 return MD;
1017 }
1018 DefMacroDirective *appendDefMacroDirective(IdentifierInfo *II,
1019 MacroInfo *MI) {
1020 return appendDefMacroDirective(II, MI, MI->getDefinitionLoc());
1021 }
1022
1023 /// Set a MacroDirective that was loaded from a PCH file.
1024 void setLoadedMacroDirective(IdentifierInfo *II, MacroDirective *ED,
1025 MacroDirective *MD);
1026
1027 /// Register an exported macro for a module and identifier.
1028 ModuleMacro *addModuleMacro(Module *Mod, IdentifierInfo *II, MacroInfo *Macro,
1029 ArrayRef<ModuleMacro *> Overrides, bool &IsNew);
1030 ModuleMacro *getModuleMacro(Module *Mod, IdentifierInfo *II);
1031
1032 /// Get the list of leaf (non-overridden) module macros for a name.
1033 ArrayRef<ModuleMacro*> getLeafModuleMacros(const IdentifierInfo *II) const {
1034 if (II->isOutOfDate())
1035 updateOutOfDateIdentifier(const_cast<IdentifierInfo&>(*II));
1036 auto I = LeafModuleMacros.find(II);
1037 if (I != LeafModuleMacros.end())
1038 return I->second;
1039 return None;
1040 }
1041
1042 /// \{
1043 /// Iterators for the macro history table. Currently defined macros have
1044 /// IdentifierInfo::hasMacroDefinition() set and an empty
1045 /// MacroInfo::getUndefLoc() at the head of the list.
1046 using macro_iterator = MacroMap::const_iterator;
1047
1048 macro_iterator macro_begin(bool IncludeExternalMacros = true) const;
1049 macro_iterator macro_end(bool IncludeExternalMacros = true) const;
1050
1051 llvm::iterator_range<macro_iterator>
1052 macros(bool IncludeExternalMacros = true) const {
1053 macro_iterator begin = macro_begin(IncludeExternalMacros);
1054 macro_iterator end = macro_end(IncludeExternalMacros);
1055 return llvm::make_range(begin, end);
1056 }
1057
1058 /// \}
1059
1060 /// Return the name of the macro defined before \p Loc that has
1061 /// spelling \p Tokens. If there are multiple macros with same spelling,
1062 /// return the last one defined.
1063 StringRef getLastMacroWithSpelling(SourceLocation Loc,
1064 ArrayRef<TokenValue> Tokens) const;
1065
1066 const std::string &getPredefines() const { return Predefines; }
1067
1068 /// Set the predefines for this Preprocessor.
1069 ///
1070 /// These predefines are automatically injected when parsing the main file.
1071 void setPredefines(const char *P) { Predefines = P; }
1072 void setPredefines(StringRef P) { Predefines = P; }
1073
1074 /// Return information about the specified preprocessor
1075 /// identifier token.
1076 IdentifierInfo *getIdentifierInfo(StringRef Name) const {
1077 return &Identifiers.get(Name);
1078 }
1079
1080 /// Add the specified pragma handler to this preprocessor.
1081 ///
1082 /// If \p Namespace is non-null, then it is a token required to exist on the
1083 /// pragma line before the pragma string starts, e.g. "STDC" or "GCC".
1084 void AddPragmaHandler(StringRef Namespace, PragmaHandler *Handler);
1085 void AddPragmaHandler(PragmaHandler *Handler) {
1086 AddPragmaHandler(StringRef(), Handler);
1087 }
1088
1089 /// Remove the specific pragma handler from this preprocessor.
1090 ///
1091 /// If \p Namespace is non-null, then it should be the namespace that
1092 /// \p Handler was added to. It is an error to remove a handler that
1093 /// has not been registered.
1094 void RemovePragmaHandler(StringRef Namespace, PragmaHandler *Handler);
1095 void RemovePragmaHandler(PragmaHandler *Handler) {
1096 RemovePragmaHandler(StringRef(), Handler);
1097 }
1098
1099 /// Install empty handlers for all pragmas (making them ignored).
1100 void IgnorePragmas();
1101
1102 /// Add the specified comment handler to the preprocessor.
1103 void addCommentHandler(CommentHandler *Handler);
1104
1105 /// Remove the specified comment handler.
1106 ///
1107 /// It is an error to remove a handler that has not been registered.
1108 void removeCommentHandler(CommentHandler *Handler);
1109
1110 /// Set the code completion handler to the given object.
1111 void setCodeCompletionHandler(CodeCompletionHandler &Handler) {
1112 CodeComplete = &Handler;
1113 }
1114
1115 /// Retrieve the current code-completion handler.
1116 CodeCompletionHandler *getCodeCompletionHandler() const {
1117 return CodeComplete;
1118 }
1119
1120 /// Clear out the code completion handler.
1121 void clearCodeCompletionHandler() {
1122 CodeComplete = nullptr;
1123 }
1124
1125 /// Hook used by the lexer to invoke the "natural language" code
1126 /// completion point.
1127 void CodeCompleteNaturalLanguage();
1128
1129 /// Set the code completion token for filtering purposes.
1130 void setCodeCompletionIdentifierInfo(IdentifierInfo *Filter) {
1131 CodeCompletionII = Filter;
1132 }
1133
1134 /// Get the code completion token for filtering purposes.
1135 StringRef getCodeCompletionFilter() {
1136 if (CodeCompletionII)
1137 return CodeCompletionII->getName();
1138 return {};
1139 }
1140
1141 /// Retrieve the preprocessing record, or NULL if there is no
1142 /// preprocessing record.
1143 PreprocessingRecord *getPreprocessingRecord() const { return Record; }
1144
1145 /// Create a new preprocessing record, which will keep track of
1146 /// all macro expansions, macro definitions, etc.
1147 void createPreprocessingRecord();
1148
1149 /// Returns true if the FileEntry is the PCH through header.
1150 bool isPCHThroughHeader(const FileEntry *File);
1151
1152 /// True if creating a PCH with a through header.
1153 bool creatingPCHWithThroughHeader();
1154
1155 /// True if using a PCH with a through header.
1156 bool usingPCHWithThroughHeader();
1157
1158 /// Skip tokens until after the #include of the through header.
1159 void SkipTokensUntilPCHThroughHeader();
1160
1161 /// Process directives while skipping until the through header is found.
1162 void HandleSkippedThroughHeaderDirective(Token &Result,
1163 SourceLocation HashLoc);
1164
1165 /// Enter the specified FileID as the main source file,
1166 /// which implicitly adds the builtin defines etc.
1167 void EnterMainSourceFile();
1168
1169 /// Inform the preprocessor callbacks that processing is complete.
1170 void EndSourceFile();
1171
1172 /// Add a source file to the top of the include stack and
1173 /// start lexing tokens from it instead of the current buffer.
1174 ///
1175 /// Emits a diagnostic, doesn't enter the file, and returns true on error.
