1	//===- Preprocessor.h - C Language Family Preprocessor ----------*- C++ -*-===//
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	/// \file
10	/// Defines the clang::Preprocessor interface.
11	//
12	//===----------------------------------------------------------------------===//
13
14	#ifndef LLVM_CLANG_LEX_PREPROCESSOR_H
15	#define LLVM_CLANG_LEX_PREPROCESSOR_H
16
17	#include "clang/Basic/Diagnostic.h"
18	#include "clang/Basic/DiagnosticIDs.h"
19	#include "clang/Basic/IdentifierTable.h"
20	#include "clang/Basic/LLVM.h"
21	#include "clang/Basic/LangOptions.h"
22	#include "clang/Basic/Module.h"
23	#include "clang/Basic/SourceLocation.h"
24	#include "clang/Basic/SourceManager.h"
25	#include "clang/Basic/TokenKinds.h"
26	#include "clang/Lex/HeaderSearch.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/Token.h"
33	#include "clang/Lex/TokenLexer.h"
34	#include "llvm/ADT/ArrayRef.h"
35	#include "llvm/ADT/DenseMap.h"
36	#include "llvm/ADT/FoldingSet.h"
37	#include "llvm/ADT/FunctionExtras.h"
38	#include "llvm/ADT/PointerUnion.h"
39	#include "llvm/ADT/STLExtras.h"
40	#include "llvm/ADT/SmallPtrSet.h"
41	#include "llvm/ADT/SmallVector.h"
42	#include "llvm/ADT/StringRef.h"
43	#include "llvm/ADT/TinyPtrVector.h"
44	#include "llvm/ADT/iterator_range.h"
45	#include "llvm/Support/Allocator.h"
46	#include "llvm/Support/Casting.h"
47	#include "llvm/Support/Registry.h"
48	#include <cassert>
49	#include <cstddef>
50	#include <cstdint>
51	#include <map>
52	#include <memory>
53	#include <optional>
54	#include <string>
55	#include <utility>
56	#include <vector>
57
58	namespace llvm {
59
60	template<unsigned InternalLen> class SmallString;
61
62	} // namespace llvm
63
64	namespace clang {
65
66	class CodeCompletionHandler;
67	class CommentHandler;
68	class DirectoryEntry;
69	class EmptylineHandler;
70	class ExternalPreprocessorSource;
71	class FileEntry;
72	class FileManager;
73	class HeaderSearch;
74	class MacroArgs;
75	class PragmaHandler;
76	class PragmaNamespace;
77	class PreprocessingRecord;
78	class PreprocessorLexer;
79	class PreprocessorOptions;
80	class ScratchBuffer;
81	class TargetInfo;
82
83	namespace Builtin {
84	class Context;
85	}
86
87	/// Stores token information for comparing actual tokens with
88	/// predefined values. Only handles simple tokens and identifiers.
89	class TokenValue {
90	tok::TokenKind Kind;
91	IdentifierInfo *II;
92
93	public:
94	TokenValue(tok::TokenKind Kind) : Kind(Kind), II(nullptr) {
95	assert(Kind != tok::raw_identifier && "Raw identifiers are not supported.");
96	assert(Kind != tok::identifier &&
97	"Identifiers should be created by TokenValue(IdentifierInfo *)");
98	assert(!tok::isLiteral(Kind) && "Literals are not supported.");
99	assert(!tok::isAnnotation(Kind) && "Annotations are not supported.");
100	}
101
102	TokenValue(IdentifierInfo *II) : Kind(tok::identifier), II(II) {}
103
104	bool operator==(const Token &Tok) const {
105	return Tok.getKind() == Kind &&
106	(!II || II == Tok.getIdentifierInfo());
107	}
108	};
109
110	/// Context in which macro name is used.
111	enum MacroUse {
112	// other than #define or #undef
113	MU_Other = 0,
114
115	// macro name specified in #define
116	MU_Define = 1,
117
118	// macro name specified in #undef
119	MU_Undef = 2
120	};
121
122	/// Engages in a tight little dance with the lexer to efficiently
123	/// preprocess tokens.
124	///
125	/// Lexers know only about tokens within a single source file, and don't
126	/// know anything about preprocessor-level issues like the \#include stack,
127	/// token expansion, etc.
128	class Preprocessor {
129	friend class VAOptDefinitionContext;
130	friend class VariadicMacroScopeGuard;
131
132	llvm::unique_function<void(const clang::Token &)> OnToken;
133	std::shared_ptr<PreprocessorOptions> PPOpts;
134	DiagnosticsEngine *Diags;
135	const LangOptions &LangOpts;
136	const TargetInfo *Target = nullptr;
137	const TargetInfo *AuxTarget = nullptr;
138	FileManager &FileMgr;
139	SourceManager &SourceMgr;
140	std::unique_ptr<ScratchBuffer> ScratchBuf;
141	HeaderSearch &HeaderInfo;
142	ModuleLoader &TheModuleLoader;
143
144	/// External source of macros.
145	ExternalPreprocessorSource *ExternalSource;
146
147	/// A BumpPtrAllocator object used to quickly allocate and release
148	/// objects internal to the Preprocessor.
149	llvm::BumpPtrAllocator BP;
150
151	/// Identifiers for builtin macros and other builtins.
152	IdentifierInfo *Ident__LINE__, *Ident__FILE__; // __LINE__, __FILE__
153	IdentifierInfo *Ident__DATE__, *Ident__TIME__; // __DATE__, __TIME__
154	IdentifierInfo *Ident__INCLUDE_LEVEL__; // __INCLUDE_LEVEL__
155	IdentifierInfo *Ident__BASE_FILE__; // __BASE_FILE__
156	IdentifierInfo *Ident__FILE_NAME__; // __FILE_NAME__
157	IdentifierInfo *Ident__TIMESTAMP__; // __TIMESTAMP__
158	IdentifierInfo *Ident__COUNTER__; // __COUNTER__
159	IdentifierInfo *Ident_Pragma, *Ident__pragma; // _Pragma, __pragma
160	IdentifierInfo *Ident__identifier; // __identifier
161	IdentifierInfo *Ident__VA_ARGS__; // __VA_ARGS__
162	IdentifierInfo *Ident__VA_OPT__; // __VA_OPT__
163	IdentifierInfo *Ident__has_feature; // __has_feature
164	IdentifierInfo *Ident__has_extension; // __has_extension
165	IdentifierInfo *Ident__has_builtin; // __has_builtin
166	IdentifierInfo *Ident__has_constexpr_builtin; // __has_constexpr_builtin
167	IdentifierInfo *Ident__has_attribute; // __has_attribute
168	IdentifierInfo *Ident__has_include; // __has_include
169	IdentifierInfo *Ident__has_include_next; // __has_include_next
170	IdentifierInfo *Ident__has_warning; // __has_warning
171	IdentifierInfo *Ident__is_identifier; // __is_identifier
172	IdentifierInfo *Ident__building_module; // __building_module
173	IdentifierInfo *Ident__MODULE__; // __MODULE__
174	IdentifierInfo *Ident__has_cpp_attribute; // __has_cpp_attribute
175	IdentifierInfo *Ident__has_c_attribute; // __has_c_attribute
176	IdentifierInfo *Ident__has_declspec; // __has_declspec_attribute
177	IdentifierInfo *Ident__is_target_arch; // __is_target_arch
178	IdentifierInfo *Ident__is_target_vendor; // __is_target_vendor
179	IdentifierInfo *Ident__is_target_os; // __is_target_os
180	IdentifierInfo *Ident__is_target_environment; // __is_target_environment
181	IdentifierInfo *Ident__is_target_variant_os;
182	IdentifierInfo *Ident__is_target_variant_environment;
183	IdentifierInfo *Ident__FLT_EVAL_METHOD__; // __FLT_EVAL_METHOD
184
185	// Weak, only valid (and set) while InMacroArgs is true.
186	Token* ArgMacro;
187
188	SourceLocation DATELoc, TIMELoc;
189
190	// FEM_UnsetOnCommandLine means that an explicit evaluation method was
191	// not specified on the command line. The target is queried to set the
192	// default evaluation method.
193	LangOptions::FPEvalMethodKind CurrentFPEvalMethod =
194	LangOptions::FPEvalMethodKind::FEM_UnsetOnCommandLine;
195
196	// The most recent pragma location where the floating point evaluation
197	// method was modified. This is used to determine whether the
198	// 'pragma clang fp eval_method' was used whithin the current scope.
199	SourceLocation LastFPEvalPragmaLocation;
200
201	LangOptions::FPEvalMethodKind TUFPEvalMethod =
202	LangOptions::FPEvalMethodKind::FEM_UnsetOnCommandLine;
203
204	// Next __COUNTER__ value, starts at 0.
205	unsigned CounterValue = 0;
206
207	enum {
208	/// Maximum depth of \#includes.
209	MaxAllowedIncludeStackDepth = 200
210	};
211
212	// State that is set before the preprocessor begins.
213	bool KeepComments : 1;
214	bool KeepMacroComments : 1;
215	bool SuppressIncludeNotFoundError : 1;
216
217	// State that changes while the preprocessor runs:
218	bool InMacroArgs : 1; // True if parsing fn macro invocation args.
219
220	/// Whether the preprocessor owns the header search object.
221	bool OwnsHeaderSearch : 1;
222
223	/// True if macro expansion is disabled.
224	bool DisableMacroExpansion : 1;
225
226	/// Temporarily disables DisableMacroExpansion (i.e. enables expansion)
227	/// when parsing preprocessor directives.
228	bool MacroExpansionInDirectivesOverride : 1;
229
230	class ResetMacroExpansionHelper;
231
232	/// Whether we have already loaded macros from the external source.
233	mutable bool ReadMacrosFromExternalSource : 1;
234
235	/// True if pragmas are enabled.
236	bool PragmasEnabled : 1;
237
238	/// True if the current build action is a preprocessing action.
239	bool PreprocessedOutput : 1;
240
241	/// True if we are currently preprocessing a #if or #elif directive
242	bool ParsingIfOrElifDirective;
243
244	/// True if we are pre-expanding macro arguments.
245	bool InMacroArgPreExpansion;
246
247	/// Mapping/lookup information for all identifiers in
248	/// the program, including program keywords.
249	mutable IdentifierTable Identifiers;
250
251	/// This table contains all the selectors in the program.
252	///
253	/// Unlike IdentifierTable above, this table *isn't* populated by the
254	/// preprocessor. It is declared/expanded here because its role/lifetime is
255	/// conceptually similar to the IdentifierTable. In addition, the current
256	/// control flow (in clang::ParseAST()), make it convenient to put here.
257	///
258	/// FIXME: Make sure the lifetime of Identifiers/Selectors *isn't* tied to
259	/// the lifetime of the preprocessor.
260	SelectorTable Selectors;
261
262	/// Information about builtins.
263	std::unique_ptr<Builtin::Context> BuiltinInfo;
264
265	/// Tracks all of the pragmas that the client registered
266	/// with this preprocessor.
267	std::unique_ptr<PragmaNamespace> PragmaHandlers;
268
269	/// Pragma handlers of the original source is stored here during the
270	/// parsing of a model file.
271	std::unique_ptr<PragmaNamespace> PragmaHandlersBackup;
272
273	/// Tracks all of the comment handlers that the client registered
274	/// with this preprocessor.
275	std::vector<CommentHandler *> CommentHandlers;
276
277	/// Empty line handler.
278	EmptylineHandler *Emptyline = nullptr;
279
280	/// True to avoid tearing down the lexer etc on EOF
281	bool IncrementalProcessing = false;
282
283	public:
284	/// The kind of translation unit we are processing.
285	const TranslationUnitKind TUKind;
286
287	/// Returns a pointer into the given file's buffer that's guaranteed
288	/// to be between tokens. The returned pointer is always before \p Start.
289	/// The maximum distance betweenthe returned pointer and \p Start is
290	/// limited by a constant value, but also an implementation detail.
291	/// If no such check point exists, \c nullptr is returned.
292	const char *getCheckPoint(FileID FID, const char *Start) const;
293
294	private:
295	/// The code-completion handler.
296	CodeCompletionHandler *CodeComplete = nullptr;
297
298	/// The file that we're performing code-completion for, if any.
299	const FileEntry *CodeCompletionFile = nullptr;
300
301	/// The offset in file for the code-completion point.
302	unsigned CodeCompletionOffset = 0;
303
304	/// The location for the code-completion point. This gets instantiated
305	/// when the CodeCompletionFile gets \#include'ed for preprocessing.
306	SourceLocation CodeCompletionLoc;
307
308	/// The start location for the file of the code-completion point.
309	///
310	/// This gets instantiated when the CodeCompletionFile gets \#include'ed
311	/// for preprocessing.
312	SourceLocation CodeCompletionFileLoc;
313
314	/// The source location of the \c import contextual keyword we just
315	/// lexed, if any.
316	SourceLocation ModuleImportLoc;
317
318	/// The import path for named module that we're currently processing.
319	SmallVector<std::pair<IdentifierInfo *, SourceLocation>, 2> NamedModuleImportPath;
320
321	llvm::DenseMap<FileID, SmallVector<const char *>> CheckPoints;
322	unsigned CheckPointCounter = 0;
323
324	/// Whether the import is an `@import` or a standard c++ modules import.
325	bool IsAtImport = false;
326
327	/// Whether the last token we lexed was an '@'.
328	bool LastTokenWasAt = false;
329
330	/// A position within a C++20 import-seq.
331	class StdCXXImportSeq {
332	public:
333	enum State : int {
334	// Positive values represent a number of unclosed brackets.
335	AtTopLevel = 0,
336	AfterTopLevelTokenSeq = -1,
337	AfterExport = -2,
338	AfterImportSeq = -3,
339	};
340
341	StdCXXImportSeq(State S) : S(S) {}
342
343	/// Saw any kind of open bracket.
344	void handleOpenBracket() {
345	S = static_cast<State>(std::max<int>(a: S, b: 0) + 1);
346	}
347	/// Saw any kind of close bracket other than '}'.
348	void handleCloseBracket() {
349	S = static_cast<State>(std::max<int>(a: S, b: 1) - 1);
350	}
351	/// Saw a close brace.
352	void handleCloseBrace() {
353	handleCloseBracket();
354	if (S == AtTopLevel && !AfterHeaderName)
355	S = AfterTopLevelTokenSeq;
356	}
357	/// Saw a semicolon.
358	void handleSemi() {
359	if (atTopLevel()) {
360	S = AfterTopLevelTokenSeq;
361	AfterHeaderName = false;
362	}
363	}
364
365	/// Saw an 'export' identifier.
366	void handleExport() {
367	if (S == AfterTopLevelTokenSeq)
368	S = AfterExport;
369	else if (S <= 0)
370	S = AtTopLevel;
371	}
372	/// Saw an 'import' identifier.
373	void handleImport() {
374	if (S == AfterTopLevelTokenSeq || S == AfterExport)
375	S = AfterImportSeq;
376	else if (S <= 0)
377	S = AtTopLevel;
378	}
379
380	/// Saw a 'header-name' token; do not recognize any more 'import' tokens
381	/// until we reach a top-level semicolon.
382	void handleHeaderName() {
383	if (S == AfterImportSeq)
384	AfterHeaderName = true;
385	handleMisc();
386	}
387
388	/// Saw any other token.
389	void handleMisc() {
390	if (S <= 0)
391	S = AtTopLevel;
392	}
393
394	bool atTopLevel() { return S <= 0; }
395	bool afterImportSeq() { return S == AfterImportSeq; }
396	bool afterTopLevelSeq() { return S == AfterTopLevelTokenSeq; }
397
398	private:
399	State S;
400	/// Whether we're in the pp-import-suffix following the header-name in a
401	/// pp-import. If so, a close-brace is not sufficient to end the
402	/// top-level-token-seq of an import-seq.
403	bool AfterHeaderName = false;
404	};
405
406	/// Our current position within a C++20 import-seq.
