1 //===--- PPDirectives.cpp - Directive Handling for Preprocessor -----------===//
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 /// Implements # directive processing for the Preprocessor.
11 ///
12 //===----------------------------------------------------------------------===//
13 
14 #include "clang/Basic/CharInfo.h"
15 #include "clang/Basic/FileManager.h"
16 #include "clang/Basic/IdentifierTable.h"
17 #include "clang/Basic/LangOptions.h"
18 #include "clang/Basic/Module.h"
19 #include "clang/Basic/SourceLocation.h"
20 #include "clang/Basic/SourceManager.h"
21 #include "clang/Basic/TokenKinds.h"
22 #include "clang/Lex/CodeCompletionHandler.h"
23 #include "clang/Lex/HeaderSearch.h"
24 #include "clang/Lex/LexDiagnostic.h"
25 #include "clang/Lex/LiteralSupport.h"
26 #include "clang/Lex/MacroInfo.h"
27 #include "clang/Lex/ModuleLoader.h"
28 #include "clang/Lex/ModuleMap.h"
29 #include "clang/Lex/PPCallbacks.h"
30 #include "clang/Lex/Pragma.h"
31 #include "clang/Lex/Preprocessor.h"
32 #include "clang/Lex/PreprocessorOptions.h"
33 #include "clang/Lex/Token.h"
34 #include "clang/Lex/VariadicMacroSupport.h"
35 #include "llvm/ADT/ArrayRef.h"
36 #include "llvm/ADT/ScopeExit.h"
37 #include "llvm/ADT/SmallString.h"
38 #include "llvm/ADT/SmallVector.h"
39 #include "llvm/ADT/STLExtras.h"
40 #include "llvm/ADT/StringSwitch.h"
41 #include "llvm/ADT/StringRef.h"
42 #include "llvm/Support/AlignOf.h"
43 #include "llvm/Support/ErrorHandling.h"
44 #include "llvm/Support/Path.h"
45 #include <algorithm>
46 #include <cassert>
47 #include <cstring>
48 #include <new>
49 #include <string>
50 #include <utility>
51 
52 using namespace clang;
53 
54 //===----------------------------------------------------------------------===//
55 // Utility Methods for Preprocessor Directive Handling.
56 //===----------------------------------------------------------------------===//
57 
58 MacroInfo *Preprocessor::AllocateMacroInfo(SourceLocation L) {
59   auto *MIChain = new (BP) MacroInfoChain{L, MIChainHead};
60   MIChainHead = MIChain;
61   return &MIChain->MI;
62 }
63 
64 DefMacroDirective *Preprocessor::AllocateDefMacroDirective(MacroInfo *MI,
65                                                            SourceLocation Loc) {
66   return new (BP) DefMacroDirective(MI, Loc);
67 }
68 
69 UndefMacroDirective *
70 Preprocessor::AllocateUndefMacroDirective(SourceLocation UndefLoc) {
71   return new (BP) UndefMacroDirective(UndefLoc);
72 }
73 
74 VisibilityMacroDirective *
75 Preprocessor::AllocateVisibilityMacroDirective(SourceLocation Loc,
76                                                bool isPublic) {
77   return new (BP) VisibilityMacroDirective(Loc, isPublic);
78 }
79 
80 /// Read and discard all tokens remaining on the current line until
81 /// the tok::eod token is found.
82 SourceRange Preprocessor::DiscardUntilEndOfDirective() {
83   Token Tmp;
84   SourceRange Res;
85 
86   LexUnexpandedToken(Tmp);
87   Res.setBegin(Tmp.getLocation());
88   while (Tmp.isNot(tok::eod)) {
89     assert(Tmp.isNot(tok::eof) && "EOF seen while discarding directive tokens");
90     LexUnexpandedToken(Tmp);
91   }
92   Res.setEnd(Tmp.getLocation());
93   return Res;
94 }
95 
96 /// Enumerates possible cases of #define/#undef a reserved identifier.
97 enum MacroDiag {
98   MD_NoWarn,        //> Not a reserved identifier
99   MD_KeywordDef,    //> Macro hides keyword, enabled by default
100   MD_ReservedMacro  //> #define of #undef reserved id, disabled by default
101 };
102 
103 /// Checks if the specified identifier is reserved in the specified
104 /// language.