1176 bool EnterSourceFile(FileID CurFileID, const DirectoryLookup *Dir,
1177 SourceLocation Loc);
1178
1179 /// Add a Macro to the top of the include stack and start lexing
1180 /// tokens from it instead of the current buffer.
1181 ///
1182 /// \param Args specifies the tokens input to a function-like macro.
1183 /// \param ILEnd specifies the location of the ')' for a function-like macro
1184 /// or the identifier for an object-like macro.
1185 void EnterMacro(Token &Identifier, SourceLocation ILEnd, MacroInfo *Macro,
1186 MacroArgs *Args);
1187
1188 /// Add a "macro" context to the top of the include stack,
1189 /// which will cause the lexer to start returning the specified tokens.
1190 ///
1191 /// If \p DisableMacroExpansion is true, tokens lexed from the token stream
1192 /// will not be subject to further macro expansion. Otherwise, these tokens
1193 /// will be re-macro-expanded when/if expansion is enabled.
1194 ///
1195 /// If \p OwnsTokens is false, this method assumes that the specified stream
1196 /// of tokens has a permanent owner somewhere, so they do not need to be
1197 /// copied. If it is true, it assumes the array of tokens is allocated with
1198 /// \c new[] and the Preprocessor will delete[] it.
1199private:
1200 void EnterTokenStream(const Token *Toks, unsigned NumToks,
1201 bool DisableMacroExpansion, bool OwnsTokens);
1202
1203public:
1204 void EnterTokenStream(std::unique_ptr<Token[]> Toks, unsigned NumToks,
1205 bool DisableMacroExpansion) {
1206 EnterTokenStream(Toks.release(), NumToks, DisableMacroExpansion, true);
1207 }
1208
1209 void EnterTokenStream(ArrayRef<Token> Toks, bool DisableMacroExpansion) {
1210 EnterTokenStream(Toks.data(), Toks.size(), DisableMacroExpansion, false);
1211 }
1212
1213 /// Pop the current lexer/macro exp off the top of the lexer stack.
1214 ///
1215 /// This should only be used in situations where the current state of the
1216 /// top-of-stack lexer is known.
1217 void RemoveTopOfLexerStack();
1218
1219 /// From the point that this method is called, and until
1220 /// CommitBacktrackedTokens() or Backtrack() is called, the Preprocessor
1221 /// keeps track of the lexed tokens so that a subsequent Backtrack() call will
1222 /// make the Preprocessor re-lex the same tokens.
1223 ///
1224 /// Nested backtracks are allowed, meaning that EnableBacktrackAtThisPos can
1225 /// be called multiple times and CommitBacktrackedTokens/Backtrack calls will
1226 /// be combined with the EnableBacktrackAtThisPos calls in reverse order.
1227 ///
1228 /// NOTE: *DO NOT* forget to call either CommitBacktrackedTokens or Backtrack
1229 /// at some point after EnableBacktrackAtThisPos. If you don't, caching of
1230 /// tokens will continue indefinitely.
1231 ///
1232 void EnableBacktrackAtThisPos();
1233
1234 /// Disable the last EnableBacktrackAtThisPos call.
1235 void CommitBacktrackedTokens();
1236
1237 struct CachedTokensRange {
1238 CachedTokensTy::size_type Begin, End;
1239 };
1240
1241private:
1242 /// A range of cached tokens that should be erased after lexing
1243 /// when backtracking requires the erasure of such cached tokens.
1244 Optional<CachedTokensRange> CachedTokenRangeToErase;
1245
1246public:
1247 /// Returns the range of cached tokens that were lexed since
1248 /// EnableBacktrackAtThisPos() was previously called.
1249 CachedTokensRange LastCachedTokenRange();
1250
1251 /// Erase the range of cached tokens that were lexed since
1252 /// EnableBacktrackAtThisPos() was previously called.
1253 void EraseCachedTokens(CachedTokensRange TokenRange);
1254
1255 /// Make Preprocessor re-lex the tokens that were lexed since
1256 /// EnableBacktrackAtThisPos() was previously called.
1257 void Backtrack();
1258
1259 /// True if EnableBacktrackAtThisPos() was called and
1260 /// caching of tokens is on.
1261 bool isBacktrackEnabled() const { return !BacktrackPositions.empty(); }
1262
1263 /// Lex the next token for this preprocessor.
1264 void Lex(Token &Result);
1265
1266 void LexAfterModuleImport(Token &Result);
1267
1268 void makeModuleVisible(Module *M, SourceLocation Loc);
1269
1270 SourceLocation getModuleImportLoc(Module *M) const {
1271 return CurSubmoduleState->VisibleModules.getImportLoc(M);
1272 }
1273
1274 /// Lex a string literal, which may be the concatenation of multiple
1275 /// string literals and may even come from macro expansion.
1276 /// \returns true on success, false if a error diagnostic has been generated.
1277 bool LexStringLiteral(Token &Result, std::string &String,
1278 const char *DiagnosticTag, bool AllowMacroExpansion) {
1279 if (AllowMacroExpansion)
1280 Lex(Result);
1281 else
1282 LexUnexpandedToken(Result);
1283 return FinishLexStringLiteral(Result, String, DiagnosticTag,
1284 AllowMacroExpansion);
1285 }
1286
1287 /// Complete the lexing of a string literal where the first token has
1288 /// already been lexed (see LexStringLiteral).
1289 bool FinishLexStringLiteral(Token &Result, std::string &String,
1290 const char *DiagnosticTag,
1291 bool AllowMacroExpansion);
1292
1293 /// Lex a token. If it's a comment, keep lexing until we get
1294 /// something not a comment.
1295 ///
1296 /// This is useful in -E -C mode where comments would foul up preprocessor
1297 /// directive handling.
1298 void LexNonComment(Token &Result) {
1299 do
1300 Lex(Result);
1301 while (Result.getKind() == tok::comment);
1302 }
1303
1304 /// Just like Lex, but disables macro expansion of identifier tokens.
1305 void LexUnexpandedToken(Token &Result) {
1306 // Disable macro expansion.
1307 bool OldVal = DisableMacroExpansion;
1308 DisableMacroExpansion = true;
1309 // Lex the token.
1310 Lex(Result);
1311
1312 // Reenable it.
1313 DisableMacroExpansion = OldVal;
1314 }
1315
1316 /// Like LexNonComment, but this disables macro expansion of
1317 /// identifier tokens.
1318 void LexUnexpandedNonComment(Token &Result) {
1319 do
1320 LexUnexpandedToken(Result);
1321 while (Result.getKind() == tok::comment);
1322 }
1323
1324 /// Parses a simple integer literal to get its numeric value. Floating
1325 /// point literals and user defined literals are rejected. Used primarily to
1326 /// handle pragmas that accept integer arguments.
1327 bool parseSimpleIntegerLiteral(Token &Tok, uint64_t &Value);
1328
1329 /// Disables macro expansion everywhere except for preprocessor directives.