407	StdCXXImportSeq StdCXXImportSeqState = StdCXXImportSeq::AfterTopLevelTokenSeq;
408
409	/// Track whether we are in a Global Module Fragment
410	class TrackGMF {
411	public:
412	enum GMFState : int {
413	GMFActive = 1,
414	MaybeGMF = 0,
415	BeforeGMFIntroducer = -1,
416	GMFAbsentOrEnded = -2,
417	};
418
419	TrackGMF(GMFState S) : S(S) {}
420
421	/// Saw a semicolon.
422	void handleSemi() {
423	// If it is immediately after the first instance of the module keyword,
424	// then that introduces the GMF.
425	if (S == MaybeGMF)
426	S = GMFActive;
427	}
428
429	/// Saw an 'export' identifier.
430	void handleExport() {
431	// The presence of an 'export' keyword always ends or excludes a GMF.
432	S = GMFAbsentOrEnded;
433	}
434
435	/// Saw an 'import' identifier.
436	void handleImport(bool AfterTopLevelTokenSeq) {
437	// If we see this before any 'module' kw, then we have no GMF.
438	if (AfterTopLevelTokenSeq && S == BeforeGMFIntroducer)
439	S = GMFAbsentOrEnded;
440	}
441
442	/// Saw a 'module' identifier.
443	void handleModule(bool AfterTopLevelTokenSeq) {
444	// This was the first module identifier and not preceded by any token
445	// that would exclude a GMF. It could begin a GMF, but only if directly
446	// followed by a semicolon.
447	if (AfterTopLevelTokenSeq && S == BeforeGMFIntroducer)
448	S = MaybeGMF;
449	else
450	S = GMFAbsentOrEnded;
451	}
452
453	/// Saw any other token.
454	void handleMisc() {
455	// We saw something other than ; after the 'module' kw, so not a GMF.
456	if (S == MaybeGMF)
457	S = GMFAbsentOrEnded;
458	}
459
460	bool inGMF() { return S == GMFActive; }
461
462	private:
463	/// Track the transitions into and out of a Global Module Fragment,
464	/// if one is present.
465	GMFState S;
466	};
467
468	TrackGMF TrackGMFState = TrackGMF::BeforeGMFIntroducer;
469
470	/// Track the status of the c++20 module decl.
471	///
472	/// module-declaration:
473	/// 'export'[opt] 'module' module-name module-partition[opt]
474	/// attribute-specifier-seq[opt] ';'
475	///
476	/// module-name:
477	/// module-name-qualifier[opt] identifier
478	///
479	/// module-partition:
480	/// ':' module-name-qualifier[opt] identifier
481	///
482	/// module-name-qualifier:
483	/// identifier '.'
484	/// module-name-qualifier identifier '.'
485	///
486	/// Transition state:
487	///
488	/// NotAModuleDecl --- export ---> FoundExport
489	/// NotAModuleDecl --- module ---> ImplementationCandidate
490	/// FoundExport --- module ---> InterfaceCandidate
491	/// ImplementationCandidate --- Identifier ---> ImplementationCandidate
492	/// ImplementationCandidate --- period ---> ImplementationCandidate
493	/// ImplementationCandidate --- colon ---> ImplementationCandidate
494	/// InterfaceCandidate --- Identifier ---> InterfaceCandidate
495	/// InterfaceCandidate --- period ---> InterfaceCandidate
496	/// InterfaceCandidate --- colon ---> InterfaceCandidate
497	/// ImplementationCandidate --- Semi ---> NamedModuleImplementation
498	/// NamedModuleInterface --- Semi ---> NamedModuleInterface
499	/// NamedModuleImplementation --- Anything ---> NamedModuleImplementation
500	/// NamedModuleInterface --- Anything ---> NamedModuleInterface
501	///
502	/// FIXME: We haven't handle attribute-specifier-seq here. It may not be bad
503	/// soon since we don't support any module attributes yet.
504	class ModuleDeclSeq {
505	enum ModuleDeclState : int {
506	NotAModuleDecl,
507	FoundExport,
508	InterfaceCandidate,
509	ImplementationCandidate,
510	NamedModuleInterface,
511	NamedModuleImplementation,
512	};
513
514	public:
515	ModuleDeclSeq() = default;
516
517	void handleExport() {
518	if (State == NotAModuleDecl)
519	State = FoundExport;
520	else if (!isNamedModule())
521	reset();
522	}
523
524	void handleModule() {
525	if (State == FoundExport)
526	State = InterfaceCandidate;
527	else if (State == NotAModuleDecl)
528	State = ImplementationCandidate;
529	else if (!isNamedModule())
530	reset();
531	}
532
533	void handleIdentifier(IdentifierInfo *Identifier) {
534	if (isModuleCandidate() && Identifier)
535	Name += Identifier->getName().str();
536	else if (!isNamedModule())
537	reset();
538	}
539
540	void handleColon() {
541	if (isModuleCandidate())
542	Name += ":";
543	else if (!isNamedModule())
544	reset();
545	}
546
547	void handlePeriod() {
548	if (isModuleCandidate())
549	Name += ".";
550	else if (!isNamedModule())
551	reset();
552	}
553
554	void handleSemi() {
555	if (!Name.empty() && isModuleCandidate()) {
556	if (State == InterfaceCandidate)
557	State = NamedModuleInterface;
558	else if (State == ImplementationCandidate)
559	State = NamedModuleImplementation;
560	else
561	llvm_unreachable("Unimaged ModuleDeclState.");
562	} else if (!isNamedModule())
563	reset();
564	}
565
566	void handleMisc() {
567	if (!isNamedModule())
568	reset();
569	}
570
571	bool isModuleCandidate() const {
572	return State == InterfaceCandidate || State == ImplementationCandidate;
573	}
574
575	bool isNamedModule() const {
576	return State == NamedModuleInterface ||
577	State == NamedModuleImplementation;
578	}
579
580	bool isNamedInterface() const { return State == NamedModuleInterface; }
581
582	bool isImplementationUnit() const {
583	return State == NamedModuleImplementation && !getName().contains(C: ':');
584	}
585
586	StringRef getName() const {
587	assert(isNamedModule() && "Can't get name from a non named module");
588	return Name;
589	}
590
591	StringRef getPrimaryName() const {
592	assert(isNamedModule() && "Can't get name from a non named module");
593	return getName().split(Separator: ':').first;
594	}
595
596	void reset() {
597	Name.clear();
598	State = NotAModuleDecl;
599	}
600
601	private:
602	ModuleDeclState State = NotAModuleDecl;
603	std::string Name;
604	};
605
606	ModuleDeclSeq ModuleDeclState;
607
608	/// Whether the module import expects an identifier next. Otherwise,
609	/// it expects a '.' or ';'.
610	bool ModuleImportExpectsIdentifier = false;
611
612	/// The identifier and source location of the currently-active
613	/// \#pragma clang arc_cf_code_audited begin.
614	std::pair<IdentifierInfo *, SourceLocation> PragmaARCCFCodeAuditedInfo;
615
616	/// The source location of the currently-active
617	/// \#pragma clang assume_nonnull begin.
618	SourceLocation PragmaAssumeNonNullLoc;
619
620	/// Set only for preambles which end with an active
621	/// \#pragma clang assume_nonnull begin.
622	///
623	/// When the preamble is loaded into the main file,
624	/// `PragmaAssumeNonNullLoc` will be set to this to
625	/// replay the unterminated assume_nonnull.
626	SourceLocation PreambleRecordedPragmaAssumeNonNullLoc;
627
628	/// True if we hit the code-completion point.
629	bool CodeCompletionReached = false;
630
631	/// The code completion token containing the information
632	/// on the stem that is to be code completed.
633	IdentifierInfo *CodeCompletionII = nullptr;
634
635	/// Range for the code completion token.
636	SourceRange CodeCompletionTokenRange;
637
638	/// The directory that the main file should be considered to occupy,
639	/// if it does not correspond to a real file (as happens when building a
640	/// module).
641	OptionalDirectoryEntryRef MainFileDir;
642
643	/// The number of bytes that we will initially skip when entering the
644	/// main file, along with a flag that indicates whether skipping this number
645	/// of bytes will place the lexer at the start of a line.
646	///
647	/// This is used when loading a precompiled preamble.
648	std::pair<int, bool> SkipMainFilePreamble;
649
650	/// Whether we hit an error due to reaching max allowed include depth. Allows
651	/// to avoid hitting the same error over and over again.
652	bool HasReachedMaxIncludeDepth = false;
653
654	/// The number of currently-active calls to Lex.
655	///
656	/// Lex is reentrant, and asking for an (end-of-phase-4) token can often
657	/// require asking for multiple additional tokens. This counter makes it
658	/// possible for Lex to detect whether it's producing a token for the end
659	/// of phase 4 of translation or for some other situation.
660	unsigned LexLevel = 0;
661
662	/// The number of (LexLevel 0) preprocessor tokens.
663	unsigned TokenCount = 0;
664
665	/// Preprocess every token regardless of LexLevel.
666	bool PreprocessToken = false;
667
668	/// The maximum number of (LexLevel 0) tokens before issuing a -Wmax-tokens
669	/// warning, or zero for unlimited.
670	unsigned MaxTokens = 0;
671	SourceLocation MaxTokensOverrideLoc;
672
673	public:
674	struct PreambleSkipInfo {
675	SourceLocation HashTokenLoc;
676	SourceLocation IfTokenLoc;
677	bool FoundNonSkipPortion;
678	bool FoundElse;
679	SourceLocation ElseLoc;
680
681	PreambleSkipInfo(SourceLocation HashTokenLoc, SourceLocation IfTokenLoc,
682	bool FoundNonSkipPortion, bool FoundElse,
683	SourceLocation ElseLoc)
684	: HashTokenLoc(HashTokenLoc), IfTokenLoc(IfTokenLoc),
685	FoundNonSkipPortion(FoundNonSkipPortion), FoundElse(FoundElse),
686	ElseLoc(ElseLoc) {}
687	};
688
689	using IncludedFilesSet = llvm::DenseSet<const FileEntry *>;
690
691	private:
692	friend class ASTReader;
693	friend class MacroArgs;
694
695	class PreambleConditionalStackStore {
696	enum State {
697	Off = 0,
698	Recording = 1,
699	Replaying = 2,
700	};
701
702	public:
703	PreambleConditionalStackStore() = default;
704
705	void startRecording() { ConditionalStackState = Recording; }
706	void startReplaying() { ConditionalStackState = Replaying; }
707	bool isRecording() const { return ConditionalStackState == Recording; }
708	bool isReplaying() const { return ConditionalStackState == Replaying; }
709
710	ArrayRef<PPConditionalInfo> getStack() const {
711	return ConditionalStack;
712	}
713
714	void doneReplaying() {
715	ConditionalStack.clear();
716	ConditionalStackState = Off;
717	}
718
719	void setStack(ArrayRef<PPConditionalInfo> s) {
720	if (!isRecording() && !isReplaying())
721	return;
722	ConditionalStack.clear();
723	ConditionalStack.append(in_start: s.begin(), in_end: s.end());
724	}
725
726	bool hasRecordedPreamble() const { return !ConditionalStack.empty(); }
727
728	bool reachedEOFWhileSkipping() const { return SkipInfo.has_value(); }
729
730	void clearSkipInfo() { SkipInfo.reset(); }
731
732	std::optional<PreambleSkipInfo> SkipInfo;
733
734	private:
735	SmallVector<PPConditionalInfo, 4> ConditionalStack;
736	State ConditionalStackState = Off;
737	} PreambleConditionalStack;
738
739	/// The current top of the stack that we're lexing from if
740	/// not expanding a macro and we are lexing directly from source code.
741	///
742	/// Only one of CurLexer, or CurTokenLexer will be non-null.
743	std::unique_ptr<Lexer> CurLexer;
744
745	/// The current top of the stack that we're lexing from
746	/// if not expanding a macro.
747	///
748	/// This is an alias for CurLexer.
749	PreprocessorLexer *CurPPLexer = nullptr;
750
751	/// Used to find the current FileEntry, if CurLexer is non-null
752	/// and if applicable.
753	///
754	/// This allows us to implement \#include_next and find directory-specific
755	/// properties.
756	ConstSearchDirIterator CurDirLookup = nullptr;
757
758	/// The current macro we are expanding, if we are expanding a macro.
759	///
760	/// One of CurLexer and CurTokenLexer must be null.
761	std::unique_ptr<TokenLexer> CurTokenLexer;
762
763	/// The kind of lexer we're currently working with.
764	typedef bool (*LexerCallback)(Preprocessor &, Token &);
765	LexerCallback CurLexerCallback = &CLK_Lexer;
766
767	/// If the current lexer is for a submodule that is being built, this
768	/// is that submodule.
769	Module *CurLexerSubmodule = nullptr;
770
771	/// Keeps track of the stack of files currently
772	/// \#included, and macros currently being expanded from, not counting
773	/// CurLexer/CurTokenLexer.
774	struct IncludeStackInfo {
775	LexerCallback CurLexerCallback;
776	Module *TheSubmodule;
777	std::unique_ptr<Lexer> TheLexer;
778	PreprocessorLexer *ThePPLexer;
779	std::unique_ptr<TokenLexer> TheTokenLexer;
780	ConstSearchDirIterator TheDirLookup;
781
782	// The following constructors are completely useless copies of the default
783	// versions, only needed to pacify MSVC.
784	IncludeStackInfo(LexerCallback CurLexerCallback, Module *TheSubmodule,
785	std::unique_ptr<Lexer> &&TheLexer,
786	PreprocessorLexer *ThePPLexer,
787	std::unique_ptr<TokenLexer> &&TheTokenLexer,
788	ConstSearchDirIterator TheDirLookup)
789	: CurLexerCallback(std::move(CurLexerCallback)),
790	TheSubmodule(std::move(TheSubmodule)), TheLexer(std::move(TheLexer)),
791	ThePPLexer(std::move(ThePPLexer)),
792	TheTokenLexer(std::move(TheTokenLexer)),
793	TheDirLookup(std::move(TheDirLookup)) {}
794	};
795	std::vector<IncludeStackInfo> IncludeMacroStack;
796
797	/// Actions invoked when some preprocessor activity is
798	/// encountered (e.g. a file is \#included, etc).
799	std::unique_ptr<PPCallbacks> Callbacks;
800
801	struct MacroExpandsInfo {
802	Token Tok;
803	MacroDefinition MD;
804	SourceRange Range;
805
806	MacroExpandsInfo(Token Tok, MacroDefinition MD, SourceRange Range)
807	: Tok(Tok), MD(MD), Range(Range) {}
808	};
809	SmallVector<MacroExpandsInfo, 2> DelayedMacroExpandsCallbacks;
810
811	/// Information about a name that has been used to define a module macro.
812	struct ModuleMacroInfo {
813	/// The most recent macro directive for this identifier.
814	MacroDirective *MD;
815
816	/// The active module macros for this identifier.
817	llvm::TinyPtrVector<ModuleMacro *> ActiveModuleMacros;
818
819	/// The generation number at which we last updated ActiveModuleMacros.
820	/// \see Preprocessor::VisibleModules.
821	unsigned ActiveModuleMacrosGeneration = 0;
822
823	/// Whether this macro name is ambiguous.
824	bool IsAmbiguous = false;
825
826	/// The module macros that are overridden by this macro.
827	llvm::TinyPtrVector<ModuleMacro *> OverriddenMacros;
828
829	ModuleMacroInfo(MacroDirective *MD) : MD(MD) {}
830	};
831
832	/// The state of a macro for an identifier.
833	class MacroState {
834	mutable llvm::PointerUnion<MacroDirective *, ModuleMacroInfo *> State;
835
836	ModuleMacroInfo *getModuleInfo(Preprocessor &PP,
837	const IdentifierInfo *II) const {
838	if (II->isOutOfDate())
839	PP.updateOutOfDateIdentifier(II&: const_cast<IdentifierInfo&>(*II));
840	// FIXME: Find a spare bit on IdentifierInfo and store a
841	// HasModuleMacros flag.