105 /// This function does not check if the identifier is a keyword.
106 static bool isReservedId(StringRef Text, const LangOptions &Lang) {
107   // C++ [macro.names], C11 7.1.3:
108   // All identifiers that begin with an underscore and either an uppercase
109   // letter or another underscore are always reserved for any use.
110   if (Text.size() >= 2 && Text[0] == '_' &&
111       (isUppercase(Text[1]) || Text[1] == '_'))
112       return true;
113   // C++ [global.names]
114   // Each name that contains a double underscore ... is reserved to the
115   // implementation for any use.
116   if (Lang.CPlusPlus) {
117     if (Text.find("__") != StringRef::npos)
118       return true;
119   }
120   return false;
121 }
122 
123 // The -fmodule-name option tells the compiler to textually include headers in
124 // the specified module, meaning clang won't build the specified module. This is
125 // useful in a number of situations, for instance, when building a library that
126 // vends a module map, one might want to avoid hitting intermediate build
127 // products containimg the the module map or avoid finding the system installed
128 // modulemap for that library.
129 static bool isForModuleBuilding(Module *M, StringRef CurrentModule,
130                                 StringRef ModuleName) {
131   StringRef TopLevelName = M->getTopLevelModuleName();
132 
133   // When building framework Foo, we wanna make sure that Foo *and* Foo_Private
134   // are textually included and no modules are built for both.
135   if (M->getTopLevelModule()->IsFramework && CurrentModule == ModuleName &&
136       !CurrentModule.endswith("_Private") && TopLevelName.endswith("_Private"))
137     TopLevelName = TopLevelName.drop_back(8);
138 
139   return TopLevelName == CurrentModule;
140 }
141 
142 static MacroDiag shouldWarnOnMacroDef(Preprocessor &PP, IdentifierInfo *II) {
143   const LangOptions &Lang = PP.getLangOpts();
144   StringRef Text = II->getName();
145   if (isReservedId(Text, Lang))
146     return MD_ReservedMacro;
147   if (II->isKeyword(Lang))
148     return MD_KeywordDef;
149   if (Lang.CPlusPlus11 && (Text.equals("override") || Text.equals("final")))
150     return MD_KeywordDef;
151   return MD_NoWarn;
152 }
153 
154 static MacroDiag shouldWarnOnMacroUndef(Preprocessor &PP, IdentifierInfo *II) {
155   const LangOptions &Lang = PP.getLangOpts();
156   StringRef Text = II->getName();
157   // Do not warn on keyword undef.  It is generally harmless and widely used.
158   if (isReservedId(Text, Lang))
159     return MD_ReservedMacro;
160   return MD_NoWarn;
161 }
162 
163 // Return true if we want to issue a diagnostic by default if we
164 // encounter this name in a #include with the wrong case. For now,
165 // this includes the standard C and C++ headers, Posix headers,
166 // and Boost headers. Improper case for these #includes is a
167 // potential portability issue.
168 static bool warnByDefaultOnWrongCase(StringRef Include) {
169   // If the first component of the path is "boost", treat this like a standard header
170   // for the purposes of diagnostics.
171   if (::llvm::sys::path::begin(Include)->equals_lower("boost"))
172     return true;
173 
174   // "condition_variable" is the longest standard header name at 18 characters.
175   // If the include file name is longer than that, it can't be a standard header.
176   static const size_t MaxStdHeaderNameLen = 18u;
177   if (Include.size() > MaxStdHeaderNameLen)
178     return false;
179 
180   // Lowercase and normalize the search string.
181   SmallString<32> LowerInclude{Include};
182   for (char &Ch : LowerInclude) {
183     // In the ASCII range?
184     if (static_cast<unsigned char>(Ch) > 0x7f)
185       return false; // Can't be a standard header
186     // ASCII lowercase:
187     if (Ch >= 'A' && Ch <= 'Z')
188       Ch += 'a' - 'A';
189     // Normalize path separators for comparison purposes.