1330 void SetMacroExpansionOnlyInDirectives() {
1331 DisableMacroExpansion = true;
1332 MacroExpansionInDirectivesOverride = true;
1333 }
1334
1335 /// Peeks ahead N tokens and returns that token without consuming any
1336 /// tokens.
1337 ///
1338 /// LookAhead(0) returns the next token that would be returned by Lex(),
1339 /// LookAhead(1) returns the token after it, etc. This returns normal
1340 /// tokens after phase 5. As such, it is equivalent to using
1341 /// 'Lex', not 'LexUnexpandedToken'.
1342 const Token &LookAhead(unsigned N) {
1343 if (CachedLexPos + N < CachedTokens.size())
1344 return CachedTokens[CachedLexPos+N];
1345 else
1346 return PeekAhead(N+1);
1347 }
1348
1349 /// When backtracking is enabled and tokens are cached,
1350 /// this allows to revert a specific number of tokens.
1351 ///
1352 /// Note that the number of tokens being reverted should be up to the last
1353 /// backtrack position, not more.
1354 void RevertCachedTokens(unsigned N) {
1355 assert(isBacktrackEnabled() &&
1356 "Should only be called when tokens are cached for backtracking");
1357 assert(signed(CachedLexPos) - signed(N) >= signed(BacktrackPositions.back())
1358 && "Should revert tokens up to the last backtrack position, not more");
1359 assert(signed(CachedLexPos) - signed(N) >= 0 &&
1360 "Corrupted backtrack positions ?");
1361 CachedLexPos -= N;
1362 }
1363
1364 /// Enters a token in the token stream to be lexed next.
1365 ///
1366 /// If BackTrack() is called afterwards, the token will remain at the
1367 /// insertion point.
1368 void EnterToken(const Token &Tok) {
1369 EnterCachingLexMode();
1370 CachedTokens.insert(CachedTokens.begin()+CachedLexPos, Tok);
1371 }
1372
1373 /// We notify the Preprocessor that if it is caching tokens (because
1374 /// backtrack is enabled) it should replace the most recent cached tokens
1375 /// with the given annotation token. This function has no effect if
1376 /// backtracking is not enabled.
1377 ///
1378 /// Note that the use of this function is just for optimization, so that the
1379 /// cached tokens doesn't get re-parsed and re-resolved after a backtrack is
1380 /// invoked.
1381 void AnnotateCachedTokens(const Token &Tok) {
1382 assert(Tok.isAnnotation() && "Expected annotation token");
1383 if (CachedLexPos != 0 && isBacktrackEnabled())
1384 AnnotatePreviousCachedTokens(Tok);
1385 }
1386
1387 /// Get the location of the last cached token, suitable for setting the end
1388 /// location of an annotation token.
1389 SourceLocation getLastCachedTokenLocation() const {
1390 assert(CachedLexPos != 0);
1391 return CachedTokens[CachedLexPos-1].getLastLoc();
1392 }
1393
1394 /// Whether \p Tok is the most recent token (`CachedLexPos - 1`) in
1395 /// CachedTokens.
1396 bool IsPreviousCachedToken(const Token &Tok) const;
1397
1398 /// Replace token in `CachedLexPos - 1` in CachedTokens by the tokens
1399 /// in \p NewToks.
1400 ///
1401 /// Useful when a token needs to be split in smaller ones and CachedTokens
1402 /// most recent token must to be updated to reflect that.
1403 void ReplacePreviousCachedToken(ArrayRef<Token> NewToks);
1404
1405 /// Replace the last token with an annotation token.
1406 ///
1407 /// Like AnnotateCachedTokens(), this routine replaces an
1408 /// already-parsed (and resolved) token with an annotation
1409 /// token. However, this routine only replaces the last token with
1410 /// the annotation token; it does not affect any other cached
1411 /// tokens. This function has no effect if backtracking is not
1412 /// enabled.
1413 void ReplaceLastTokenWithAnnotation(const Token &Tok) {
1414 assert(Tok.isAnnotation() && "Expected annotation token");
1415 if (CachedLexPos != 0 && isBacktrackEnabled())
1416 CachedTokens[CachedLexPos-1] = Tok;
1417 }
1418
1419 /// Enter an annotation token into the token stream.
1420 void EnterAnnotationToken(SourceRange Range, tok::TokenKind Kind,
1421 void *AnnotationVal);
1422
1423 /// Update the current token to represent the provided
1424 /// identifier, in order to cache an action performed by typo correction.
1425 void TypoCorrectToken(const Token &Tok) {
1426 assert(Tok.getIdentifierInfo() && "Expected identifier token");
1427 if (CachedLexPos != 0 && isBacktrackEnabled())
1428 CachedTokens[CachedLexPos-1] = Tok;
1429 }
1430
1431 /// Recompute the current lexer kind based on the CurLexer/CurPTHLexer/
1432 /// CurTokenLexer pointers.
1433 void recomputeCurLexerKind();
1434
1435 /// Returns true if incremental processing is enabled
1436 bool isIncrementalProcessingEnabled() const { return IncrementalProcessing; }
1437
1438 /// Enables the incremental processing
1439 void enableIncrementalProcessing(bool value = true) {
1440 IncrementalProcessing = value;
1441 }
1442
1443 /// Specify the point at which code-completion will be performed.
1444 ///
1445 /// \param File the file in which code completion should occur. If
1446 /// this file is included multiple times, code-completion will
1447 /// perform completion the first time it is included. If NULL, this
1448 /// function clears out the code-completion point.
1449 ///
1450 /// \param Line the line at which code completion should occur
1451 /// (1-based).
1452 ///
1453 /// \param Column the column at which code completion should occur
1454 /// (1-based).
1455 ///
1456 /// \returns true if an error occurred, false otherwise.
1457 bool SetCodeCompletionPoint(const FileEntry *File,
1458 unsigned Line, unsigned Column);
1459
1460 /// Determine if we are performing code completion.
1461 bool isCodeCompletionEnabled() const { return CodeCompletionFile != nullptr; }
1462
1463 /// Returns the location of the code-completion point.
1464 ///
1465 /// Returns an invalid location if code-completion is not enabled or the file
1466 /// containing the code-completion point has not been lexed yet.
1467 SourceLocation getCodeCompletionLoc() const { return CodeCompletionLoc; }
1468
1469 /// Returns the start location of the file of code-completion point.
1470 ///
1471 /// Returns an invalid location if code-completion is not enabled or the file
1472 /// containing the code-completion point has not been lexed yet.
1473 SourceLocation getCodeCompletionFileLoc() const {
1474 return CodeCompletionFileLoc;
1475 }
1476
1477 /// Returns true if code-completion is enabled and we have hit the
1478 /// code-completion point.
1479 bool isCodeCompletionReached() const { return CodeCompletionReached; }
1480
1481 /// Note that we hit the code-completion point.
1482 void setCodeCompletionReached() {
1483 assert(isCodeCompletionEnabled() && "Code-completion not enabled!");
1484 CodeCompletionReached = true;
1485 // Silence any diagnostics that occur after we hit the code-completion.