842	if (!II->hasMacroDefinition() ||
843	(!PP.getLangOpts().Modules &&
844	!PP.getLangOpts().ModulesLocalVisibility) ||
845	!PP.CurSubmoduleState->VisibleModules.getGeneration())
846	return nullptr;
847
848	auto *Info = State.dyn_cast<ModuleMacroInfo*>();
849	if (!Info) {
850	Info = new (PP.getPreprocessorAllocator())
851	ModuleMacroInfo(State.get<MacroDirective *>());
852	State = Info;
853	}
854
855	if (PP.CurSubmoduleState->VisibleModules.getGeneration() !=
856	Info->ActiveModuleMacrosGeneration)
857	PP.updateModuleMacroInfo(II, Info&: *Info);
858	return Info;
859	}
860
861	public:
862	MacroState() : MacroState(nullptr) {}
863	MacroState(MacroDirective *MD) : State(MD) {}
864
865	MacroState(MacroState &&O) noexcept : State(O.State) {
866	O.State = (MacroDirective *)nullptr;
867	}
868
869	MacroState &operator=(MacroState &&O) noexcept {
870	auto S = O.State;
871	O.State = (MacroDirective *)nullptr;
872	State = S;
873	return *this;
874	}
875
876	~MacroState() {
877	if (auto *Info = State.dyn_cast<ModuleMacroInfo*>())
878	Info->~ModuleMacroInfo();
879	}
880
881	MacroDirective *getLatest() const {
882	if (auto *Info = State.dyn_cast<ModuleMacroInfo*>())
883	return Info->MD;
884	return State.get<MacroDirective*>();
885	}
886
887	void setLatest(MacroDirective *MD) {
888	if (auto *Info = State.dyn_cast<ModuleMacroInfo*>())
889	Info->MD = MD;
890	else
891	State = MD;
892	}
893
894	bool isAmbiguous(Preprocessor &PP, const IdentifierInfo *II) const {
895	auto *Info = getModuleInfo(PP, II);
896	return Info ? Info->IsAmbiguous : false;
897	}
898
899	ArrayRef<ModuleMacro *>
900	getActiveModuleMacros(Preprocessor &PP, const IdentifierInfo *II) const {
901	if (auto *Info = getModuleInfo(PP, II))
902	return Info->ActiveModuleMacros;
903	return std::nullopt;
904	}
905
906	MacroDirective::DefInfo findDirectiveAtLoc(SourceLocation Loc,
907	SourceManager &SourceMgr) const {
908	// FIXME: Incorporate module macros into the result of this.
909	if (auto *Latest = getLatest())
910	return Latest->findDirectiveAtLoc(L: Loc, SM: SourceMgr);
911	return {};
912	}
913
914	void overrideActiveModuleMacros(Preprocessor &PP, IdentifierInfo *II) {
915	if (auto *Info = getModuleInfo(PP, II)) {
916	Info->OverriddenMacros.insert(I: Info->OverriddenMacros.end(),
917	From: Info->ActiveModuleMacros.begin(),
918	To: Info->ActiveModuleMacros.end());
919	Info->ActiveModuleMacros.clear();
920	Info->IsAmbiguous = false;
921	}
922	}
923
924	ArrayRef<ModuleMacro*> getOverriddenMacros() const {
925	if (auto *Info = State.dyn_cast<ModuleMacroInfo*>())
926	return Info->OverriddenMacros;
927	return std::nullopt;
928	}
929
930	void setOverriddenMacros(Preprocessor &PP,
931	ArrayRef<ModuleMacro *> Overrides) {
932	auto *Info = State.dyn_cast<ModuleMacroInfo*>();
933	if (!Info) {
934	if (Overrides.empty())
935	return;
936	Info = new (PP.getPreprocessorAllocator())
937	ModuleMacroInfo(State.get<MacroDirective *>());
938	State = Info;
939	}
940	Info->OverriddenMacros.clear();
941	Info->OverriddenMacros.insert(I: Info->OverriddenMacros.end(),
942	From: Overrides.begin(), To: Overrides.end());
943	Info->ActiveModuleMacrosGeneration = 0;
944	}
945	};
946
947	/// For each IdentifierInfo that was associated with a macro, we
948	/// keep a mapping to the history of all macro definitions and #undefs in
949	/// the reverse order (the latest one is in the head of the list).
950	///
951	/// This mapping lives within the \p CurSubmoduleState.
952	using MacroMap = llvm::DenseMap<const IdentifierInfo *, MacroState>;
953
954	struct SubmoduleState;
955
956	/// Information about a submodule that we're currently building.
957	struct BuildingSubmoduleInfo {
958	/// The module that we are building.
959	Module *M;
960
961	/// The location at which the module was included.
962	SourceLocation ImportLoc;
963
964	/// Whether we entered this submodule via a pragma.
965	bool IsPragma;
966
967	/// The previous SubmoduleState.
968	SubmoduleState *OuterSubmoduleState;
969
970	/// The number of pending module macro names when we started building this.
971	unsigned OuterPendingModuleMacroNames;
972
973	BuildingSubmoduleInfo(Module *M, SourceLocation ImportLoc, bool IsPragma,
974	SubmoduleState *OuterSubmoduleState,
975	unsigned OuterPendingModuleMacroNames)
976	: M(M), ImportLoc(ImportLoc), IsPragma(IsPragma),
977	OuterSubmoduleState(OuterSubmoduleState),
978	OuterPendingModuleMacroNames(OuterPendingModuleMacroNames) {}
979	};
980	SmallVector<BuildingSubmoduleInfo, 8> BuildingSubmoduleStack;
981
982	/// Information about a submodule's preprocessor state.
983	struct SubmoduleState {
984	/// The macros for the submodule.
985	MacroMap Macros;
986
987	/// The set of modules that are visible within the submodule.
988	VisibleModuleSet VisibleModules;
989
990	// FIXME: CounterValue?
991	// FIXME: PragmaPushMacroInfo?
992	};
993	std::map<Module *, SubmoduleState> Submodules;
994
995	/// The preprocessor state for preprocessing outside of any submodule.
996	SubmoduleState NullSubmoduleState;
997
998	/// The current submodule state. Will be \p NullSubmoduleState if we're not
999	/// in a submodule.
1000	SubmoduleState *CurSubmoduleState;
1001
1002	/// The files that have been included.
1003	IncludedFilesSet IncludedFiles;
1004
1005	/// The set of top-level modules that affected preprocessing, but were not
1006	/// imported.
1007	llvm::SmallSetVector<Module *, 2> AffectingClangModules;
1008
1009	/// The set of known macros exported from modules.
1010	llvm::FoldingSet<ModuleMacro> ModuleMacros;
1011
1012	/// The names of potential module macros that we've not yet processed.
1013	llvm::SmallVector<const IdentifierInfo *, 32> PendingModuleMacroNames;
1014
1015	/// The list of module macros, for each identifier, that are not overridden by
1016	/// any other module macro.
1017	llvm::DenseMap<const IdentifierInfo *, llvm::TinyPtrVector<ModuleMacro *>>
1018	LeafModuleMacros;
1019
1020	/// Macros that we want to warn because they are not used at the end
1021	/// of the translation unit.
1022	///
1023	/// We store just their SourceLocations instead of
1024	/// something like MacroInfo*. The benefit of this is that when we are
1025	/// deserializing from PCH, we don't need to deserialize identifier & macros
1026	/// just so that we can report that they are unused, we just warn using
1027	/// the SourceLocations of this set (that will be filled by the ASTReader).
1028	using WarnUnusedMacroLocsTy = llvm::SmallDenseSet<SourceLocation, 32>;
1029	WarnUnusedMacroLocsTy WarnUnusedMacroLocs;
1030
1031	/// This is a pair of an optional message and source location used for pragmas
1032	/// that annotate macros like pragma clang restrict_expansion and pragma clang
1033	/// deprecated. This pair stores the optional message and the location of the
1034	/// annotation pragma for use producing diagnostics and notes.
1035	using MsgLocationPair = std::pair<std::string, SourceLocation>;
1036
1037	struct MacroAnnotationInfo {
1038	SourceLocation Location;
1039	std::string Message;
1040	};
1041
1042	struct MacroAnnotations {
1043	std::optional<MacroAnnotationInfo> DeprecationInfo;
1044	std::optional<MacroAnnotationInfo> RestrictExpansionInfo;
1045	std::optional<SourceLocation> FinalAnnotationLoc;
1046
1047	static MacroAnnotations makeDeprecation(SourceLocation Loc,
1048	std::string Msg) {
1049	return MacroAnnotations{.DeprecationInfo: MacroAnnotationInfo{.Location: Loc, .Message: std::move(Msg)},
1050	.RestrictExpansionInfo: std::nullopt, .FinalAnnotationLoc: std::nullopt};
1051	}
1052
1053	static MacroAnnotations makeRestrictExpansion(SourceLocation Loc,
1054	std::string Msg) {
1055	return MacroAnnotations{
1056	.DeprecationInfo: std::nullopt, .RestrictExpansionInfo: MacroAnnotationInfo{.Location: Loc, .Message: std::move(Msg)}, .FinalAnnotationLoc: std::nullopt};
1057	}
1058
1059	static MacroAnnotations makeFinal(SourceLocation Loc) {
1060	return MacroAnnotations{.DeprecationInfo: std::nullopt, .RestrictExpansionInfo: std::nullopt, .FinalAnnotationLoc: Loc};
1061	}
1062	};
1063
1064	/// Warning information for macro annotations.
1065	llvm::DenseMap<const IdentifierInfo *, MacroAnnotations> AnnotationInfos;
1066
1067	/// A "freelist" of MacroArg objects that can be
1068	/// reused for quick allocation.
1069	MacroArgs *MacroArgCache = nullptr;
1070
1071	/// For each IdentifierInfo used in a \#pragma push_macro directive,
1072	/// we keep a MacroInfo stack used to restore the previous macro value.
1073	llvm::DenseMap<IdentifierInfo *, std::vector<MacroInfo *>>
1074	PragmaPushMacroInfo;
1075
1076	// Various statistics we track for performance analysis.
1077	unsigned NumDirectives = 0;
1078	unsigned NumDefined = 0;
1079	unsigned NumUndefined = 0;
1080	unsigned NumPragma = 0;
1081	unsigned NumIf = 0;
1082	unsigned NumElse = 0;
1083	unsigned NumEndif = 0;
1084	unsigned NumEnteredSourceFiles = 0;
1085	unsigned MaxIncludeStackDepth = 0;
1086	unsigned NumMacroExpanded = 0;
1087	unsigned NumFnMacroExpanded = 0;
1088	unsigned NumBuiltinMacroExpanded = 0;
1089	unsigned NumFastMacroExpanded = 0;
1090	unsigned NumTokenPaste = 0;
1091	unsigned NumFastTokenPaste = 0;
1092	unsigned NumSkipped = 0;
1093
1094	/// The predefined macros that preprocessor should use from the
1095	/// command line etc.
1096	std::string Predefines;
1097
1098	/// The file ID for the preprocessor predefines.
1099	FileID PredefinesFileID;
1100
1101	/// The file ID for the PCH through header.
1102	FileID PCHThroughHeaderFileID;
1103
1104	/// Whether tokens are being skipped until a #pragma hdrstop is seen.
1105	bool SkippingUntilPragmaHdrStop = false;
1106
1107	/// Whether tokens are being skipped until the through header is seen.
1108	bool SkippingUntilPCHThroughHeader = false;
1109
1110	/// \{
1111	/// Cache of macro expanders to reduce malloc traffic.
1112	enum { TokenLexerCacheSize = 8 };
1113	unsigned NumCachedTokenLexers;
1114	std::unique_ptr<TokenLexer> TokenLexerCache[TokenLexerCacheSize];
1115	/// \}
1116
1117	/// Keeps macro expanded tokens for TokenLexers.
1118	//
1119	/// Works like a stack; a TokenLexer adds the macro expanded tokens that is
1120	/// going to lex in the cache and when it finishes the tokens are removed
1121	/// from the end of the cache.
1122	SmallVector<Token, 16> MacroExpandedTokens;
1123	std::vector<std::pair<TokenLexer *, size_t>> MacroExpandingLexersStack;
1124
1125	/// A record of the macro definitions and expansions that
1126	/// occurred during preprocessing.
1127	///
1128	/// This is an optional side structure that can be enabled with
1129	/// \c createPreprocessingRecord() prior to preprocessing.
1130	PreprocessingRecord *Record = nullptr;
1131
1132	/// Cached tokens state.
1133	using CachedTokensTy = SmallVector<Token, 1>;
1134
1135	/// Cached tokens are stored here when we do backtracking or
1136	/// lookahead. They are "lexed" by the CachingLex() method.
1137	CachedTokensTy CachedTokens;
1138
1139	/// The position of the cached token that CachingLex() should
1140	/// "lex" next.
1141	///
1142	/// If it points beyond the CachedTokens vector, it means that a normal
1143	/// Lex() should be invoked.
1144	CachedTokensTy::size_type CachedLexPos = 0;
1145
1146	/// Stack of backtrack positions, allowing nested backtracks.
1147	///
1148	/// The EnableBacktrackAtThisPos() method pushes a position to
1149	/// indicate where CachedLexPos should be set when the BackTrack() method is
1150	/// invoked (at which point the last position is popped).
1151	std::vector<CachedTokensTy::size_type> BacktrackPositions;
1152
1153	/// True if \p Preprocessor::SkipExcludedConditionalBlock() is running.
1154	/// This is used to guard against calling this function recursively.
1155	///
1156	/// See comments at the use-site for more context about why it is needed.
1157	bool SkippingExcludedConditionalBlock = false;
1158
1159	/// Keeps track of skipped range mappings that were recorded while skipping
1160	/// excluded conditional directives. It maps the source buffer pointer at
1161	/// the beginning of a skipped block, to the number of bytes that should be
1162	/// skipped.
1163	llvm::DenseMap<const char *, unsigned> RecordedSkippedRanges;
1164
1165	void updateOutOfDateIdentifier(IdentifierInfo &II) const;
1166
1167	public:
1168	Preprocessor(std::shared_ptr<PreprocessorOptions> PPOpts,
1169	DiagnosticsEngine &diags, const LangOptions &LangOpts,
1170	SourceManager &SM, HeaderSearch &Headers,
1171	ModuleLoader &TheModuleLoader,
1172	IdentifierInfoLookup *IILookup = nullptr,
1173	bool OwnsHeaderSearch = false,
1174	TranslationUnitKind TUKind = TU_Complete);
1175
1176	~Preprocessor();
1177
1178	/// Initialize the preprocessor using information about the target.
1179	///
1180	/// \param Target is owned by the caller and must remain valid for the
1181	/// lifetime of the preprocessor.
1182	/// \param AuxTarget is owned by the caller and must remain valid for
1183	/// the lifetime of the preprocessor.
1184	void Initialize(const TargetInfo &Target,
1185	const TargetInfo *AuxTarget = nullptr);
1186
1187	/// Initialize the preprocessor to parse a model file
1188	///
1189	/// To parse model files the preprocessor of the original source is reused to
1190	/// preserver the identifier table. However to avoid some duplicate
1191	/// information in the preprocessor some cleanup is needed before it is used
1192	/// to parse model files. This method does that cleanup.
1193	void InitializeForModelFile();
1194
1195	/// Cleanup after model file parsing
1196	void FinalizeForModelFile();
1197
1198	/// Retrieve the preprocessor options used to initialize this
1199	/// preprocessor.
1200	PreprocessorOptions &getPreprocessorOpts() const { return *PPOpts; }
1201
1202	DiagnosticsEngine &getDiagnostics() const { return *Diags; }
1203	void setDiagnostics(DiagnosticsEngine &D) { Diags = &D; }
1204
1205	const LangOptions &getLangOpts() const { return LangOpts; }
1206	const TargetInfo &getTargetInfo() const { return *Target; }
1207	const TargetInfo *getAuxTargetInfo() const { return AuxTarget; }
1208	FileManager &getFileManager() const { return FileMgr; }
1209	SourceManager &getSourceManager() const { return SourceMgr; }
1210	HeaderSearch &getHeaderSearchInfo() const { return HeaderInfo; }
1211
1212	IdentifierTable &getIdentifierTable() { return Identifiers; }
1213	const IdentifierTable &getIdentifierTable() const { return Identifiers; }
1214	SelectorTable &getSelectorTable() { return Selectors; }
1215	Builtin::Context &getBuiltinInfo() { return *BuiltinInfo; }
1216	llvm::BumpPtrAllocator &getPreprocessorAllocator() { return BP; }
1217
1218	void setExternalSource(ExternalPreprocessorSource *Source) {
1219	ExternalSource = Source;
1220	}
1221
1222	ExternalPreprocessorSource *getExternalSource() const {
1223	return ExternalSource;
1224	}
1225
1226	/// Retrieve the module loader associated with this preprocessor.