190     else if (::llvm::sys::path::is_separator(Ch))
191       Ch = '/';
192   }
193 
194   // The standard C/C++ and Posix headers
195   return llvm::StringSwitch<bool>(LowerInclude)
196     // C library headers
197     .Cases("assert.h", "complex.h", "ctype.h", "errno.h", "fenv.h", true)
198     .Cases("float.h", "inttypes.h", "iso646.h", "limits.h", "locale.h", true)
199     .Cases("math.h", "setjmp.h", "signal.h", "stdalign.h", "stdarg.h", true)
200     .Cases("stdatomic.h", "stdbool.h", "stddef.h", "stdint.h", "stdio.h", true)
201     .Cases("stdlib.h", "stdnoreturn.h", "string.h", "tgmath.h", "threads.h", true)
202     .Cases("time.h", "uchar.h", "wchar.h", "wctype.h", true)
203 
204     // C++ headers for C library facilities
205     .Cases("cassert", "ccomplex", "cctype", "cerrno", "cfenv", true)
206     .Cases("cfloat", "cinttypes", "ciso646", "climits", "clocale", true)
207     .Cases("cmath", "csetjmp", "csignal", "cstdalign", "cstdarg", true)
208     .Cases("cstdbool", "cstddef", "cstdint", "cstdio", "cstdlib", true)
209     .Cases("cstring", "ctgmath", "ctime", "cuchar", "cwchar", true)
210     .Case("cwctype", true)
211 
212     // C++ library headers
213     .Cases("algorithm", "fstream", "list", "regex", "thread", true)
214     .Cases("array", "functional", "locale", "scoped_allocator", "tuple", true)
215     .Cases("atomic", "future", "map", "set", "type_traits", true)
216     .Cases("bitset", "initializer_list", "memory", "shared_mutex", "typeindex", true)
217     .Cases("chrono", "iomanip", "mutex", "sstream", "typeinfo", true)
218     .Cases("codecvt", "ios", "new", "stack", "unordered_map", true)
219     .Cases("complex", "iosfwd", "numeric", "stdexcept", "unordered_set", true)
220     .Cases("condition_variable", "iostream", "ostream", "streambuf", "utility", true)
221     .Cases("deque", "istream", "queue", "string", "valarray", true)
222     .Cases("exception", "iterator", "random", "strstream", "vector", true)
223     .Cases("forward_list", "limits", "ratio", "system_error", true)
224 
225     // POSIX headers (which aren't also C headers)
226     .Cases("aio.h", "arpa/inet.h", "cpio.h", "dirent.h", "dlfcn.h", true)
227     .Cases("fcntl.h", "fmtmsg.h", "fnmatch.h", "ftw.h", "glob.h", true)
228     .Cases("grp.h", "iconv.h", "langinfo.h", "libgen.h", "monetary.h", true)
229     .Cases("mqueue.h", "ndbm.h", "net/if.h", "netdb.h", "netinet/in.h", true)
230     .Cases("netinet/tcp.h", "nl_types.h", "poll.h", "pthread.h", "pwd.h", true)
231     .Cases("regex.h", "sched.h", "search.h", "semaphore.h", "spawn.h", true)
232     .Cases("strings.h", "stropts.h", "sys/ipc.h", "sys/mman.h", "sys/msg.h", true)
233     .Cases("sys/resource.h", "sys/select.h",  "sys/sem.h", "sys/shm.h", "sys/socket.h", true)
234     .Cases("sys/stat.h", "sys/statvfs.h", "sys/time.h", "sys/times.h", "sys/types.h", true)
235     .Cases("sys/uio.h", "sys/un.h", "sys/utsname.h", "sys/wait.h", "syslog.h", true)
236     .Cases("tar.h", "termios.h", "trace.h", "ulimit.h", true)
237     .Cases("unistd.h", "utime.h", "utmpx.h", "wordexp.h", true)
238     .Default(false);
239 }
240 
241 bool Preprocessor::CheckMacroName(Token &MacroNameTok, MacroUse isDefineUndef,
242                                   bool *ShadowFlag) {
243   // Missing macro name?
244   if (MacroNameTok.is(tok::eod))
245     return Diag(MacroNameTok, diag::err_pp_missing_macro_name);
246 
247   IdentifierInfo *II = MacroNameTok.getIdentifierInfo();
248   if (!II)
249     return Diag(MacroNameTok, diag::err_pp_macro_not_identifier);
250 
251   if (II->isCPlusPlusOperatorKeyword()) {
252     // C++ 2.5p2: Alternative tokens behave the same as its primary token
253     // except for their spellings.