1486 getDiagnostics().setSuppressAllDiagnostics(true);
1487 }
1488
1489 /// The location of the currently-active \#pragma clang
1490 /// arc_cf_code_audited begin.
1491 ///
1492 /// Returns an invalid location if there is no such pragma active.
1493 SourceLocation getPragmaARCCFCodeAuditedLoc() const {
1494 return PragmaARCCFCodeAuditedLoc;
1495 }
1496
1497 /// Set the location of the currently-active \#pragma clang
1498 /// arc_cf_code_audited begin. An invalid location ends the pragma.
1499 void setPragmaARCCFCodeAuditedLoc(SourceLocation Loc) {
1500 PragmaARCCFCodeAuditedLoc = Loc;
1501 }
1502
1503 /// The location of the currently-active \#pragma clang
1504 /// assume_nonnull begin.
1505 ///
1506 /// Returns an invalid location if there is no such pragma active.
1507 SourceLocation getPragmaAssumeNonNullLoc() const {
1508 return PragmaAssumeNonNullLoc;
1509 }
1510
1511 /// Set the location of the currently-active \#pragma clang
1512 /// assume_nonnull begin. An invalid location ends the pragma.
1513 void setPragmaAssumeNonNullLoc(SourceLocation Loc) {
1514 PragmaAssumeNonNullLoc = Loc;
1515 }
1516
1517 /// Set the directory in which the main file should be considered
1518 /// to have been found, if it is not a real file.
1519 void setMainFileDir(const DirectoryEntry *Dir) {
1520 MainFileDir = Dir;
1521 }
1522
1523 /// Instruct the preprocessor to skip part of the main source file.
1524 ///
1525 /// \param Bytes The number of bytes in the preamble to skip.
1526 ///
1527 /// \param StartOfLine Whether skipping these bytes puts the lexer at the
1528 /// start of a line.
1529 void setSkipMainFilePreamble(unsigned Bytes, bool StartOfLine) {
1530 SkipMainFilePreamble.first = Bytes;
1531 SkipMainFilePreamble.second = StartOfLine;
1532 }
1533
1534 /// Forwarding function for diagnostics. This emits a diagnostic at
1535 /// the specified Token's location, translating the token's start
1536 /// position in the current buffer into a SourcePosition object for rendering.
1537 DiagnosticBuilder Diag(SourceLocation Loc, unsigned DiagID) const {
1538 return Diags->Report(Loc, DiagID);
1539 }
1540
1541 DiagnosticBuilder Diag(const Token &Tok, unsigned DiagID) const {
1542 return Diags->Report(Tok.getLocation(), DiagID);
1543 }
1544
1545 /// Return the 'spelling' of the token at the given
1546 /// location; does not go up to the spelling location or down to the
1547 /// expansion location.
1548 ///
1549 /// \param buffer A buffer which will be used only if the token requires
1550 /// "cleaning", e.g. if it contains trigraphs or escaped newlines
1551 /// \param invalid If non-null, will be set \c true if an error occurs.
1552 StringRef getSpelling(SourceLocation loc,
1553 SmallVectorImpl<char> &buffer,
1554 bool *invalid = nullptr) const {
1555 return Lexer::getSpelling(loc, buffer, SourceMgr, LangOpts, invalid);
1556 }
1557
1558 /// Return the 'spelling' of the Tok token.
1559 ///
1560 /// The spelling of a token is the characters used to represent the token in
1561 /// the source file after trigraph expansion and escaped-newline folding. In
1562 /// particular, this wants to get the true, uncanonicalized, spelling of
1563 /// things like digraphs, UCNs, etc.
1564 ///
1565 /// \param Invalid If non-null, will be set \c true if an error occurs.
1566 std::string getSpelling(const Token &Tok, bool *Invalid = nullptr) const {
1567 return Lexer::getSpelling(Tok, SourceMgr, LangOpts, Invalid);
1568 }
1569
1570 /// Get the spelling of a token into a preallocated buffer, instead
1571 /// of as an std::string.
1572 ///
1573 /// The caller is required to allocate enough space for the token, which is
1574 /// guaranteed to be at least Tok.getLength() bytes long. The length of the
1575 /// actual result is returned.
1576 ///
1577 /// Note that this method may do two possible things: it may either fill in
1578 /// the buffer specified with characters, or it may *change the input pointer*
1579 /// to point to a constant buffer with the data already in it (avoiding a
1580 /// copy). The caller is not allowed to modify the returned buffer pointer
1581 /// if an internal buffer is returned.
1582 unsigned getSpelling(const Token &Tok, const char *&Buffer,
1583 bool *Invalid = nullptr) const {
1584 return Lexer::getSpelling(Tok, Buffer, SourceMgr, LangOpts, Invalid);
1585 }
1586
1587 /// Get the spelling of a token into a SmallVector.
1588 ///
1589 /// Note that the returned StringRef may not point to the
1590 /// supplied buffer if a copy can be avoided.
1591 StringRef getSpelling(const Token &Tok,
1592 SmallVectorImpl<char> &Buffer,
1593 bool *Invalid = nullptr) const;
1594
1595 /// Relex the token at the specified location.
1596 /// \returns true if there was a failure, false on success.
1597 bool getRawToken(SourceLocation Loc, Token &Result,
1598 bool IgnoreWhiteSpace = false) {
1599 return Lexer::getRawToken(Loc, Result, SourceMgr, LangOpts, IgnoreWhiteSpace);
1600 }
1601
1602 /// Given a Token \p Tok that is a numeric constant with length 1,
1603 /// return the character.
1604 char
1605 getSpellingOfSingleCharacterNumericConstant(const Token &Tok,
1606 bool *Invalid = nullptr) const {
1607 assert(Tok.is(tok::numeric_constant) &&
1608 Tok.getLength() == 1 && "Called on unsupported token");
1609 assert(!Tok.needsCleaning() && "Token can't need cleaning with length 1");
1610
1611 // If the token is carrying a literal data pointer, just use it.
1612 if (const char *D = Tok.getLiteralData())
1613 return *D;
1614
1615 // Otherwise, fall back on getCharacterData, which is slower, but always
1616 // works.
1617 return *SourceMgr.getCharacterData(Tok.getLocation(), Invalid);
1618 }
1619
1620 /// Retrieve the name of the immediate macro expansion.
1621 ///
1622 /// This routine starts from a source location, and finds the name of the
1623 /// macro responsible for its immediate expansion. It looks through any
1624 /// intervening macro argument expansions to compute this. It returns a
1625 /// StringRef that refers to the SourceManager-owned buffer of the source
1626 /// where that macro name is spelled. Thus, the result shouldn't out-live
1627 /// the SourceManager.
1628 StringRef getImmediateMacroName(SourceLocation Loc) {
1629 return Lexer::getImmediateMacroName(Loc, SourceMgr, getLangOpts());
1630 }
1631
1632 /// Plop the specified string into a scratch buffer and set the
1633 /// specified token's location and length to it.