1227	ModuleLoader &getModuleLoader() const { return TheModuleLoader; }
1228
1229	bool hadModuleLoaderFatalFailure() const {
1230	return TheModuleLoader.HadFatalFailure;
1231	}
1232
1233	/// Retrieve the number of Directives that have been processed by the
1234	/// Preprocessor.
1235	unsigned getNumDirectives() const {
1236	return NumDirectives;
1237	}
1238
1239	/// True if we are currently preprocessing a #if or #elif directive
1240	bool isParsingIfOrElifDirective() const {
1241	return ParsingIfOrElifDirective;
1242	}
1243
1244	/// Control whether the preprocessor retains comments in output.
1245	void SetCommentRetentionState(bool KeepComments, bool KeepMacroComments) {
1246	this->KeepComments = KeepComments | KeepMacroComments;
1247	this->KeepMacroComments = KeepMacroComments;
1248	}
1249
1250	bool getCommentRetentionState() const { return KeepComments; }
1251
1252	void setPragmasEnabled(bool Enabled) { PragmasEnabled = Enabled; }
1253	bool getPragmasEnabled() const { return PragmasEnabled; }
1254
1255	void SetSuppressIncludeNotFoundError(bool Suppress) {
1256	SuppressIncludeNotFoundError = Suppress;
1257	}
1258
1259	bool GetSuppressIncludeNotFoundError() {
1260	return SuppressIncludeNotFoundError;
1261	}
1262
1263	/// Sets whether the preprocessor is responsible for producing output or if
1264	/// it is producing tokens to be consumed by Parse and Sema.
1265	void setPreprocessedOutput(bool IsPreprocessedOutput) {
1266	PreprocessedOutput = IsPreprocessedOutput;
1267	}
1268
1269	/// Returns true if the preprocessor is responsible for generating output,
1270	/// false if it is producing tokens to be consumed by Parse and Sema.
1271	bool isPreprocessedOutput() const { return PreprocessedOutput; }
1272
1273	/// Return true if we are lexing directly from the specified lexer.
1274	bool isCurrentLexer(const PreprocessorLexer *L) const {
1275	return CurPPLexer == L;
1276	}
1277
1278	/// Return the current lexer being lexed from.
1279	///
1280	/// Note that this ignores any potentially active macro expansions and _Pragma
1281	/// expansions going on at the time.
1282	PreprocessorLexer *getCurrentLexer() const { return CurPPLexer; }
1283
1284	/// Return the current file lexer being lexed from.
1285	///
1286	/// Note that this ignores any potentially active macro expansions and _Pragma
1287	/// expansions going on at the time.
1288	PreprocessorLexer *getCurrentFileLexer() const;
1289
1290	/// Return the submodule owning the file being lexed. This may not be
1291	/// the current module if we have changed modules since entering the file.
1292	Module *getCurrentLexerSubmodule() const { return CurLexerSubmodule; }
1293
1294	/// Returns the FileID for the preprocessor predefines.
1295	FileID getPredefinesFileID() const { return PredefinesFileID; }
1296
1297	/// \{
1298	/// Accessors for preprocessor callbacks.
1299	///
1300	/// Note that this class takes ownership of any PPCallbacks object given to
1301	/// it.
1302	PPCallbacks *getPPCallbacks() const { return Callbacks.get(); }
1303	void addPPCallbacks(std::unique_ptr<PPCallbacks> C) {
1304	if (Callbacks)
1305	C = std::make_unique<PPChainedCallbacks>(args: std::move(C),
1306	args: std::move(Callbacks));
1307	Callbacks = std::move(C);
1308	}
1309	/// \}
1310
1311	/// Get the number of tokens processed so far.
1312	unsigned getTokenCount() const { return TokenCount; }
1313
1314	/// Get the max number of tokens before issuing a -Wmax-tokens warning.
1315	unsigned getMaxTokens() const { return MaxTokens; }
1316
1317	void overrideMaxTokens(unsigned Value, SourceLocation Loc) {
1318	MaxTokens = Value;
1319	MaxTokensOverrideLoc = Loc;
1320	};
1321
1322	SourceLocation getMaxTokensOverrideLoc() const { return MaxTokensOverrideLoc; }
1323
1324	/// Register a function that would be called on each token in the final
1325	/// expanded token stream.
1326	/// This also reports annotation tokens produced by the parser.
1327	void setTokenWatcher(llvm::unique_function<void(const clang::Token &)> F) {
1328	OnToken = std::move(F);
1329	}
1330
1331	void setPreprocessToken(bool Preprocess) { PreprocessToken = Preprocess; }
1332
1333	bool isMacroDefined(StringRef Id) {
1334	return isMacroDefined(&Identifiers.get(Id));
1335	}
1336	bool isMacroDefined(const IdentifierInfo *II) {
1337	return II->hasMacroDefinition() &&
1338	(!getLangOpts().Modules || (bool)getMacroDefinition(II));
1339	}
1340
1341	/// Determine whether II is defined as a macro within the module M,
1342	/// if that is a module that we've already preprocessed. Does not check for
1343	/// macros imported into M.
1344	bool isMacroDefinedInLocalModule(const IdentifierInfo *II, Module *M) {
1345	if (!II->hasMacroDefinition())
1346	return false;
1347	auto I = Submodules.find(x: M);
1348	if (I == Submodules.end())
1349	return false;
1350	auto J = I->second.Macros.find(Val: II);
1351	if (J == I->second.Macros.end())
1352	return false;
1353	auto *MD = J->second.getLatest();
1354	return MD && MD->isDefined();
1355	}
1356
1357	MacroDefinition getMacroDefinition(const IdentifierInfo *II) {
1358	if (!II->hasMacroDefinition())
1359	return {};
1360
1361	MacroState &S = CurSubmoduleState->Macros[II];
1362	auto *MD = S.getLatest();
1363	while (MD && isa<VisibilityMacroDirective>(Val: MD))
1364	MD = MD->getPrevious();
1365	return MacroDefinition(dyn_cast_or_null<DefMacroDirective>(Val: MD),
1366	S.getActiveModuleMacros(PP&: *this, II),
1367	S.isAmbiguous(PP&: *this, II));
1368	}
1369
1370	MacroDefinition getMacroDefinitionAtLoc(const IdentifierInfo *II,
1371	SourceLocation Loc) {
1372	if (!II->hadMacroDefinition())
1373	return {};
1374
1375	MacroState &S = CurSubmoduleState->Macros[II];
1376	MacroDirective::DefInfo DI;
1377	if (auto *MD = S.getLatest())
1378	DI = MD->findDirectiveAtLoc(L: Loc, SM: getSourceManager());
1379	// FIXME: Compute the set of active module macros at the specified location.
1380	return MacroDefinition(DI.getDirective(),
1381	S.getActiveModuleMacros(PP&: *this, II),
1382	S.isAmbiguous(PP&: *this, II));
1383	}
1384
1385	/// Given an identifier, return its latest non-imported MacroDirective
1386	/// if it is \#define'd and not \#undef'd, or null if it isn't \#define'd.
1387	MacroDirective *getLocalMacroDirective(const IdentifierInfo *II) const {
1388	if (!II->hasMacroDefinition())
1389	return nullptr;
1390
1391	auto *MD = getLocalMacroDirectiveHistory(II);
1392	if (!MD || MD->getDefinition().isUndefined())
1393	return nullptr;
1394
1395	return MD;
1396	}
1397
1398	const MacroInfo *getMacroInfo(const IdentifierInfo *II) const {
1399	return const_cast<Preprocessor*>(this)->getMacroInfo(II);
1400	}
1401
1402	MacroInfo *getMacroInfo(const IdentifierInfo *II) {
1403	if (!II->hasMacroDefinition())
1404	return nullptr;
1405	if (auto MD = getMacroDefinition(II))
1406	return MD.getMacroInfo();
1407	return nullptr;
1408	}
1409
1410	/// Given an identifier, return the latest non-imported macro
1411	/// directive for that identifier.
1412	///
1413	/// One can iterate over all previous macro directives from the most recent
1414	/// one.
1415	MacroDirective *getLocalMacroDirectiveHistory(const IdentifierInfo *II) const;
1416
1417	/// Add a directive to the macro directive history for this identifier.
1418	void appendMacroDirective(IdentifierInfo *II, MacroDirective *MD);
1419	DefMacroDirective *appendDefMacroDirective(IdentifierInfo *II, MacroInfo *MI,
1420	SourceLocation Loc) {
1421	DefMacroDirective *MD = AllocateDefMacroDirective(MI, Loc);
1422	appendMacroDirective(II, MD);
1423	return MD;
1424	}
1425	DefMacroDirective *appendDefMacroDirective(IdentifierInfo *II,
1426	MacroInfo *MI) {
1427	return appendDefMacroDirective(II, MI, Loc: MI->getDefinitionLoc());
1428	}
1429
1430	/// Set a MacroDirective that was loaded from a PCH file.
1431	void setLoadedMacroDirective(IdentifierInfo *II, MacroDirective *ED,
1432	MacroDirective *MD);
1433
1434	/// Register an exported macro for a module and identifier.
1435	ModuleMacro *addModuleMacro(Module *Mod, IdentifierInfo *II, MacroInfo *Macro,
1436	ArrayRef<ModuleMacro *> Overrides, bool &IsNew);
1437	ModuleMacro *getModuleMacro(Module *Mod, const IdentifierInfo *II);
1438
1439	/// Get the list of leaf (non-overridden) module macros for a name.
1440	ArrayRef<ModuleMacro*> getLeafModuleMacros(const IdentifierInfo *II) const {
1441	if (II->isOutOfDate())
1442	updateOutOfDateIdentifier(II&: const_cast<IdentifierInfo&>(*II));
1443	auto I = LeafModuleMacros.find(Val: II);
1444	if (I != LeafModuleMacros.end())
1445	return I->second;
1446	return std::nullopt;
1447	}
1448
1449	/// Get the list of submodules that we're currently building.
1450	ArrayRef<BuildingSubmoduleInfo> getBuildingSubmodules() const {
1451	return BuildingSubmoduleStack;
1452	}
1453
1454	/// \{
1455	/// Iterators for the macro history table. Currently defined macros have
1456	/// IdentifierInfo::hasMacroDefinition() set and an empty
1457	/// MacroInfo::getUndefLoc() at the head of the list.
1458	using macro_iterator = MacroMap::const_iterator;
1459
1460	macro_iterator macro_begin(bool IncludeExternalMacros = true) const;
1461	macro_iterator macro_end(bool IncludeExternalMacros = true) const;
1462
1463	llvm::iterator_range<macro_iterator>
1464	macros(bool IncludeExternalMacros = true) const {
1465	macro_iterator begin = macro_begin(IncludeExternalMacros);
1466	macro_iterator end = macro_end(IncludeExternalMacros);
1467	return llvm::make_range(x: begin, y: end);
1468	}
1469
1470	/// \}
1471
1472	/// Mark the given clang module as affecting the current clang module or translation unit.
1473	void markClangModuleAsAffecting(Module *M) {
1474	assert(M->isModuleMapModule());
1475	if (!BuildingSubmoduleStack.empty()) {
1476	if (M != BuildingSubmoduleStack.back().M)
1477	BuildingSubmoduleStack.back().M->AffectingClangModules.insert(X: M);
1478	} else {
1479	AffectingClangModules.insert(X: M);
1480	}
1481	}
1482
1483	/// Get the set of top-level clang modules that affected preprocessing, but were not
1484	/// imported.
1485	const llvm::SmallSetVector<Module *, 2> &getAffectingClangModules() const {
1486	return AffectingClangModules;
1487	}
1488
1489	/// Mark the file as included.
1490	/// Returns true if this is the first time the file was included.
1491	bool markIncluded(FileEntryRef File) {
1492	HeaderInfo.getFileInfo(FE: File);
1493	return IncludedFiles.insert(V: File).second;
1494	}
1495
1496	/// Return true if this header has already been included.
1497	bool alreadyIncluded(FileEntryRef File) const {
1498	HeaderInfo.getFileInfo(FE: File);
1499	return IncludedFiles.count(V: File);
1500	}
1501
1502	/// Get the set of included files.
1503	IncludedFilesSet &getIncludedFiles() { return IncludedFiles; }
1504	const IncludedFilesSet &getIncludedFiles() const { return IncludedFiles; }
1505
1506	/// Return the name of the macro defined before \p Loc that has
1507	/// spelling \p Tokens. If there are multiple macros with same spelling,
1508	/// return the last one defined.
1509	StringRef getLastMacroWithSpelling(SourceLocation Loc,
1510	ArrayRef<TokenValue> Tokens) const;
1511
1512	/// Get the predefines for this processor.
1513	/// Used by some third-party tools to inspect and add predefines (see
1514	/// https://github.com/llvm/llvm-project/issues/57483).
1515	const std::string &getPredefines() const { return Predefines; }
1516
1517	/// Set the predefines for this Preprocessor.
1518	///
1519	/// These predefines are automatically injected when parsing the main file.
1520	void setPredefines(std::string P) { Predefines = std::move(P); }
1521
1522	/// Return information about the specified preprocessor
1523	/// identifier token.
1524	IdentifierInfo *getIdentifierInfo(StringRef Name) const {
1525	return &Identifiers.get(Name);
1526	}
1527
1528	/// Add the specified pragma handler to this preprocessor.
1529	///
1530	/// If \p Namespace is non-null, then it is a token required to exist on the
1531	/// pragma line before the pragma string starts, e.g. "STDC" or "GCC".
1532	void AddPragmaHandler(StringRef Namespace, PragmaHandler *Handler);
1533	void AddPragmaHandler(PragmaHandler *Handler) {
1534	AddPragmaHandler(Namespace: StringRef(), Handler);
1535	}
1536
1537	/// Remove the specific pragma handler from this preprocessor.
1538	///
1539	/// If \p Namespace is non-null, then it should be the namespace that
1540	/// \p Handler was added to. It is an error to remove a handler that
1541	/// has not been registered.
1542	void RemovePragmaHandler(StringRef Namespace, PragmaHandler *Handler);
1543	void RemovePragmaHandler(PragmaHandler *Handler) {
1544	RemovePragmaHandler(Namespace: StringRef(), Handler);
1545	}
1546
1547	/// Install empty handlers for all pragmas (making them ignored).
1548	void IgnorePragmas();
1549
1550	/// Set empty line handler.
1551	void setEmptylineHandler(EmptylineHandler *Handler) { Emptyline = Handler; }
1552
1553	EmptylineHandler *getEmptylineHandler() const { return Emptyline; }
1554
1555	/// Add the specified comment handler to the preprocessor.
1556	void addCommentHandler(CommentHandler *Handler);
1557
1558	/// Remove the specified comment handler.
1559	///
1560	/// It is an error to remove a handler that has not been registered.
1561	void removeCommentHandler(CommentHandler *Handler);
1562
1563	/// Set the code completion handler to the given object.
1564	void setCodeCompletionHandler(CodeCompletionHandler &Handler) {
1565	CodeComplete = &Handler;
1566	}
1567
1568	/// Retrieve the current code-completion handler.
1569	CodeCompletionHandler *getCodeCompletionHandler() const {
1570	return CodeComplete;
1571	}
1572
1573	/// Clear out the code completion handler.
1574	void clearCodeCompletionHandler() {
1575	CodeComplete = nullptr;
1576	}
1577
1578	/// Hook used by the lexer to invoke the "included file" code
1579	/// completion point.