254     Diag(MacroNameTok, getLangOpts().MicrosoftExt
255                            ? diag::ext_pp_operator_used_as_macro_name
256                            : diag::err_pp_operator_used_as_macro_name)
257         << II << MacroNameTok.getKind();
258     // Allow #defining |and| and friends for Microsoft compatibility or
259     // recovery when legacy C headers are included in C++.
260   }
261 
262   if ((isDefineUndef != MU_Other) && II->getPPKeywordID() == tok::pp_defined) {
263     // Error if defining "defined": C99 6.10.8/4, C++ [cpp.predefined]p4.
264     return Diag(MacroNameTok, diag::err_defined_macro_name);
265   }
266 
267   if (isDefineUndef == MU_Undef) {
268     auto *MI = getMacroInfo(II);
269     if (MI && MI->isBuiltinMacro()) {
270       // Warn if undefining "__LINE__" and other builtins, per C99 6.10.8/4
271       // and C++ [cpp.predefined]p4], but allow it as an extension.
272       Diag(MacroNameTok, diag::ext_pp_undef_builtin_macro);
273     }
274   }
275 
276   // If defining/undefining reserved identifier or a keyword, we need to issue
277   // a warning.
278   SourceLocation MacroNameLoc = MacroNameTok.getLocation();
279   if (ShadowFlag)
280     *ShadowFlag = false;
281   if (!SourceMgr.isInSystemHeader(MacroNameLoc) &&
282       (SourceMgr.getBufferName(MacroNameLoc) != "<built-in>")) {
283     MacroDiag D = MD_NoWarn;
284     if (isDefineUndef == MU_Define) {
285       D = shouldWarnOnMacroDef(*this, II);
286     }
287     else if (isDefineUndef == MU_Undef)
288       D = shouldWarnOnMacroUndef(*this, II);
289     if (D == MD_KeywordDef) {
290       // We do not want to warn on some patterns widely used in configuration
291       // scripts.  This requires analyzing next tokens, so do not issue warnings
292       // now, only inform caller.
293       if (ShadowFlag)
294         *ShadowFlag = true;
295     }
296     if (D == MD_ReservedMacro)
297       Diag(MacroNameTok, diag::warn_pp_macro_is_reserved_id);
298   }
299 
300   // Okay, we got a good identifier.
301   return false;
302 }
303 
304 /// Lex and validate a macro name, which occurs after a
305 /// \#define or \#undef.
306 ///
307 /// This sets the token kind to eod and discards the rest of the macro line if
308 /// the macro name is invalid.
309 ///
310 /// \param MacroNameTok Token that is expected to be a macro name.
311 /// \param isDefineUndef Context in which macro is used.
312 /// \param ShadowFlag Points to a flag that is set if macro shadows a keyword.
313 void Preprocessor::ReadMacroName(Token &MacroNameTok, MacroUse isDefineUndef,
314                                  bool *ShadowFlag) {
315   // Read the token, don't allow macro expansion on it.
316   LexUnexpandedToken(MacroNameTok);
317 
318   if (MacroNameTok.is(tok::code_completion)) {
319     if (CodeComplete)
320       CodeComplete->CodeCompleteMacroName(isDefineUndef == MU_Define);
321     setCodeCompletionReached();
322     LexUnexpandedToken(MacroNameTok);
323   }
324 
325   if (!CheckMacroName(MacroNameTok, isDefineUndef, ShadowFlag))
326     return;
327 
328   // Invalid macro name, read and discard the rest of the line and set the
329   // token kind to tok::eod if necessary.
330   if (MacroNameTok.isNot(tok::eod)) {
331     MacroNameTok.setKind(tok::eod);
332     DiscardUntilEndOfDirective();
333   }
334 }
335 
336 /// Ensure that the next token is a tok::eod token.
337 ///
338 /// If not, emit a diagnostic and consume up until the eod.  If EnableMacros is
339 /// true, then we consider macros that expand to zero tokens as being ok.
340 ///
341 /// Returns the location of the end of the directive.