1634 ///
1635 /// If specified, the source location provides a location of the expansion
1636 /// point of the token.
1637 void CreateString(StringRef Str, Token &Tok,
1638 SourceLocation ExpansionLocStart = SourceLocation(),
1639 SourceLocation ExpansionLocEnd = SourceLocation());
1640
1641 /// Split the first Length characters out of the token starting at TokLoc
1642 /// and return a location pointing to the split token. Re-lexing from the
1643 /// split token will return the split token rather than the original.
1644 SourceLocation SplitToken(SourceLocation TokLoc, unsigned Length);
1645
1646 /// Computes the source location just past the end of the
1647 /// token at this source location.
1648 ///
1649 /// This routine can be used to produce a source location that
1650 /// points just past the end of the token referenced by \p Loc, and
1651 /// is generally used when a diagnostic needs to point just after a
1652 /// token where it expected something different that it received. If
1653 /// the returned source location would not be meaningful (e.g., if
1654 /// it points into a macro), this routine returns an invalid
1655 /// source location.
1656 ///
1657 /// \param Offset an offset from the end of the token, where the source
1658 /// location should refer to. The default offset (0) produces a source
1659 /// location pointing just past the end of the token; an offset of 1 produces
1660 /// a source location pointing to the last character in the token, etc.
1661 SourceLocation getLocForEndOfToken(SourceLocation Loc, unsigned Offset = 0) {
1662 return Lexer::getLocForEndOfToken(Loc, Offset, SourceMgr, LangOpts);
1663 }
1664
1665 /// Returns true if the given MacroID location points at the first
1666 /// token of the macro expansion.
1667 ///
1668 /// \param MacroBegin If non-null and function returns true, it is set to
1669 /// begin location of the macro.
1670 bool isAtStartOfMacroExpansion(SourceLocation loc,
1671 SourceLocation *MacroBegin = nullptr) const {
1672 return Lexer::isAtStartOfMacroExpansion(loc, SourceMgr, LangOpts,
1673 MacroBegin);
1674 }
1675
1676 /// Returns true if the given MacroID location points at the last
1677 /// token of the macro expansion.
1678 ///
1679 /// \param MacroEnd If non-null and function returns true, it is set to
1680 /// end location of the macro.
1681 bool isAtEndOfMacroExpansion(SourceLocation loc,
1682 SourceLocation *MacroEnd = nullptr) const {
1683 return Lexer::isAtEndOfMacroExpansion(loc, SourceMgr, LangOpts, MacroEnd);
1684 }
1685
1686 /// Print the token to stderr, used for debugging.
1687 void DumpToken(const Token &Tok, bool DumpFlags = false) const;
1688 void DumpLocation(SourceLocation Loc) const;
1689 void DumpMacro(const MacroInfo &MI) const;
1690 void dumpMacroInfo(const IdentifierInfo *II);
1691
1692 /// Given a location that specifies the start of a
1693 /// token, return a new location that specifies a character within the token.
1694 SourceLocation AdvanceToTokenCharacter(SourceLocation TokStart,
1695 unsigned Char) const {
1696 return Lexer::AdvanceToTokenCharacter(TokStart, Char, SourceMgr, LangOpts);
1697 }
1698
1699 /// Increment the counters for the number of token paste operations
1700 /// performed.
1701 ///
1702 /// If fast was specified, this is a 'fast paste' case we handled.
1703 void IncrementPasteCounter(bool isFast) {
1704 if (isFast)
1705 ++NumFastTokenPaste;
1706 else
1707 ++NumTokenPaste;
1708 }
1709
1710 void PrintStats();
1711
1712 size_t getTotalMemory() const;
1713
1714 /// When the macro expander pastes together a comment (/##/) in Microsoft
1715 /// mode, this method handles updating the current state, returning the
1716 /// token on the next source line.
1717 void HandleMicrosoftCommentPaste(Token &Tok);
1718
1719 //===--------------------------------------------------------------------===//
1720 // Preprocessor callback methods. These are invoked by a lexer as various
1721 // directives and events are found.
1722
1723 /// Given a tok::raw_identifier token, look up the
1724 /// identifier information for the token and install it into the token,
1725 /// updating the token kind accordingly.
1726 IdentifierInfo *LookUpIdentifierInfo(Token &Identifier) const;
1727
1728private:
1729 llvm::DenseMap<IdentifierInfo*,unsigned> PoisonReasons;
1730
1731public:
1732 /// Specifies the reason for poisoning an identifier.
1733 ///
1734 /// If that identifier is accessed while poisoned, then this reason will be
1735 /// used instead of the default "poisoned" diagnostic.
1736 void SetPoisonReason(IdentifierInfo *II, unsigned DiagID);
1737
1738 /// Display reason for poisoned identifier.
1739 void HandlePoisonedIdentifier(Token & Tok);
1740
1741 void MaybeHandlePoisonedIdentifier(Token & Identifier) {
1742 if(IdentifierInfo * II = Identifier.getIdentifierInfo()) {
1743 if(II->isPoisoned()) {
1744 HandlePoisonedIdentifier(Identifier);
1745 }
1746 }
1747 }
1748
1749private:
1750 /// Identifiers used for SEH handling in Borland. These are only
1751 /// allowed in particular circumstances
1752 // __except block
1753 IdentifierInfo *Ident__exception_code,
1754 *Ident___exception_code,
1755 *Ident_GetExceptionCode;
1756 // __except filter expression
1757 IdentifierInfo *Ident__exception_info,
1758 *Ident___exception_info,
1759 *Ident_GetExceptionInfo;
1760 // __finally
1761 IdentifierInfo *Ident__abnormal_termination,
1762 *Ident___abnormal_termination,
1763 *Ident_AbnormalTermination;
1764
1765 const char *getCurLexerEndPos();
1766 void diagnoseMissingHeaderInUmbrellaDir(const Module &Mod);
1767
1768public:
1769 void PoisonSEHIdentifiers(bool Poison = true); // Borland
1770
1771 /// Callback invoked when the lexer reads an identifier and has
1772 /// filled in the tokens IdentifierInfo member.
1773 ///
1774 /// This callback potentially macro expands it or turns it into a named
1775 /// token (like 'for').
1776 ///
1777 /// \returns true if we actually computed a token, false if we need to
1778 /// lex again.
1779 bool HandleIdentifier(Token &Identifier);
1780
1781 /// Callback invoked when the lexer hits the end of the current file.
1782 ///
1783 /// This either returns the EOF token and returns true, or
1784 /// pops a level off the include stack and returns false, at which point the
1785 /// client should call lex again.
1786 bool HandleEndOfFile(Token &Result, bool isEndOfMacro = false);
1787
1788 /// Callback invoked when the current TokenLexer hits the end of its
1789 /// token stream.
1790 bool HandleEndOfTokenLexer(Token &Result);
1791
1792 /// Callback invoked when the lexer sees a # token at the start of a
1793 /// line.