1580	void CodeCompleteIncludedFile(llvm::StringRef Dir, bool IsAngled);
1581
1582	/// Hook used by the lexer to invoke the "natural language" code
1583	/// completion point.
1584	void CodeCompleteNaturalLanguage();
1585
1586	/// Set the code completion token for filtering purposes.
1587	void setCodeCompletionIdentifierInfo(IdentifierInfo *Filter) {
1588	CodeCompletionII = Filter;
1589	}
1590
1591	/// Set the code completion token range for detecting replacement range later
1592	/// on.
1593	void setCodeCompletionTokenRange(const SourceLocation Start,
1594	const SourceLocation End) {
1595	CodeCompletionTokenRange = {Start, End};
1596	}
1597	SourceRange getCodeCompletionTokenRange() const {
1598	return CodeCompletionTokenRange;
1599	}
1600
1601	/// Get the code completion token for filtering purposes.
1602	StringRef getCodeCompletionFilter() {
1603	if (CodeCompletionII)
1604	return CodeCompletionII->getName();
1605	return {};
1606	}
1607
1608	/// Retrieve the preprocessing record, or NULL if there is no
1609	/// preprocessing record.
1610	PreprocessingRecord *getPreprocessingRecord() const { return Record; }
1611
1612	/// Create a new preprocessing record, which will keep track of
1613	/// all macro expansions, macro definitions, etc.
1614	void createPreprocessingRecord();
1615
1616	/// Returns true if the FileEntry is the PCH through header.
1617	bool isPCHThroughHeader(const FileEntry *FE);
1618
1619	/// True if creating a PCH with a through header.
1620	bool creatingPCHWithThroughHeader();
1621
1622	/// True if using a PCH with a through header.
1623	bool usingPCHWithThroughHeader();
1624
1625	/// True if creating a PCH with a #pragma hdrstop.
1626	bool creatingPCHWithPragmaHdrStop();
1627
1628	/// True if using a PCH with a #pragma hdrstop.
1629	bool usingPCHWithPragmaHdrStop();
1630
1631	/// Skip tokens until after the #include of the through header or
1632	/// until after a #pragma hdrstop.
1633	void SkipTokensWhileUsingPCH();
1634
1635	/// Process directives while skipping until the through header or
1636	/// #pragma hdrstop is found.
1637	void HandleSkippedDirectiveWhileUsingPCH(Token &Result,
1638	SourceLocation HashLoc);
1639
1640	/// Enter the specified FileID as the main source file,
1641	/// which implicitly adds the builtin defines etc.
1642	void EnterMainSourceFile();
1643
1644	/// Inform the preprocessor callbacks that processing is complete.
1645	void EndSourceFile();
1646
1647	/// Add a source file to the top of the include stack and
1648	/// start lexing tokens from it instead of the current buffer.
1649	///
1650	/// Emits a diagnostic, doesn't enter the file, and returns true on error.
1651	bool EnterSourceFile(FileID FID, ConstSearchDirIterator Dir,
1652	SourceLocation Loc, bool IsFirstIncludeOfFile = true);
1653
1654	/// Add a Macro to the top of the include stack and start lexing
1655	/// tokens from it instead of the current buffer.
1656	///
1657	/// \param Args specifies the tokens input to a function-like macro.
1658	/// \param ILEnd specifies the location of the ')' for a function-like macro
1659	/// or the identifier for an object-like macro.
1660	void EnterMacro(Token &Tok, SourceLocation ILEnd, MacroInfo *Macro,
1661	MacroArgs *Args);
1662
1663	private:
1664	/// Add a "macro" context to the top of the include stack,
1665	/// which will cause the lexer to start returning the specified tokens.
1666	///
1667	/// If \p DisableMacroExpansion is true, tokens lexed from the token stream
1668	/// will not be subject to further macro expansion. Otherwise, these tokens
1669	/// will be re-macro-expanded when/if expansion is enabled.
1670	///
1671	/// If \p OwnsTokens is false, this method assumes that the specified stream
1672	/// of tokens has a permanent owner somewhere, so they do not need to be
1673	/// copied. If it is true, it assumes the array of tokens is allocated with
1674	/// \c new[] and the Preprocessor will delete[] it.
1675	///
1676	/// If \p IsReinject the resulting tokens will have Token::IsReinjected flag
1677	/// set, see the flag documentation for details.
1678	void EnterTokenStream(const Token *Toks, unsigned NumToks,
1679	bool DisableMacroExpansion, bool OwnsTokens,
1680	bool IsReinject);
1681
1682	public:
1683	void EnterTokenStream(std::unique_ptr<Token[]> Toks, unsigned NumToks,
1684	bool DisableMacroExpansion, bool IsReinject) {
1685	EnterTokenStream(Toks: Toks.release(), NumToks, DisableMacroExpansion, OwnsTokens: true,
1686	IsReinject);
1687	}
1688
1689	void EnterTokenStream(ArrayRef<Token> Toks, bool DisableMacroExpansion,
1690	bool IsReinject) {
1691	EnterTokenStream(Toks: Toks.data(), NumToks: Toks.size(), DisableMacroExpansion, OwnsTokens: false,
1692	IsReinject);
1693	}
1694
1695	/// Pop the current lexer/macro exp off the top of the lexer stack.
1696	///
1697	/// This should only be used in situations where the current state of the
1698	/// top-of-stack lexer is known.
1699	void RemoveTopOfLexerStack();
1700
1701	/// From the point that this method is called, and until
1702	/// CommitBacktrackedTokens() or Backtrack() is called, the Preprocessor
1703	/// keeps track of the lexed tokens so that a subsequent Backtrack() call will
1704	/// make the Preprocessor re-lex the same tokens.
1705	///
1706	/// Nested backtracks are allowed, meaning that EnableBacktrackAtThisPos can
1707	/// be called multiple times and CommitBacktrackedTokens/Backtrack calls will
1708	/// be combined with the EnableBacktrackAtThisPos calls in reverse order.
1709	///
1710	/// NOTE: *DO NOT* forget to call either CommitBacktrackedTokens or Backtrack
1711	/// at some point after EnableBacktrackAtThisPos. If you don't, caching of
1712	/// tokens will continue indefinitely.
1713	///
1714	void EnableBacktrackAtThisPos();
1715
1716	/// Disable the last EnableBacktrackAtThisPos call.
1717	void CommitBacktrackedTokens();
1718
1719	/// Make Preprocessor re-lex the tokens that were lexed since
1720	/// EnableBacktrackAtThisPos() was previously called.
1721	void Backtrack();
1722
1723	/// True if EnableBacktrackAtThisPos() was called and
1724	/// caching of tokens is on.
1725	bool isBacktrackEnabled() const { return !BacktrackPositions.empty(); }
1726
1727	/// Lex the next token for this preprocessor.
1728	void Lex(Token &Result);
1729
1730	/// Lex all tokens for this preprocessor until (and excluding) end of file.
1731	void LexTokensUntilEOF(std::vector<Token> *Tokens = nullptr);
1732
1733	/// Lex a token, forming a header-name token if possible.
1734	bool LexHeaderName(Token &Result, bool AllowMacroExpansion = true);
1735
1736	bool LexAfterModuleImport(Token &Result);
1737	void CollectPpImportSuffix(SmallVectorImpl<Token> &Toks);
1738
1739	void makeModuleVisible(Module *M, SourceLocation Loc);
1740
1741	SourceLocation getModuleImportLoc(Module *M) const {
1742	return CurSubmoduleState->VisibleModules.getImportLoc(M);
1743	}
1744
1745	/// Lex a string literal, which may be the concatenation of multiple
1746	/// string literals and may even come from macro expansion.
1747	/// \returns true on success, false if a error diagnostic has been generated.
1748	bool LexStringLiteral(Token &Result, std::string &String,
1749	const char *DiagnosticTag, bool AllowMacroExpansion) {
1750	if (AllowMacroExpansion)
1751	Lex(Result);
1752	else
1753	LexUnexpandedToken(Result);
1754	return FinishLexStringLiteral(Result, String, DiagnosticTag,
1755	AllowMacroExpansion);
1756	}
1757
1758	/// Complete the lexing of a string literal where the first token has
1759	/// already been lexed (see LexStringLiteral).
1760	bool FinishLexStringLiteral(Token &Result, std::string &String,
1761	const char *DiagnosticTag,
1762	bool AllowMacroExpansion);
1763
1764	/// Lex a token. If it's a comment, keep lexing until we get
1765	/// something not a comment.
1766	///
1767	/// This is useful in -E -C mode where comments would foul up preprocessor
1768	/// directive handling.
1769	void LexNonComment(Token &Result) {
1770	do
1771	Lex(Result);
1772	while (Result.getKind() == tok::comment);
1773	}
1774
1775	/// Just like Lex, but disables macro expansion of identifier tokens.
1776	void LexUnexpandedToken(Token &Result) {
1777	// Disable macro expansion.
1778	bool OldVal = DisableMacroExpansion;
1779	DisableMacroExpansion = true;
1780	// Lex the token.
1781	Lex(Result);
1782
1783	// Reenable it.
1784	DisableMacroExpansion = OldVal;
1785	}
1786
1787	/// Like LexNonComment, but this disables macro expansion of
1788	/// identifier tokens.
1789	void LexUnexpandedNonComment(Token &Result) {
1790	do
1791	LexUnexpandedToken(Result);
1792	while (Result.getKind() == tok::comment);
1793	}
1794
1795	/// Parses a simple integer literal to get its numeric value. Floating
1796	/// point literals and user defined literals are rejected. Used primarily to
1797	/// handle pragmas that accept integer arguments.
1798	bool parseSimpleIntegerLiteral(Token &Tok, uint64_t &Value);
1799
1800	/// Disables macro expansion everywhere except for preprocessor directives.
1801	void SetMacroExpansionOnlyInDirectives() {
1802	DisableMacroExpansion = true;
1803	MacroExpansionInDirectivesOverride = true;
1804	}
1805
1806	/// Peeks ahead N tokens and returns that token without consuming any
1807	/// tokens.
1808	///
1809	/// LookAhead(0) returns the next token that would be returned by Lex(),
1810	/// LookAhead(1) returns the token after it, etc. This returns normal
1811	/// tokens after phase 5. As such, it is equivalent to using
1812	/// 'Lex', not 'LexUnexpandedToken'.
1813	const Token &LookAhead(unsigned N) {
1814	assert(LexLevel == 0 && "cannot use lookahead while lexing");
1815	if (CachedLexPos + N < CachedTokens.size())
1816	return CachedTokens[CachedLexPos+N];
1817	else
1818	return PeekAhead(N: N+1);
1819	}
1820
1821	/// When backtracking is enabled and tokens are cached,
1822	/// this allows to revert a specific number of tokens.
1823	///
1824	/// Note that the number of tokens being reverted should be up to the last
1825	/// backtrack position, not more.
1826	void RevertCachedTokens(unsigned N) {
1827	assert(isBacktrackEnabled() &&
1828	"Should only be called when tokens are cached for backtracking");
1829	assert(signed(CachedLexPos) - signed(N) >= signed(BacktrackPositions.back())
1830	&& "Should revert tokens up to the last backtrack position, not more");
1831	assert(signed(CachedLexPos) - signed(N) >= 0 &&
1832	"Corrupted backtrack positions ?");
1833	CachedLexPos -= N;
1834	}
1835
1836	/// Enters a token in the token stream to be lexed next.
1837	///
1838	/// If BackTrack() is called afterwards, the token will remain at the
1839	/// insertion point.
1840	/// If \p IsReinject is true, resulting token will have Token::IsReinjected
1841	/// flag set. See the flag documentation for details.
1842	void EnterToken(const Token &Tok, bool IsReinject) {
1843	if (LexLevel) {
1844	// It's not correct in general to enter caching lex mode while in the
1845	// middle of a nested lexing action.
1846	auto TokCopy = std::make_unique<Token[]>(num: 1);
1847	TokCopy[0] = Tok;
1848	EnterTokenStream(Toks: std::move(TokCopy), NumToks: 1, DisableMacroExpansion: true, IsReinject);
1849	} else {
1850	EnterCachingLexMode();
1851	assert(IsReinject && "new tokens in the middle of cached stream");
1852	CachedTokens.insert(I: CachedTokens.begin()+CachedLexPos, Elt: Tok);
1853	}
1854	}
1855
1856	/// We notify the Preprocessor that if it is caching tokens (because
1857	/// backtrack is enabled) it should replace the most recent cached tokens
1858	/// with the given annotation token. This function has no effect if
1859	/// backtracking is not enabled.
1860	///
1861	/// Note that the use of this function is just for optimization, so that the
1862	/// cached tokens doesn't get re-parsed and re-resolved after a backtrack is
1863	/// invoked.
1864	void AnnotateCachedTokens(const Token &Tok) {
1865	assert(Tok.isAnnotation() && "Expected annotation token");
1866	if (CachedLexPos != 0 && isBacktrackEnabled())
1867	AnnotatePreviousCachedTokens(Tok);
1868	}
1869
1870	/// Get the location of the last cached token, suitable for setting the end
1871	/// location of an annotation token.
1872	SourceLocation getLastCachedTokenLocation() const {
1873	assert(CachedLexPos != 0);
1874	return CachedTokens[CachedLexPos-1].getLastLoc();
1875	}
1876
1877	/// Whether \p Tok is the most recent token (`CachedLexPos - 1`) in
1878	/// CachedTokens.
1879	bool IsPreviousCachedToken(const Token &Tok) const;
1880
1881	/// Replace token in `CachedLexPos - 1` in CachedTokens by the tokens
1882	/// in \p NewToks.
1883	///
1884	/// Useful when a token needs to be split in smaller ones and CachedTokens
1885	/// most recent token must to be updated to reflect that.
1886	void ReplacePreviousCachedToken(ArrayRef<Token> NewToks);
1887
1888	/// Replace the last token with an annotation token.
1889	///
1890	/// Like AnnotateCachedTokens(), this routine replaces an
1891	/// already-parsed (and resolved) token with an annotation
1892	/// token. However, this routine only replaces the last token with
1893	/// the annotation token; it does not affect any other cached
1894	/// tokens. This function has no effect if backtracking is not
1895	/// enabled.
1896	void ReplaceLastTokenWithAnnotation(const Token &Tok) {
1897	assert(Tok.isAnnotation() && "Expected annotation token");
1898	if (CachedLexPos != 0 && isBacktrackEnabled())
1899	CachedTokens[CachedLexPos-1] = Tok;
1900	}
1901
1902	/// Enter an annotation token into the token stream.
1903	void EnterAnnotationToken(SourceRange Range, tok::TokenKind Kind,
1904	void *AnnotationVal);
1905
1906	/// Determine whether it's possible for a future call to Lex to produce an
1907	/// annotation token created by a previous call to EnterAnnotationToken.
1908	bool mightHavePendingAnnotationTokens() {
1909	return CurLexerCallback != CLK_Lexer;
1910	}
1911
1912	/// Update the current token to represent the provided
1913	/// identifier, in order to cache an action performed by typo correction.
1914	void TypoCorrectToken(const Token &Tok) {
1915	assert(Tok.getIdentifierInfo() && "Expected identifier token");
1916	if (CachedLexPos != 0 && isBacktrackEnabled())
1917	CachedTokens[CachedLexPos-1] = Tok;
1918	}
1919
1920	/// Recompute the current lexer kind based on the CurLexer/
1921	/// CurTokenLexer pointers.
1922	void recomputeCurLexerKind();
1923
1924	/// Returns true if incremental processing is enabled
1925	bool isIncrementalProcessingEnabled() const { return IncrementalProcessing; }
1926
1927	/// Enables the incremental processing
1928	void enableIncrementalProcessing(bool value = true) {
1929	IncrementalProcessing = value;
1930	}
1931
1932	/// Specify the point at which code-completion will be performed.
1933	///
1934	/// \param File the file in which code completion should occur. If
1935	/// this file is included multiple times, code-completion will
1936	/// perform completion the first time it is included. If NULL, this
1937	/// function clears out the code-completion point.
1938	///
1939	/// \param Line the line at which code completion should occur
1940	/// (1-based).
1941	///
1942	/// \param Column the column at which code completion should occur
1943	/// (1-based).