1794 ///
1795 /// This consumes the directive, modifies the lexer/preprocessor state, and
1796 /// advances the lexer(s) so that the next token read is the correct one.
1797 void HandleDirective(Token &Result);
1798
1799 /// Ensure that the next token is a tok::eod token.
1800 ///
1801 /// If not, emit a diagnostic and consume up until the eod.
1802 /// If \p EnableMacros is true, then we consider macros that expand to zero
1803 /// tokens as being ok.
1804 void CheckEndOfDirective(const char *Directive, bool EnableMacros = false);
1805
1806 /// Read and discard all tokens remaining on the current line until
1807 /// the tok::eod token is found.
1808 void DiscardUntilEndOfDirective();
1809
1810 /// Returns true if the preprocessor has seen a use of
1811 /// __DATE__ or __TIME__ in the file so far.
1812 bool SawDateOrTime() const {
1813 return DATELoc != SourceLocation() || TIMELoc != SourceLocation();
1814 }
1815 unsigned getCounterValue() const { return CounterValue; }
1816 void setCounterValue(unsigned V) { CounterValue = V; }
1817
1818 /// Retrieves the module that we're currently building, if any.
1819 Module *getCurrentModule();
1820
1821 /// Allocate a new MacroInfo object with the provided SourceLocation.
1822 MacroInfo *AllocateMacroInfo(SourceLocation L);
1823
1824 /// Turn the specified lexer token into a fully checked and spelled
1825 /// filename, e.g. as an operand of \#include.
1826 ///
1827 /// The caller is expected to provide a buffer that is large enough to hold
1828 /// the spelling of the filename, but is also expected to handle the case
1829 /// when this method decides to use a different buffer.
1830 ///
1831 /// \returns true if the input filename was in <>'s or false if it was
1832 /// in ""'s.
1833 bool GetIncludeFilenameSpelling(SourceLocation Loc,StringRef &Filename);
1834
1835 /// Given a "foo" or \<foo> reference, look up the indicated file.
1836 ///
1837 /// Returns null on failure. \p isAngled indicates whether the file
1838 /// reference is for system \#include's or not (i.e. using <> instead of "").
1839 const FileEntry *LookupFile(SourceLocation FilenameLoc, StringRef Filename,
1840 bool isAngled, const DirectoryLookup *FromDir,
1841 const FileEntry *FromFile,
1842 const DirectoryLookup *&CurDir,
1843 SmallVectorImpl<char> *SearchPath,
1844 SmallVectorImpl<char> *RelativePath,
1845 ModuleMap::KnownHeader *SuggestedModule,
1846 bool *IsMapped, bool SkipCache = false);
1847
1848 /// Get the DirectoryLookup structure used to find the current
1849 /// FileEntry, if CurLexer is non-null and if applicable.
1850 ///
1851 /// This allows us to implement \#include_next and find directory-specific
1852 /// properties.
1853 const DirectoryLookup *GetCurDirLookup() { return CurDirLookup; }
1854
1855 /// Return true if we're in the top-level file, not in a \#include.
1856 bool isInPrimaryFile() const;
1857
1858 /// Handle cases where the \#include name is expanded
1859 /// from a macro as multiple tokens, which need to be glued together.
1860 ///
1861 /// This occurs for code like:
1862 /// \code
1863 /// \#define FOO <x/y.h>
1864 /// \#include FOO
1865 /// \endcode
1866 /// because in this case, "<x/y.h>" is returned as 7 tokens, not one.
1867 ///
1868 /// This code concatenates and consumes tokens up to the '>' token. It
1869 /// returns false if the > was found, otherwise it returns true if it finds
1870 /// and consumes the EOD marker.
1871 bool ConcatenateIncludeName(SmallString<128> &FilenameBuffer,
1872 SourceLocation &End);
1873
1874 /// Lex an on-off-switch (C99 6.10.6p2) and verify that it is
1875 /// followed by EOD. Return true if the token is not a valid on-off-switch.
1876 bool LexOnOffSwitch(tok::OnOffSwitch &OOS);
1877
1878 bool CheckMacroName(Token &MacroNameTok, MacroUse isDefineUndef,
1879 bool *ShadowFlag = nullptr);
1880
1881 void EnterSubmodule(Module *M, SourceLocation ImportLoc, bool ForPragma);
1882 Module *LeaveSubmodule(bool ForPragma);
1883
1884private:
1885 friend void TokenLexer::ExpandFunctionArguments();
1886
1887 void PushIncludeMacroStack() {
1888 assert(CurLexerKind != CLK_CachingLexer && "cannot push a caching lexer");
1889 IncludeMacroStack.emplace_back(CurLexerKind, CurLexerSubmodule,
1890 std::move(CurLexer), std::move(CurPTHLexer),
1891 CurPPLexer, std::move(CurTokenLexer),
1892 CurDirLookup);
1893 CurPPLexer = nullptr;
1894 }
1895
1896 void PopIncludeMacroStack() {
1897 CurLexer = std::move(IncludeMacroStack.back().TheLexer);
1898 CurPTHLexer = std::move(IncludeMacroStack.back().ThePTHLexer);
1899 CurPPLexer = IncludeMacroStack.back().ThePPLexer;
1900 CurTokenLexer = std::move(IncludeMacroStack.back().TheTokenLexer);
1901 CurDirLookup = IncludeMacroStack.back().TheDirLookup;
1902 CurLexerSubmodule = IncludeMacroStack.back().TheSubmodule;
1903 CurLexerKind = IncludeMacroStack.back().CurLexerKind;
1904 IncludeMacroStack.pop_back();
1905 }
1906
1907 void PropagateLineStartLeadingSpaceInfo(Token &Result);
1908
1909 /// Determine whether we need to create module macros for #defines in the
1910 /// current context.
1911 bool needModuleMacros() const;
1912
1913 /// Update the set of active module macros and ambiguity flag for a module
1914 /// macro name.
1915 void updateModuleMacroInfo(const IdentifierInfo *II, ModuleMacroInfo &Info);
1916
1917 DefMacroDirective *AllocateDefMacroDirective(MacroInfo *MI,
1918 SourceLocation Loc);
1919 UndefMacroDirective *AllocateUndefMacroDirective(SourceLocation UndefLoc);
1920 VisibilityMacroDirective *AllocateVisibilityMacroDirective(SourceLocation Loc,
1921 bool isPublic);
1922
1923 /// Lex and validate a macro name, which occurs after a
1924 /// \#define or \#undef.
1925 ///
1926 /// \param MacroNameTok Token that represents the name defined or undefined.
1927 /// \param IsDefineUndef Kind if preprocessor directive.
1928 /// \param ShadowFlag Points to flag that is set if macro name shadows
1929 /// a keyword.
1930 ///
1931 /// This emits a diagnostic, sets the token kind to eod,
1932 /// and discards the rest of the macro line if the macro name is invalid.
1933 void ReadMacroName(Token &MacroNameTok, MacroUse IsDefineUndef = MU_Other,
1934 bool *ShadowFlag = nullptr);
1935
1936 /// ReadOptionalMacroParameterListAndBody - This consumes all (i.e. the
1937 /// entire line) of the macro's tokens and adds them to MacroInfo, and while
1938 /// doing so performs certain validity checks including (but not limited to):
1939 /// - # (stringization) is followed by a macro parameter
1940 /// \param MacroNameTok - Token that represents the macro name
1941 /// \param ImmediatelyAfterHeaderGuard - Macro follows an #ifdef header guard
1942 ///
1943 /// Either returns a pointer to a MacroInfo object OR emits a diagnostic and
1944 /// returns a nullptr if an invalid sequence of tokens is encountered.
1945 MacroInfo *ReadOptionalMacroParameterListAndBody(
1946 const Token &MacroNameTok, bool ImmediatelyAfterHeaderGuard);
1947
1948 /// The ( starting an argument list of a macro definition has just been read.
1949 /// Lex the rest of the parameters and the closing ), updating \p MI with
1950 /// what we learn and saving in \p LastTok the last token read.
1951 /// Return true if an error occurs parsing the arg list.
1952 bool ReadMacroParameterList(MacroInfo *MI, Token& LastTok);
1953
1954 /// We just read a \#if or related directive and decided that the
1955 /// subsequent tokens are in the \#if'd out portion of the
1956 /// file. Lex the rest of the file, until we see an \#endif. If \p
1957 /// FoundNonSkipPortion is true, then we have already emitted code for part of
1958 /// this \#if directive, so \#else/\#elif blocks should never be entered. If
1959 /// \p FoundElse is false, then \#else directives are ok, if not, then we have
1960 /// already seen one so a \#else directive is a duplicate. When this returns,
1961 /// the caller can lex the first valid token.
1962 void SkipExcludedConditionalBlock(SourceLocation HashTokenLoc,
1963 SourceLocation IfTokenLoc,
1964 bool FoundNonSkipPortion, bool FoundElse,
1965 SourceLocation ElseLoc = SourceLocation());
1966
1967 /// A fast PTH version of SkipExcludedConditionalBlock.
1968 void PTHSkipExcludedConditionalBlock();
1969
1970 /// Information about the result for evaluating an expression for a
1971 /// preprocessor directive.
1972 struct DirectiveEvalResult {
1973 /// Whether the expression was evaluated as true or not.
1974 bool Conditional;
1975
1976 /// True if the expression contained identifiers that were undefined.
1977 bool IncludedUndefinedIds;
1978 };
1979
1980 /// Evaluate an integer constant expression that may occur after a
1981 /// \#if or \#elif directive and return a \p DirectiveEvalResult object.
1982 ///
1983 /// If the expression is equivalent to "!defined(X)" return X in IfNDefMacro.
1984 DirectiveEvalResult EvaluateDirectiveExpression(IdentifierInfo *&IfNDefMacro);
1985
1986 /// Install the standard preprocessor pragmas:
1987 /// \#pragma GCC poison/system_header/dependency and \#pragma once.
1988 void RegisterBuiltinPragmas();
1989
1990 /// Register builtin macros such as __LINE__ with the identifier table.
1991 void RegisterBuiltinMacros();
1992
1993 /// If an identifier token is read that is to be expanded as a macro, handle
1994 /// it and return the next token as 'Tok'. If we lexed a token, return true;
1995 /// otherwise the caller should lex again.
1996 bool HandleMacroExpandedIdentifier(Token &Tok, const MacroDefinition &MD);
1997
1998 /// Cache macro expanded tokens for TokenLexers.
1999 //
2000 /// Works like a stack; a TokenLexer adds the macro expanded tokens that is
2001 /// going to lex in the cache and when it finishes the tokens are removed
2002 /// from the end of the cache.
2003 Token *cacheMacroExpandedTokens(TokenLexer *tokLexer,
2004 ArrayRef<Token> tokens);
2005
2006 void removeCachedMacroExpandedTokensOfLastLexer();
2007
2008 /// Determine whether the next preprocessor token to be
2009 /// lexed is a '('. If so, consume the token and return true, if not, this
2010 /// method should have no observable side-effect on the lexed tokens.
2011 bool isNextPPTokenLParen();
2012
2013 /// After reading "MACRO(", this method is invoked to read all of the formal
2014 /// arguments specified for the macro invocation. Returns null on error.
2015 MacroArgs *ReadMacroCallArgumentList(Token &MacroName, MacroInfo *MI,
2016 SourceLocation &ExpansionEnd);
2017
2018 /// If an identifier token is read that is to be expanded
2019 /// as a builtin macro, handle it and return the next token as 'Tok'.
2020 void ExpandBuiltinMacro(Token &Tok);
2021
2022 /// Read a \c _Pragma directive, slice it up, process it, then
2023 /// return the first token after the directive.
2024 /// This assumes that the \c _Pragma token has just been read into \p Tok.
2025 void Handle_Pragma(Token &Tok);
2026
2027 /// Like Handle_Pragma except the pragma text is not enclosed within
2028 /// a string literal.
2029 void HandleMicrosoft__pragma(Token &Tok);
2030
2031 /// Add a lexer to the top of the include stack and
2032 /// start lexing tokens from it instead of the current buffer.
2033 void EnterSourceFileWithLexer(Lexer *TheLexer, const DirectoryLookup *Dir);
2034
2035 /// Add a lexer to the top of the include stack and
2036 /// start getting tokens from it using the PTH cache.
2037 void EnterSourceFileWithPTH(PTHLexer *PL, const DirectoryLookup *Dir);
2038
2039 /// Set the FileID for the preprocessor predefines.
2040 void setPredefinesFileID(FileID FID) {
2041 assert(PredefinesFileID.isInvalid() && "PredefinesFileID already set!");
2042 PredefinesFileID = FID;
2043 }
2044
2045 /// Set the FileID for the PCH through header.
2046 void setPCHThroughHeaderFileID(FileID FID);
2047
2048 /// Returns true if we are lexing from a file and not a
2049 /// pragma or a macro.
2050 static bool IsFileLexer(const Lexer* L, const PreprocessorLexer* P) {
2051 return L ? !L->isPragmaLexer() : P != nullptr;
2052 }
2053
2054 static bool IsFileLexer(const IncludeStackInfo& I) {
2055 return IsFileLexer(I.TheLexer.get(), I.ThePPLexer);
2056 }
2057
2058 bool IsFileLexer() const {
2059 return IsFileLexer(CurLexer.get(), CurPPLexer);
2060 }
2061
2062 //===--------------------------------------------------------------------===//
2063 // Caching stuff.
2064 void CachingLex(Token &Result);
2065
2066 bool InCachingLexMode() const {
2067 // If the Lexer pointers are 0 and IncludeMacroStack is empty, it means
2068 // that we are past EOF, not that we are in CachingLex mode.