1944	///
1945	/// \returns true if an error occurred, false otherwise.
1946	bool SetCodeCompletionPoint(FileEntryRef File, unsigned Line,
1947	unsigned Column);
1948
1949	/// Determine if we are performing code completion.
1950	bool isCodeCompletionEnabled() const { return CodeCompletionFile != nullptr; }
1951
1952	/// Returns the location of the code-completion point.
1953	///
1954	/// Returns an invalid location if code-completion is not enabled or the file
1955	/// containing the code-completion point has not been lexed yet.
1956	SourceLocation getCodeCompletionLoc() const { return CodeCompletionLoc; }
1957
1958	/// Returns the start location of the file of code-completion point.
1959	///
1960	/// Returns an invalid location if code-completion is not enabled or the file
1961	/// containing the code-completion point has not been lexed yet.
1962	SourceLocation getCodeCompletionFileLoc() const {
1963	return CodeCompletionFileLoc;
1964	}
1965
1966	/// Returns true if code-completion is enabled and we have hit the
1967	/// code-completion point.
1968	bool isCodeCompletionReached() const { return CodeCompletionReached; }
1969
1970	/// Note that we hit the code-completion point.
1971	void setCodeCompletionReached() {
1972	assert(isCodeCompletionEnabled() && "Code-completion not enabled!");
1973	CodeCompletionReached = true;
1974	// Silence any diagnostics that occur after we hit the code-completion.
1975	getDiagnostics().setSuppressAllDiagnostics(true);
1976	}
1977
1978	/// The location of the currently-active \#pragma clang
1979	/// arc_cf_code_audited begin.
1980	///
1981	/// Returns an invalid location if there is no such pragma active.
1982	std::pair<IdentifierInfo *, SourceLocation>
1983	getPragmaARCCFCodeAuditedInfo() const {
1984	return PragmaARCCFCodeAuditedInfo;
1985	}
1986
1987	/// Set the location of the currently-active \#pragma clang
1988	/// arc_cf_code_audited begin. An invalid location ends the pragma.
1989	void setPragmaARCCFCodeAuditedInfo(IdentifierInfo *Ident,
1990	SourceLocation Loc) {
1991	PragmaARCCFCodeAuditedInfo = {Ident, Loc};
1992	}
1993
1994	/// The location of the currently-active \#pragma clang
1995	/// assume_nonnull begin.
1996	///
1997	/// Returns an invalid location if there is no such pragma active.
1998	SourceLocation getPragmaAssumeNonNullLoc() const {
1999	return PragmaAssumeNonNullLoc;
2000	}
2001
2002	/// Set the location of the currently-active \#pragma clang
2003	/// assume_nonnull begin. An invalid location ends the pragma.
2004	void setPragmaAssumeNonNullLoc(SourceLocation Loc) {
2005	PragmaAssumeNonNullLoc = Loc;
2006	}
2007
2008	/// Get the location of the recorded unterminated \#pragma clang
2009	/// assume_nonnull begin in the preamble, if one exists.
2010	///
2011	/// Returns an invalid location if the premable did not end with
2012	/// such a pragma active or if there is no recorded preamble.
2013	SourceLocation getPreambleRecordedPragmaAssumeNonNullLoc() const {
2014	return PreambleRecordedPragmaAssumeNonNullLoc;
2015	}
2016
2017	/// Record the location of the unterminated \#pragma clang
2018	/// assume_nonnull begin in the preamble.
2019	void setPreambleRecordedPragmaAssumeNonNullLoc(SourceLocation Loc) {
2020	PreambleRecordedPragmaAssumeNonNullLoc = Loc;
2021	}
2022
2023	/// Set the directory in which the main file should be considered
2024	/// to have been found, if it is not a real file.
2025	void setMainFileDir(DirectoryEntryRef Dir) { MainFileDir = Dir; }
2026
2027	/// Instruct the preprocessor to skip part of the main source file.
2028	///
2029	/// \param Bytes The number of bytes in the preamble to skip.
2030	///
2031	/// \param StartOfLine Whether skipping these bytes puts the lexer at the
2032	/// start of a line.
2033	void setSkipMainFilePreamble(unsigned Bytes, bool StartOfLine) {
2034	SkipMainFilePreamble.first = Bytes;
2035	SkipMainFilePreamble.second = StartOfLine;
2036	}
2037
2038	/// Forwarding function for diagnostics. This emits a diagnostic at
2039	/// the specified Token's location, translating the token's start
2040	/// position in the current buffer into a SourcePosition object for rendering.
2041	DiagnosticBuilder Diag(SourceLocation Loc, unsigned DiagID) const {
2042	return Diags->Report(Loc, DiagID);
2043	}
2044
2045	DiagnosticBuilder Diag(const Token &Tok, unsigned DiagID) const {
2046	return Diags->Report(Loc: Tok.getLocation(), DiagID);
2047	}
2048
2049	/// Return the 'spelling' of the token at the given
2050	/// location; does not go up to the spelling location or down to the
2051	/// expansion location.
2052	///
2053	/// \param buffer A buffer which will be used only if the token requires
2054	/// "cleaning", e.g. if it contains trigraphs or escaped newlines
2055	/// \param invalid If non-null, will be set \c true if an error occurs.
2056	StringRef getSpelling(SourceLocation loc,
2057	SmallVectorImpl<char> &buffer,
2058	bool *invalid = nullptr) const {
2059	return Lexer::getSpelling(loc, buffer, SM: SourceMgr, options: LangOpts, invalid);
2060	}
2061
2062	/// Return the 'spelling' of the Tok token.
2063	///
2064	/// The spelling of a token is the characters used to represent the token in
2065	/// the source file after trigraph expansion and escaped-newline folding. In
2066	/// particular, this wants to get the true, uncanonicalized, spelling of
2067	/// things like digraphs, UCNs, etc.
2068	///
2069	/// \param Invalid If non-null, will be set \c true if an error occurs.
2070	std::string getSpelling(const Token &Tok, bool *Invalid = nullptr) const {
2071	return Lexer::getSpelling(Tok, SourceMgr, LangOpts, Invalid);
2072	}
2073
2074	/// Get the spelling of a token into a preallocated buffer, instead
2075	/// of as an std::string.
2076	///
2077	/// The caller is required to allocate enough space for the token, which is
2078	/// guaranteed to be at least Tok.getLength() bytes long. The length of the
2079	/// actual result is returned.
2080	///
2081	/// Note that this method may do two possible things: it may either fill in
2082	/// the buffer specified with characters, or it may *change the input pointer*
2083	/// to point to a constant buffer with the data already in it (avoiding a
2084	/// copy). The caller is not allowed to modify the returned buffer pointer
2085	/// if an internal buffer is returned.
2086	unsigned getSpelling(const Token &Tok, const char *&Buffer,
2087	bool *Invalid = nullptr) const {
2088	return Lexer::getSpelling(Tok, Buffer, SourceMgr, LangOpts, Invalid);
2089	}
2090
2091	/// Get the spelling of a token into a SmallVector.
2092	///
2093	/// Note that the returned StringRef may not point to the
2094	/// supplied buffer if a copy can be avoided.
2095	StringRef getSpelling(const Token &Tok,
2096	SmallVectorImpl<char> &Buffer,
2097	bool *Invalid = nullptr) const;
2098
2099	/// Relex the token at the specified location.
2100	/// \returns true if there was a failure, false on success.
2101	bool getRawToken(SourceLocation Loc, Token &Result,
2102	bool IgnoreWhiteSpace = false) {
2103	return Lexer::getRawToken(Loc, Result, SM: SourceMgr, LangOpts, IgnoreWhiteSpace);
2104	}
2105
2106	/// Given a Token \p Tok that is a numeric constant with length 1,
2107	/// return the character.
2108	char
2109	getSpellingOfSingleCharacterNumericConstant(const Token &Tok,
2110	bool *Invalid = nullptr) const {
2111	assert(Tok.is(tok::numeric_constant) &&
2112	Tok.getLength() == 1 && "Called on unsupported token");
2113	assert(!Tok.needsCleaning() && "Token can't need cleaning with length 1");
2114
2115	// If the token is carrying a literal data pointer, just use it.
2116	if (const char *D = Tok.getLiteralData())
2117	return *D;
2118
2119	// Otherwise, fall back on getCharacterData, which is slower, but always
2120	// works.
2121	return *SourceMgr.getCharacterData(SL: Tok.getLocation(), Invalid);
2122	}
2123
2124	/// Retrieve the name of the immediate macro expansion.
2125	///
2126	/// This routine starts from a source location, and finds the name of the
2127	/// macro responsible for its immediate expansion. It looks through any
2128	/// intervening macro argument expansions to compute this. It returns a
2129	/// StringRef that refers to the SourceManager-owned buffer of the source
2130	/// where that macro name is spelled. Thus, the result shouldn't out-live
2131	/// the SourceManager.
2132	StringRef getImmediateMacroName(SourceLocation Loc) {
2133	return Lexer::getImmediateMacroName(Loc, SM: SourceMgr, LangOpts: getLangOpts());
2134	}
2135
2136	/// Plop the specified string into a scratch buffer and set the
2137	/// specified token's location and length to it.
2138	///
2139	/// If specified, the source location provides a location of the expansion
2140	/// point of the token.
2141	void CreateString(StringRef Str, Token &Tok,
2142	SourceLocation ExpansionLocStart = SourceLocation(),
2143	SourceLocation ExpansionLocEnd = SourceLocation());
2144
2145	/// Split the first Length characters out of the token starting at TokLoc
2146	/// and return a location pointing to the split token. Re-lexing from the
2147	/// split token will return the split token rather than the original.
2148	SourceLocation SplitToken(SourceLocation TokLoc, unsigned Length);
2149
2150	/// Computes the source location just past the end of the
2151	/// token at this source location.
2152	///
2153	/// This routine can be used to produce a source location that
2154	/// points just past the end of the token referenced by \p Loc, and
2155	/// is generally used when a diagnostic needs to point just after a
2156	/// token where it expected something different that it received. If
2157	/// the returned source location would not be meaningful (e.g., if
2158	/// it points into a macro), this routine returns an invalid
2159	/// source location.
2160	///
2161	/// \param Offset an offset from the end of the token, where the source
2162	/// location should refer to. The default offset (0) produces a source
2163	/// location pointing just past the end of the token; an offset of 1 produces
2164	/// a source location pointing to the last character in the token, etc.
2165	SourceLocation getLocForEndOfToken(SourceLocation Loc, unsigned Offset = 0) {
2166	return Lexer::getLocForEndOfToken(Loc, Offset, SM: SourceMgr, LangOpts);
2167	}
2168
2169	/// Returns true if the given MacroID location points at the first
2170	/// token of the macro expansion.
2171	///
2172	/// \param MacroBegin If non-null and function returns true, it is set to
2173	/// begin location of the macro.
2174	bool isAtStartOfMacroExpansion(SourceLocation loc,
2175	SourceLocation *MacroBegin = nullptr) const {
2176	return Lexer::isAtStartOfMacroExpansion(loc, SM: SourceMgr, LangOpts,
2177	MacroBegin);
2178	}
2179
2180	/// Returns true if the given MacroID location points at the last
2181	/// token of the macro expansion.
2182	///
2183	/// \param MacroEnd If non-null and function returns true, it is set to
2184	/// end location of the macro.
2185	bool isAtEndOfMacroExpansion(SourceLocation loc,
2186	SourceLocation *MacroEnd = nullptr) const {
2187	return Lexer::isAtEndOfMacroExpansion(loc, SM: SourceMgr, LangOpts, MacroEnd);
2188	}
2189
2190	/// Print the token to stderr, used for debugging.
2191	void DumpToken(const Token &Tok, bool DumpFlags = false) const;
2192	void DumpLocation(SourceLocation Loc) const;
2193	void DumpMacro(const MacroInfo &MI) const;
2194	void dumpMacroInfo(const IdentifierInfo *II);
2195
2196	/// Given a location that specifies the start of a
2197	/// token, return a new location that specifies a character within the token.
2198	SourceLocation AdvanceToTokenCharacter(SourceLocation TokStart,
2199	unsigned Char) const {
2200	return Lexer::AdvanceToTokenCharacter(TokStart, Characters: Char, SM: SourceMgr, LangOpts);
2201	}
2202
2203	/// Increment the counters for the number of token paste operations
2204	/// performed.
2205	///
2206	/// If fast was specified, this is a 'fast paste' case we handled.
2207	void IncrementPasteCounter(bool isFast) {
2208	if (isFast)
2209	++NumFastTokenPaste;
2210	else
2211	++NumTokenPaste;
2212	}
2213
2214	void PrintStats();
2215
2216	size_t getTotalMemory() const;
2217
2218	/// When the macro expander pastes together a comment (/##/) in Microsoft
2219	/// mode, this method handles updating the current state, returning the
2220	/// token on the next source line.
2221	void HandleMicrosoftCommentPaste(Token &Tok);
2222
2223	//===--------------------------------------------------------------------===//
2224	// Preprocessor callback methods. These are invoked by a lexer as various
2225	// directives and events are found.
2226
2227	/// Given a tok::raw_identifier token, look up the
2228	/// identifier information for the token and install it into the token,
2229	/// updating the token kind accordingly.
2230	IdentifierInfo *LookUpIdentifierInfo(Token &Identifier) const;
2231
2232	private:
2233	llvm::DenseMap<IdentifierInfo*,unsigned> PoisonReasons;
2234
2235	public:
2236	/// Specifies the reason for poisoning an identifier.
2237	///
2238	/// If that identifier is accessed while poisoned, then this reason will be
2239	/// used instead of the default "poisoned" diagnostic.
2240	void SetPoisonReason(IdentifierInfo *II, unsigned DiagID);
2241
2242	/// Display reason for poisoned identifier.
2243	void HandlePoisonedIdentifier(Token & Identifier);
2244
2245	void MaybeHandlePoisonedIdentifier(Token & Identifier) {
2246	if(IdentifierInfo * II = Identifier.getIdentifierInfo()) {
2247	if(II->isPoisoned()) {
2248	HandlePoisonedIdentifier(Identifier);
2249	}
2250	}
2251	}
2252
2253	private:
2254	/// Identifiers used for SEH handling in Borland. These are only
2255	/// allowed in particular circumstances
2256	// __except block
2257	IdentifierInfo *Ident__exception_code,
2258	*Ident___exception_code,
2259	*Ident_GetExceptionCode;
2260	// __except filter expression
2261	IdentifierInfo *Ident__exception_info,
2262	*Ident___exception_info,
2263	*Ident_GetExceptionInfo;
2264	// __finally
2265	IdentifierInfo *Ident__abnormal_termination,
2266	*Ident___abnormal_termination,
2267	*Ident_AbnormalTermination;
2268
2269	const char *getCurLexerEndPos();
2270	void diagnoseMissingHeaderInUmbrellaDir(const Module &Mod);
2271
2272	public:
2273	void PoisonSEHIdentifiers(bool Poison = true); // Borland
2274
2275	/// Callback invoked when the lexer reads an identifier and has
2276	/// filled in the tokens IdentifierInfo member.
2277	///
2278	/// This callback potentially macro expands it or turns it into a named
2279	/// token (like 'for').
2280	///
2281	/// \returns true if we actually computed a token, false if we need to
2282	/// lex again.
2283	bool HandleIdentifier(Token &Identifier);
2284
2285	/// Callback invoked when the lexer hits the end of the current file.
2286	///
2287	/// This either returns the EOF token and returns true, or
2288	/// pops a level off the include stack and returns false, at which point the
2289	/// client should call lex again.
2290	bool HandleEndOfFile(Token &Result, bool isEndOfMacro = false);
2291
2292	/// Callback invoked when the current TokenLexer hits the end of its
2293	/// token stream.
2294	bool HandleEndOfTokenLexer(Token &Result);
2295
2296	/// Callback invoked when the lexer sees a # token at the start of a
2297	/// line.
2298	///
2299	/// This consumes the directive, modifies the lexer/preprocessor state, and
2300	/// advances the lexer(s) so that the next token read is the correct one.