2069 return !CurPPLexer && !CurTokenLexer && !CurPTHLexer &&
2070 !IncludeMacroStack.empty();
2071 }
2072
2073 void EnterCachingLexMode();
2074
2075 void ExitCachingLexMode() {
2076 if (InCachingLexMode())
2077 RemoveTopOfLexerStack();
2078 }
2079
2080 const Token &PeekAhead(unsigned N);
2081 void AnnotatePreviousCachedTokens(const Token &Tok);
2082
2083 //===--------------------------------------------------------------------===//
2084 /// Handle*Directive - implement the various preprocessor directives. These
2085 /// should side-effect the current preprocessor object so that the next call
2086 /// to Lex() will return the appropriate token next.
2087 void HandleLineDirective();
2088 void HandleDigitDirective(Token &Tok);
2089 void HandleUserDiagnosticDirective(Token &Tok, bool isWarning);
2090 void HandleIdentSCCSDirective(Token &Tok);
2091 void HandleMacroPublicDirective(Token &Tok);
2092 void HandleMacroPrivateDirective();
2093
2094 // File inclusion.
2095 void HandleIncludeDirective(SourceLocation HashLoc,
2096 Token &Tok,
2097 const DirectoryLookup *LookupFrom = nullptr,
2098 const FileEntry *LookupFromFile = nullptr,
2099 bool isImport = false);
2100 void HandleIncludeNextDirective(SourceLocation HashLoc, Token &Tok);
2101 void HandleIncludeMacrosDirective(SourceLocation HashLoc, Token &Tok);
2102 void HandleImportDirective(SourceLocation HashLoc, Token &Tok);
2103 void HandleMicrosoftImportDirective(Token &Tok);
2104
2105public:
2106 /// Check that the given module is available, producing a diagnostic if not.
2107 /// \return \c true if the check failed (because the module is not available).
2108 /// \c false if the module appears to be usable.
2109 static bool checkModuleIsAvailable(const LangOptions &LangOpts,
2110 const TargetInfo &TargetInfo,
2111 DiagnosticsEngine &Diags, Module *M);
2112
2113 // Module inclusion testing.
2114 /// Find the module that owns the source or header file that
2115 /// \p Loc points to. If the location is in a file that was included
2116 /// into a module, or is outside any module, returns nullptr.
2117 Module *getModuleForLocation(SourceLocation Loc);
2118
2119 /// We want to produce a diagnostic at location IncLoc concerning a
2120 /// missing module import.
2121 ///
2122 /// \param IncLoc The location at which the missing import was detected.
2123 /// \param M The desired module.
2124 /// \param MLoc A location within the desired module at which some desired
2125 /// effect occurred (eg, where a desired entity was declared).
2126 ///
2127 /// \return A file that can be #included to import a module containing MLoc.
2128 /// Null if no such file could be determined or if a #include is not
2129 /// appropriate.
2130 const FileEntry *getModuleHeaderToIncludeForDiagnostics(SourceLocation IncLoc,
2131 Module *M,
2132 SourceLocation MLoc);
2133
2134 bool isRecordingPreamble() const {
2135 return PreambleConditionalStack.isRecording();
2136 }
2137
2138 bool hasRecordedPreamble() const {
2139 return PreambleConditionalStack.hasRecordedPreamble();
2140 }
2141
2142 ArrayRef<PPConditionalInfo> getPreambleConditionalStack() const {
2143 return PreambleConditionalStack.getStack();
2144 }
2145
2146 void setRecordedPreambleConditionalStack(ArrayRef<PPConditionalInfo> s) {
2147 PreambleConditionalStack.setStack(s);
2148 }
2149
2150 void setReplayablePreambleConditionalStack(ArrayRef<PPConditionalInfo> s,
2151 llvm::Optional<PreambleSkipInfo> SkipInfo) {
2152 PreambleConditionalStack.startReplaying();
2153 PreambleConditionalStack.setStack(s);
2154 PreambleConditionalStack.SkipInfo = SkipInfo;
2155 }
2156
2157 llvm::Optional<PreambleSkipInfo> getPreambleSkipInfo() const {
2158 return PreambleConditionalStack.SkipInfo;
2159 }
2160
2161private:
2162 /// After processing predefined file, initialize the conditional stack from
2163 /// the preamble.
2164 void replayPreambleConditionalStack();
2165
2166 // Macro handling.
2167 void HandleDefineDirective(Token &Tok, bool ImmediatelyAfterTopLevelIfndef);
2168 void HandleUndefDirective();
2169
2170 // Conditional Inclusion.
2171 void HandleIfdefDirective(Token &Tok, const Token &HashToken,
2172 bool isIfndef, bool ReadAnyTokensBeforeDirective);
2173 void HandleIfDirective(Token &Tok, const Token &HashToken,
2174 bool ReadAnyTokensBeforeDirective);
2175 void HandleEndifDirective(Token &Tok);
2176 void HandleElseDirective(Token &Tok, const Token &HashToken);
2177 void HandleElifDirective(Token &Tok, const Token &HashToken);
2178
2179 // Pragmas.
2180 void HandlePragmaDirective(SourceLocation IntroducerLoc,
2181 PragmaIntroducerKind Introducer);
2182
2183public:
2184 void HandlePragmaOnce(Token &OnceTok);
2185 void HandlePragmaMark();
2186 void HandlePragmaPoison();
2187 void HandlePragmaSystemHeader(Token &SysHeaderTok);
2188 void HandlePragmaDependency(Token &DependencyTok);
2189 void HandlePragmaPushMacro(Token &Tok);
2190 void HandlePragmaPopMacro(Token &Tok);
2191 void HandlePragmaIncludeAlias(Token &Tok);
2192 void HandlePragmaModuleBuild(Token &Tok);
2193 IdentifierInfo *ParsePragmaPushOrPopMacro(Token &Tok);
2194
2195 // Return true and store the first token only if any CommentHandler
2196 // has inserted some tokens and getCommentRetentionState() is false.
2197 bool HandleComment(Token &Token, SourceRange Comment);
2198
2199 /// A macro is used, update information about macros that need unused
2200 /// warnings.
2201 void markMacroAsUsed(MacroInfo *MI);
2202};
2203
2204/// Abstract base class that describes a handler that will receive
2205/// source ranges for each of the comments encountered in the source file.
2206class CommentHandler {
2207public:
2208 virtual ~CommentHandler();
2209
2210 // The handler shall return true if it has pushed any tokens
2211 // to be read using e.g. EnterToken or EnterTokenStream.
2212 virtual bool HandleComment(Preprocessor &PP, SourceRange Comment) = 0;
2213};
2214
2215/// Registry of pragma handlers added by plugins
2216using PragmaHandlerRegistry = llvm::Registry<PragmaHandler>;
2217
2218} // namespace clang
2219
2220#endif // LLVM_CLANG_LEX_PREPROCESSOR_H
2221

Warning: That file was not part of the compilation database. It may have many parsing errors.