2301	void HandleDirective(Token &Result);
2302
2303	/// Ensure that the next token is a tok::eod token.
2304	///
2305	/// If not, emit a diagnostic and consume up until the eod.
2306	/// If \p EnableMacros is true, then we consider macros that expand to zero
2307	/// tokens as being ok.
2308	///
2309	/// \return The location of the end of the directive (the terminating
2310	/// newline).
2311	SourceLocation CheckEndOfDirective(const char *DirType,
2312	bool EnableMacros = false);
2313
2314	/// Read and discard all tokens remaining on the current line until
2315	/// the tok::eod token is found. Returns the range of the skipped tokens.
2316	SourceRange DiscardUntilEndOfDirective();
2317
2318	/// Returns true if the preprocessor has seen a use of
2319	/// __DATE__ or __TIME__ in the file so far.
2320	bool SawDateOrTime() const {
2321	return DATELoc != SourceLocation() || TIMELoc != SourceLocation();
2322	}
2323	unsigned getCounterValue() const { return CounterValue; }
2324	void setCounterValue(unsigned V) { CounterValue = V; }
2325
2326	LangOptions::FPEvalMethodKind getCurrentFPEvalMethod() const {
2327	assert(CurrentFPEvalMethod != LangOptions::FEM_UnsetOnCommandLine &&
2328	"FPEvalMethod should be set either from command line or from the "
2329	"target info");
2330	return CurrentFPEvalMethod;
2331	}
2332
2333	LangOptions::FPEvalMethodKind getTUFPEvalMethod() const {
2334	return TUFPEvalMethod;
2335	}
2336
2337	SourceLocation getLastFPEvalPragmaLocation() const {
2338	return LastFPEvalPragmaLocation;
2339	}
2340
2341	void setCurrentFPEvalMethod(SourceLocation PragmaLoc,
2342	LangOptions::FPEvalMethodKind Val) {
2343	assert(Val != LangOptions::FEM_UnsetOnCommandLine &&
2344	"FPEvalMethod should never be set to FEM_UnsetOnCommandLine");
2345	// This is the location of the '#pragma float_control" where the
2346	// execution state is modifed.
2347	LastFPEvalPragmaLocation = PragmaLoc;
2348	CurrentFPEvalMethod = Val;
2349	TUFPEvalMethod = Val;
2350	}
2351
2352	void setTUFPEvalMethod(LangOptions::FPEvalMethodKind Val) {
2353	assert(Val != LangOptions::FEM_UnsetOnCommandLine &&
2354	"TUPEvalMethod should never be set to FEM_UnsetOnCommandLine");
2355	TUFPEvalMethod = Val;
2356	}
2357
2358	/// Retrieves the module that we're currently building, if any.
2359	Module *getCurrentModule();
2360
2361	/// Retrieves the module whose implementation we're current compiling, if any.
2362	Module *getCurrentModuleImplementation();
2363
2364	/// If we are preprocessing a named module.
2365	bool isInNamedModule() const { return ModuleDeclState.isNamedModule(); }
2366
2367	/// If we are proprocessing a named interface unit.
2368	/// Note that a module implementation partition is not considered as an
2369	/// named interface unit here although it is importable
2370	/// to ease the parsing.
2371	bool isInNamedInterfaceUnit() const {
2372	return ModuleDeclState.isNamedInterface();
2373	}
2374
2375	/// Get the named module name we're preprocessing.
2376	/// Requires we're preprocessing a named module.
2377	StringRef getNamedModuleName() const { return ModuleDeclState.getName(); }
2378
2379	/// If we are implementing an implementation module unit.
2380	/// Note that the module implementation partition is not considered as an
2381	/// implementation unit.
2382	bool isInImplementationUnit() const {
2383	return ModuleDeclState.isImplementationUnit();
2384	}
2385
2386	/// If we're importing a standard C++20 Named Modules.
2387	bool isInImportingCXXNamedModules() const {
2388	// NamedModuleImportPath will be non-empty only if we're importing
2389	// Standard C++ named modules.
2390	return !NamedModuleImportPath.empty() && getLangOpts().CPlusPlusModules &&
2391	!IsAtImport;
2392	}
2393
2394	/// Allocate a new MacroInfo object with the provided SourceLocation.
2395	MacroInfo *AllocateMacroInfo(SourceLocation L);
2396
2397	/// Turn the specified lexer token into a fully checked and spelled
2398	/// filename, e.g. as an operand of \#include.
2399	///
2400	/// The caller is expected to provide a buffer that is large enough to hold
2401	/// the spelling of the filename, but is also expected to handle the case
2402	/// when this method decides to use a different buffer.
2403	///
2404	/// \returns true if the input filename was in <>'s or false if it was
2405	/// in ""'s.
2406	bool GetIncludeFilenameSpelling(SourceLocation Loc,StringRef &Buffer);
2407
2408	/// Given a "foo" or \<foo> reference, look up the indicated file.
2409	///
2410	/// Returns std::nullopt on failure. \p isAngled indicates whether the file
2411	/// reference is for system \#include's or not (i.e. using <> instead of "").
2412	OptionalFileEntryRef
2413	LookupFile(SourceLocation FilenameLoc, StringRef Filename, bool isAngled,
2414	ConstSearchDirIterator FromDir, const FileEntry *FromFile,
2415	ConstSearchDirIterator *CurDir, SmallVectorImpl<char> *SearchPath,
2416	SmallVectorImpl<char> *RelativePath,
2417	ModuleMap::KnownHeader *SuggestedModule, bool *IsMapped,
2418	bool *IsFrameworkFound, bool SkipCache = false,
2419	bool OpenFile = true, bool CacheFailures = true);
2420
2421	/// Return true if we're in the top-level file, not in a \#include.
2422	bool isInPrimaryFile() const;
2423
2424	/// Lex an on-off-switch (C99 6.10.6p2) and verify that it is
2425	/// followed by EOD. Return true if the token is not a valid on-off-switch.
2426	bool LexOnOffSwitch(tok::OnOffSwitch &Result);
2427
2428	bool CheckMacroName(Token &MacroNameTok, MacroUse isDefineUndef,
2429	bool *ShadowFlag = nullptr);
2430
2431	void EnterSubmodule(Module *M, SourceLocation ImportLoc, bool ForPragma);
2432	Module *LeaveSubmodule(bool ForPragma);
2433
2434	private:
2435	friend void TokenLexer::ExpandFunctionArguments();
2436
2437	void PushIncludeMacroStack() {
2438	assert(CurLexerCallback != CLK_CachingLexer &&
2439	"cannot push a caching lexer");
2440	IncludeMacroStack.emplace_back(args&: CurLexerCallback, args&: CurLexerSubmodule,
2441	args: std::move(CurLexer), args&: CurPPLexer,
2442	args: std::move(CurTokenLexer), args&: CurDirLookup);
2443	CurPPLexer = nullptr;
2444	}
2445
2446	void PopIncludeMacroStack() {
2447	CurLexer = std::move(IncludeMacroStack.back().TheLexer);
2448	CurPPLexer = IncludeMacroStack.back().ThePPLexer;
2449	CurTokenLexer = std::move(IncludeMacroStack.back().TheTokenLexer);
2450	CurDirLookup = IncludeMacroStack.back().TheDirLookup;
2451	CurLexerSubmodule = IncludeMacroStack.back().TheSubmodule;
2452	CurLexerCallback = IncludeMacroStack.back().CurLexerCallback;
2453	IncludeMacroStack.pop_back();
2454	}
2455
2456	void PropagateLineStartLeadingSpaceInfo(Token &Result);
2457
2458	/// Determine whether we need to create module macros for #defines in the
2459	/// current context.
2460	bool needModuleMacros() const;
2461
2462	/// Update the set of active module macros and ambiguity flag for a module
2463	/// macro name.
2464	void updateModuleMacroInfo(const IdentifierInfo *II, ModuleMacroInfo &Info);
2465
2466	DefMacroDirective *AllocateDefMacroDirective(MacroInfo *MI,
2467	SourceLocation Loc);
2468	UndefMacroDirective *AllocateUndefMacroDirective(SourceLocation UndefLoc);
2469	VisibilityMacroDirective *AllocateVisibilityMacroDirective(SourceLocation Loc,
2470	bool isPublic);
2471
2472	/// Lex and validate a macro name, which occurs after a
2473	/// \#define or \#undef.
2474	///
2475	/// \param MacroNameTok Token that represents the name defined or undefined.
2476	/// \param IsDefineUndef Kind if preprocessor directive.
2477	/// \param ShadowFlag Points to flag that is set if macro name shadows
2478	/// a keyword.
2479	///
2480	/// This emits a diagnostic, sets the token kind to eod,
2481	/// and discards the rest of the macro line if the macro name is invalid.
2482	void ReadMacroName(Token &MacroNameTok, MacroUse IsDefineUndef = MU_Other,
2483	bool *ShadowFlag = nullptr);
2484
2485	/// ReadOptionalMacroParameterListAndBody - This consumes all (i.e. the
2486	/// entire line) of the macro's tokens and adds them to MacroInfo, and while
2487	/// doing so performs certain validity checks including (but not limited to):
2488	/// - # (stringization) is followed by a macro parameter
2489	/// \param MacroNameTok - Token that represents the macro name
2490	/// \param ImmediatelyAfterHeaderGuard - Macro follows an #ifdef header guard
2491	///
2492	/// Either returns a pointer to a MacroInfo object OR emits a diagnostic and
2493	/// returns a nullptr if an invalid sequence of tokens is encountered.
2494	MacroInfo *ReadOptionalMacroParameterListAndBody(
2495	const Token &MacroNameTok, bool ImmediatelyAfterHeaderGuard);
2496
2497	/// The ( starting an argument list of a macro definition has just been read.
2498	/// Lex the rest of the parameters and the closing ), updating \p MI with
2499	/// what we learn and saving in \p LastTok the last token read.
2500	/// Return true if an error occurs parsing the arg list.
2501	bool ReadMacroParameterList(MacroInfo *MI, Token& LastTok);
2502
2503	/// Provide a suggestion for a typoed directive. If there is no typo, then
2504	/// just skip suggesting.
2505	///
2506	/// \param Tok - Token that represents the directive
2507	/// \param Directive - String reference for the directive name
2508	void SuggestTypoedDirective(const Token &Tok, StringRef Directive) const;
2509
2510	/// We just read a \#if or related directive and decided that the
2511	/// subsequent tokens are in the \#if'd out portion of the
2512	/// file. Lex the rest of the file, until we see an \#endif. If \p
2513	/// FoundNonSkipPortion is true, then we have already emitted code for part of
2514	/// this \#if directive, so \#else/\#elif blocks should never be entered. If
2515	/// \p FoundElse is false, then \#else directives are ok, if not, then we have
2516	/// already seen one so a \#else directive is a duplicate. When this returns,
2517	/// the caller can lex the first valid token.
2518	void SkipExcludedConditionalBlock(SourceLocation HashTokenLoc,
2519	SourceLocation IfTokenLoc,
2520	bool FoundNonSkipPortion, bool FoundElse,
2521	SourceLocation ElseLoc = SourceLocation());
2522
2523	/// Information about the result for evaluating an expression for a
2524	/// preprocessor directive.
2525	struct DirectiveEvalResult {
2526	/// Whether the expression was evaluated as true or not.
2527	bool Conditional;
2528
2529	/// True if the expression contained identifiers that were undefined.
2530	bool IncludedUndefinedIds;
2531
2532	/// The source range for the expression.
2533	SourceRange ExprRange;
2534	};
2535
2536	/// Evaluate an integer constant expression that may occur after a
2537	/// \#if or \#elif directive and return a \p DirectiveEvalResult object.
2538	///
2539	/// If the expression is equivalent to "!defined(X)" return X in IfNDefMacro.
2540	DirectiveEvalResult EvaluateDirectiveExpression(IdentifierInfo *&IfNDefMacro);
2541
2542	/// Process a '__has_include("path")' expression.
2543	///
2544	/// Returns true if successful.
2545	bool EvaluateHasInclude(Token &Tok, IdentifierInfo *II);
2546
2547	/// Process '__has_include_next("path")' expression.
2548	///
2549	/// Returns true if successful.
2550	bool EvaluateHasIncludeNext(Token &Tok, IdentifierInfo *II);
2551
2552	/// Get the directory and file from which to start \#include_next lookup.
2553	std::pair<ConstSearchDirIterator, const FileEntry *>
2554	getIncludeNextStart(const Token &IncludeNextTok) const;
2555
2556	/// Install the standard preprocessor pragmas:
2557	/// \#pragma GCC poison/system_header/dependency and \#pragma once.
2558	void RegisterBuiltinPragmas();
2559
2560	/// Register builtin macros such as __LINE__ with the identifier table.
2561	void RegisterBuiltinMacros();
2562
2563	/// If an identifier token is read that is to be expanded as a macro, handle
2564	/// it and return the next token as 'Tok'. If we lexed a token, return true;
2565	/// otherwise the caller should lex again.
2566	bool HandleMacroExpandedIdentifier(Token &Identifier, const MacroDefinition &MD);
2567
2568	/// Cache macro expanded tokens for TokenLexers.
2569	//
2570	/// Works like a stack; a TokenLexer adds the macro expanded tokens that is
2571	/// going to lex in the cache and when it finishes the tokens are removed
2572	/// from the end of the cache.
2573	Token *cacheMacroExpandedTokens(TokenLexer *tokLexer,
2574	ArrayRef<Token> tokens);
2575
2576	void removeCachedMacroExpandedTokensOfLastLexer();
2577
2578	/// Determine whether the next preprocessor token to be
2579	/// lexed is a '('. If so, consume the token and return true, if not, this
2580	/// method should have no observable side-effect on the lexed tokens.
2581	bool isNextPPTokenLParen();
2582
2583	/// After reading "MACRO(", this method is invoked to read all of the formal
2584	/// arguments specified for the macro invocation. Returns null on error.
2585	MacroArgs *ReadMacroCallArgumentList(Token &MacroName, MacroInfo *MI,
2586	SourceLocation &MacroEnd);
2587
2588	/// If an identifier token is read that is to be expanded
2589	/// as a builtin macro, handle it and return the next token as 'Tok'.
2590	void ExpandBuiltinMacro(Token &Tok);
2591
2592	/// Read a \c _Pragma directive, slice it up, process it, then
2593	/// return the first token after the directive.
2594	/// This assumes that the \c _Pragma token has just been read into \p Tok.
2595	void Handle_Pragma(Token &Tok);
2596
2597	/// Like Handle_Pragma except the pragma text is not enclosed within
2598	/// a string literal.
2599	void HandleMicrosoft__pragma(Token &Tok);
2600
2601	/// Add a lexer to the top of the include stack and
2602	/// start lexing tokens from it instead of the current buffer.
2603	void EnterSourceFileWithLexer(Lexer *TheLexer, ConstSearchDirIterator Dir);
2604
2605	/// Set the FileID for the preprocessor predefines.
2606	void setPredefinesFileID(FileID FID) {
2607	assert(PredefinesFileID.isInvalid() && "PredefinesFileID already set!");
2608	PredefinesFileID = FID;
2609	}
2610
2611	/// Set the FileID for the PCH through header.
2612	void setPCHThroughHeaderFileID(FileID FID);
2613
2614	/// Returns true if we are lexing from a file and not a
2615	/// pragma or a macro.
2616	static bool IsFileLexer(const Lexer* L, const PreprocessorLexer* P) {
2617	return L ? !L->isPragmaLexer() : P != nullptr;
2618	}
2619
2620	static bool IsFileLexer(const IncludeStackInfo& I) {
2621	return IsFileLexer(L: I.TheLexer.get(), P: I.ThePPLexer);
2622	}
2623
2624	bool IsFileLexer() const {
2625	return IsFileLexer(L: CurLexer.get(), P: CurPPLexer);
2626	}
2627
2628	//===--------------------------------------------------------------------===//
2629	// Caching stuff.
2630	void CachingLex(Token &Result);
2631
2632	bool InCachingLexMode() const {
2633	// If the Lexer pointers are 0 and IncludeMacroStack is empty, it means
2634	// that we are past EOF, not that we are in CachingLex mode.
2635	return !CurPPLexer && !CurTokenLexer && !IncludeMacroStack.empty();
2636	}
2637
2638	void EnterCachingLexMode();
2639	void EnterCachingLexModeUnchecked();
2640
2641	void ExitCachingLexMode() {
2642	if (InCachingLexMode())
2643	RemoveTopOfLexerStack();
2644	}
2645
2646	const Token &PeekAhead(unsigned N);
2647	void AnnotatePreviousCachedTokens(const Token &Tok);
2648
2649	//===--------------------------------------------------------------------===//
2650	/// Handle*Directive - implement the various preprocessor directives. These
2651	/// should side-effect the current preprocessor object so that the next call
2652	/// to Lex() will return the appropriate token next.
2653	void HandleLineDirective();
2654	void HandleDigitDirective(Token &Tok);
2655	void HandleUserDiagnosticDirective(Token &Tok, bool isWarning);
2656	void HandleIdentSCCSDirective(Token &Tok);
2657	void HandleMacroPublicDirective(Token &Tok);
2658	void HandleMacroPrivateDirective();
2659
2660	/// An additional notification that can be produced by a header inclusion or
2661	/// import to tell the parser what happened.
2662	struct ImportAction {
2663	enum ActionKind {
2664	None,
2665	ModuleBegin,
2666	ModuleImport,
2667	HeaderUnitImport,
2668	SkippedModuleImport,
2669	Failure,
2670	} Kind;
2671	Module *ModuleForHeader = nullptr;
2672
2673	ImportAction(ActionKind AK, Module *Mod = nullptr)
2674	: Kind(AK), ModuleForHeader(Mod) {
2675	assert((AK == None || Mod || AK == Failure) &&
2676	"no module for module action");
2677	}
2678	};
2679
2680	OptionalFileEntryRef LookupHeaderIncludeOrImport(
2681	ConstSearchDirIterator *CurDir, StringRef &Filename,
2682	SourceLocation FilenameLoc, CharSourceRange FilenameRange,
2683	const Token &FilenameTok, bool &IsFrameworkFound, bool IsImportDecl,
2684	bool &IsMapped, ConstSearchDirIterator LookupFrom,
2685	const FileEntry *LookupFromFile, StringRef &LookupFilename,
2686	SmallVectorImpl<char> &RelativePath, SmallVectorImpl<char> &SearchPath,
2687	ModuleMap::KnownHeader &SuggestedModule, bool isAngled);
2688
2689	// File inclusion.
2690	void HandleIncludeDirective(SourceLocation HashLoc, Token &Tok,
2691	ConstSearchDirIterator LookupFrom = nullptr,
2692	const FileEntry *LookupFromFile = nullptr);
2693	ImportAction
2694	HandleHeaderIncludeOrImport(SourceLocation HashLoc, Token &IncludeTok,
2695	Token &FilenameTok, SourceLocation EndLoc,
2696	ConstSearchDirIterator LookupFrom = nullptr,
2697	const FileEntry *LookupFromFile = nullptr);
2698	void HandleIncludeNextDirective(SourceLocation HashLoc, Token &Tok);
2699	void HandleIncludeMacrosDirective(SourceLocation HashLoc, Token &Tok);
2700	void HandleImportDirective(SourceLocation HashLoc, Token &Tok);
2701	void HandleMicrosoftImportDirective(Token &Tok);
2702
2703	public:
2704	/// Check that the given module is available, producing a diagnostic if not.
2705	/// \return \c true if the check failed (because the module is not available).
2706	/// \c false if the module appears to be usable.
2707	static bool checkModuleIsAvailable(const LangOptions &LangOpts,
2708	const TargetInfo &TargetInfo,
2709	const Module &M, DiagnosticsEngine &Diags);
2710
2711	// Module inclusion testing.
2712	/// Find the module that owns the source or header file that
2713	/// \p Loc points to. If the location is in a file that was included
2714	/// into a module, or is outside any module, returns nullptr.
2715	Module *getModuleForLocation(SourceLocation Loc, bool AllowTextual);
2716
2717	/// We want to produce a diagnostic at location IncLoc concerning an
2718	/// unreachable effect at location MLoc (eg, where a desired entity was
2719	/// declared or defined). Determine whether the right way to make MLoc
2720	/// reachable is by #include, and if so, what header should be included.
2721	///
2722	/// This is not necessarily fast, and might load unexpected module maps, so
2723	/// should only be called by code that intends to produce an error.
2724	///
2725	/// \param IncLoc The location at which the missing effect was detected.
2726	/// \param MLoc A location within an unimported module at which the desired
2727	/// effect occurred.
2728	/// \return A file that can be #included to provide the desired effect. Null
2729	/// if no such file could be determined or if a #include is not
2730	/// appropriate (eg, if a module should be imported instead).
2731	OptionalFileEntryRef getHeaderToIncludeForDiagnostics(SourceLocation IncLoc,
2732	SourceLocation MLoc);
2733
2734	bool isRecordingPreamble() const {
2735	return PreambleConditionalStack.isRecording();
2736	}
2737
2738	bool hasRecordedPreamble() const {
2739	return PreambleConditionalStack.hasRecordedPreamble();
2740	}
2741
2742	ArrayRef<PPConditionalInfo> getPreambleConditionalStack() const {
2743	return PreambleConditionalStack.getStack();
2744	}
2745
2746	void setRecordedPreambleConditionalStack(ArrayRef<PPConditionalInfo> s) {
2747	PreambleConditionalStack.setStack(s);
2748	}
2749
2750	void setReplayablePreambleConditionalStack(
2751	ArrayRef<PPConditionalInfo> s, std::optional<PreambleSkipInfo> SkipInfo) {
2752	PreambleConditionalStack.startReplaying();
2753	PreambleConditionalStack.setStack(s);
2754	PreambleConditionalStack.SkipInfo = SkipInfo;
2755	}
2756
2757	std::optional<PreambleSkipInfo> getPreambleSkipInfo() const {
2758	return PreambleConditionalStack.SkipInfo;
2759	}
2760
2761	private:
2762	/// After processing predefined file, initialize the conditional stack from
2763	/// the preamble.
2764	void replayPreambleConditionalStack();
2765
2766	// Macro handling.
2767	void HandleDefineDirective(Token &Tok, bool ImmediatelyAfterHeaderGuard);
2768	void HandleUndefDirective();
2769
2770	// Conditional Inclusion.
2771	void HandleIfdefDirective(Token &Result, const Token &HashToken,
2772	bool isIfndef, bool ReadAnyTokensBeforeDirective);
2773	void HandleIfDirective(Token &IfToken, const Token &HashToken,
2774	bool ReadAnyTokensBeforeDirective);
2775	void HandleEndifDirective(Token &EndifToken);
2776	void HandleElseDirective(Token &Result, const Token &HashToken);
2777	void HandleElifFamilyDirective(Token &ElifToken, const Token &HashToken,
2778	tok::PPKeywordKind Kind);
2779
2780	// Pragmas.
2781	void HandlePragmaDirective(PragmaIntroducer Introducer);
2782
2783	public:
2784	void HandlePragmaOnce(Token &OnceTok);
2785	void HandlePragmaMark(Token &MarkTok);
2786	void HandlePragmaPoison();
2787	void HandlePragmaSystemHeader(Token &SysHeaderTok);
2788	void HandlePragmaDependency(Token &DependencyTok);
2789	void HandlePragmaPushMacro(Token &Tok);
2790	void HandlePragmaPopMacro(Token &Tok);
2791	void HandlePragmaIncludeAlias(Token &Tok);
2792	void HandlePragmaModuleBuild(Token &Tok);
2793	void HandlePragmaHdrstop(Token &Tok);
2794	IdentifierInfo *ParsePragmaPushOrPopMacro(Token &Tok);
2795
2796	// Return true and store the first token only if any CommentHandler
2797	// has inserted some tokens and getCommentRetentionState() is false.
2798	bool HandleComment(Token &result, SourceRange Comment);
2799
2800	/// A macro is used, update information about macros that need unused
2801	/// warnings.
2802	void markMacroAsUsed(MacroInfo *MI);
2803
2804	void addMacroDeprecationMsg(const IdentifierInfo *II, std::string Msg,
2805	SourceLocation AnnotationLoc) {
2806	auto Annotations = AnnotationInfos.find(Val: II);
2807	if (Annotations == AnnotationInfos.end())
2808	AnnotationInfos.insert(KV: std::make_pair(
2809	x&: II,
2810	y: MacroAnnotations::makeDeprecation(Loc: AnnotationLoc, Msg: std::move(Msg))));
2811	else
2812	Annotations->second.DeprecationInfo =
2813	MacroAnnotationInfo{.Location: AnnotationLoc, .Message: std::move(Msg)};
2814	}
2815
2816	void addRestrictExpansionMsg(const IdentifierInfo *II, std::string Msg,
2817	SourceLocation AnnotationLoc) {
2818	auto Annotations = AnnotationInfos.find(Val: II);
2819	if (Annotations == AnnotationInfos.end())
2820	AnnotationInfos.insert(
2821	KV: std::make_pair(x&: II, y: MacroAnnotations::makeRestrictExpansion(
2822	Loc: AnnotationLoc, Msg: std::move(Msg))));
2823	else
2824	Annotations->second.RestrictExpansionInfo =
2825	MacroAnnotationInfo{.Location: AnnotationLoc, .Message: std::move(Msg)};
2826	}
2827
2828	void addFinalLoc(const IdentifierInfo *II, SourceLocation AnnotationLoc) {
2829	auto Annotations = AnnotationInfos.find(Val: II);
2830	if (Annotations == AnnotationInfos.end())
2831	AnnotationInfos.insert(
2832	KV: std::make_pair(x&: II, y: MacroAnnotations::makeFinal(Loc: AnnotationLoc)));
2833	else
2834	Annotations->second.FinalAnnotationLoc = AnnotationLoc;
2835	}
2836
2837	const MacroAnnotations &getMacroAnnotations(const IdentifierInfo *II) const {
2838	return AnnotationInfos.find(Val: II)->second;
2839	}
2840
2841	void emitMacroExpansionWarnings(const Token &Identifier,
2842	bool IsIfnDef = false) const {
2843	IdentifierInfo *Info = Identifier.getIdentifierInfo();
2844	if (Info->isDeprecatedMacro())
2845	emitMacroDeprecationWarning(Identifier);
2846
2847	if (Info->isRestrictExpansion() &&
2848	!SourceMgr.isInMainFile(Loc: Identifier.getLocation()))
2849	emitRestrictExpansionWarning(Identifier);
2850
2851	if (!IsIfnDef) {
2852	if (Info->getName() == "INFINITY" && getLangOpts().NoHonorInfs)
2853	emitRestrictInfNaNWarning(Identifier, DiagSelection: 0);
2854	if (Info->getName() == "NAN" && getLangOpts().NoHonorNaNs)
2855	emitRestrictInfNaNWarning(Identifier, DiagSelection: 1);
2856	}
2857	}
2858
2859	static void processPathForFileMacro(SmallVectorImpl<char> &Path,
2860	const LangOptions &LangOpts,
2861	const TargetInfo &TI);
2862
2863	static void processPathToFileName(SmallVectorImpl<char> &FileName,
2864	const PresumedLoc &PLoc,
2865	const LangOptions &LangOpts,
2866	const TargetInfo &TI);
2867
2868	private:
2869	void emitMacroDeprecationWarning(const Token &Identifier) const;
2870	void emitRestrictExpansionWarning(const Token &Identifier) const;
2871	void emitFinalMacroWarning(const Token &Identifier, bool IsUndef) const;
2872	void emitRestrictInfNaNWarning(const Token &Identifier,
2873	unsigned DiagSelection) const;
2874
2875	/// This boolean state keeps track if the current scanned token (by this PP)
2876	/// is in an "-Wunsafe-buffer-usage" opt-out region. Assuming PP scans a
2877	/// translation unit in a linear order.
2878	bool InSafeBufferOptOutRegion = false;
2879
2880	/// Hold the start location of the current "-Wunsafe-buffer-usage" opt-out
2881	/// region if PP is currently in such a region. Hold undefined value
2882	/// otherwise.
2883	SourceLocation CurrentSafeBufferOptOutStart; // It is used to report the start location of an never-closed region.
2884
2885	// An ordered sequence of "-Wunsafe-buffer-usage" opt-out regions in one
2886	// translation unit. Each region is represented by a pair of start and end
2887	// locations. A region is "open" if its' start and end locations are
2888	// identical.
2889	SmallVector<std::pair<SourceLocation, SourceLocation>, 8> SafeBufferOptOutMap;
2890
2891	public:
2892	/// \return true iff the given `Loc` is in a "-Wunsafe-buffer-usage" opt-out
2893	/// region. This `Loc` must be a source location that has been pre-processed.
2894	bool isSafeBufferOptOut(const SourceManager&SourceMgr, const SourceLocation &Loc) const;
2895
2896	/// Alter the state of whether this PP currently is in a
2897	/// "-Wunsafe-buffer-usage" opt-out region.
2898	///
2899	/// \param isEnter true if this PP is entering a region; otherwise, this PP
2900	/// is exiting a region
2901	/// \param Loc the location of the entry or exit of a
2902	/// region
2903	/// \return true iff it is INVALID to enter or exit a region, i.e.,
2904	/// attempt to enter a region before exiting a previous region, or exiting a
2905	/// region that PP is not currently in.
2906	bool enterOrExitSafeBufferOptOutRegion(bool isEnter,
2907	const SourceLocation &Loc);
2908
2909	/// \return true iff this PP is currently in a "-Wunsafe-buffer-usage"
2910	/// opt-out region
2911	bool isPPInSafeBufferOptOutRegion();
2912
2913	/// \param StartLoc output argument. It will be set to the start location of
2914	/// the current "-Wunsafe-buffer-usage" opt-out region iff this function
2915	/// returns true.
2916	/// \return true iff this PP is currently in a "-Wunsafe-buffer-usage"
2917	/// opt-out region
2918	bool isPPInSafeBufferOptOutRegion(SourceLocation &StartLoc);
2919
2920	private:
2921	/// Helper functions to forward lexing to the actual lexer. They all share the
2922	/// same signature.
2923	static bool CLK_Lexer(Preprocessor &P, Token &Result) {
2924	return P.CurLexer->Lex(Result);
2925	}
2926	static bool CLK_TokenLexer(Preprocessor &P, Token &Result) {
2927	return P.CurTokenLexer->Lex(Tok&: Result);
2928	}
2929	static bool CLK_CachingLexer(Preprocessor &P, Token &Result) {
2930	P.CachingLex(Result);
2931	return true;
2932	}
2933	static bool CLK_DependencyDirectivesLexer(Preprocessor &P, Token &Result) {
2934	return P.CurLexer->LexDependencyDirectiveToken(Result);
2935	}
2936	static bool CLK_LexAfterModuleImport(Preprocessor &P, Token &Result) {
2937	return P.LexAfterModuleImport(Result);
2938	}
2939	};
2940
2941	/// Abstract base class that describes a handler that will receive
2942	/// source ranges for each of the comments encountered in the source file.
2943	class CommentHandler {
2944	public:
2945	virtual ~CommentHandler();
2946
2947	// The handler shall return true if it has pushed any tokens
2948	// to be read using e.g. EnterToken or EnterTokenStream.
2949	virtual bool HandleComment(Preprocessor &PP, SourceRange Comment) = 0;
2950	};
2951
2952	/// Abstract base class that describes a handler that will receive
2953	/// source ranges for empty lines encountered in the source file.
2954	class EmptylineHandler {
2955	public:
2956	virtual ~EmptylineHandler();
2957
2958	// The handler handles empty lines.
2959	virtual void HandleEmptyline(SourceRange Range) = 0;
2960	};
2961
2962	/// Registry of pragma handlers added by plugins
2963	using PragmaHandlerRegistry = llvm::Registry<PragmaHandler>;
2964
2965	} // namespace clang
2966
2967	#endif // LLVM_CLANG_LEX_PREPROCESSOR_H
2968