342 SourceLocation Preprocessor::CheckEndOfDirective(const char *DirType,
343                                                  bool EnableMacros) {
344   Token Tmp;
345   // Lex unexpanded tokens for most directives: macros might expand to zero
346   // tokens, causing us to miss diagnosing invalid lines.  Some directives (like
347   // #line) allow empty macros.
348   if (EnableMacros)
349     Lex(Tmp);
350   else
351     LexUnexpandedToken(Tmp);
352 
353   // There should be no tokens after the directive, but we allow them as an
354   // extension.
355   while (Tmp.is(tok::comment))  // Skip comments in -C mode.
356     LexUnexpandedToken(Tmp);
357 
358   if (Tmp.is(tok::eod))
359     return Tmp.getLocation();
360 
361   // Add a fixit in GNU/C99/C++ mode.  Don't offer a fixit for strict-C89,
362   // or if this is a macro-style preprocessing directive, because it is more
363   // trouble than it is worth to insert /**/ and check that there is no /**/
364   // in the range also.
365   FixItHint Hint;
366   if ((LangOpts.GNUMode || LangOpts.C99 || LangOpts.CPlusPlus) &&
367       !CurTokenLexer)
368     Hint = FixItHint::CreateInsertion(Tmp.getLocation(),"//");
369   Diag(Tmp, diag::ext_pp_extra_tokens_at_eol) << DirType << Hint;
370   return DiscardUntilEndOfDirective().getEnd();
371 }
372 
373 Optional<unsigned> Preprocessor::getSkippedRangeForExcludedConditionalBlock(
374     SourceLocation HashLoc) {
375   if (!ExcludedConditionalDirectiveSkipMappings)
376     return None;
377   if (!HashLoc.isFileID())
378     return None;
379 
380   std::pair<FileID, unsigned> HashFileOffset =
381       SourceMgr.getDecomposedLoc(HashLoc);
382   const llvm::MemoryBuffer *Buf = SourceMgr.getBuffer(HashFileOffset.first);
383   auto It = ExcludedConditionalDirectiveSkipMappings->find(Buf);
384   if (It == ExcludedConditionalDirectiveSkipMappings->end())
385     return None;
386 
387   const PreprocessorSkippedRangeMapping &SkippedRanges = *It->getSecond();
388   // Check if the offset of '#' is mapped in the skipped ranges.
389   auto MappingIt = SkippedRanges.find(HashFileOffset.second);
390   if (MappingIt == SkippedRanges.end())
391     return None;
392 
393   unsigned BytesToSkip = MappingIt->getSecond();
394   unsigned CurLexerBufferOffset = CurLexer->getCurrentBufferOffset();
395   assert(CurLexerBufferOffset >= HashFileOffset.second &&
396          "lexer is before the hash?");
397   // Take into account the fact that the lexer has already advanced, so the
398   // number of bytes to skip must be adjusted.
399   unsigned LengthDiff = CurLexerBufferOffset - HashFileOffset.second;
400   assert(BytesToSkip >= LengthDiff && "lexer is after the skipped range?");
401   return BytesToSkip - LengthDiff;
402 }
403 
404 /// SkipExcludedConditionalBlock - We just read a \#if or related directive and
405 /// decided that the subsequent tokens are in the \#if'd out portion of the
406 /// file.  Lex the rest of the file, until we see an \#endif.  If
407 /// FoundNonSkipPortion is true, then we have already emitted code for part of
408 /// this \#if directive, so \#else/\#elif blocks should never be entered.
409 /// If ElseOk is true, then \#else directives are ok, if not, then we have
410 /// already seen one so a \#else directive is a duplicate.  When this returns,
411 /// the caller can lex the first valid token.
412 void Preprocessor::SkipExcludedConditionalBlock(SourceLocation HashTokenLoc,
413                                                 SourceLocation IfTokenLoc,
414                                                 bool FoundNonSkipPortion,
415                                                 bool FoundElse,
416                                                 SourceLocation ElseLoc) {
417   ++NumSkipped;
418   assert(!CurTokenLexer && CurPPLexer && "Lexing a macro, not a file?");
419 
420   if (PreambleConditionalStack.reachedEOFWhileSkipping())
421     PreambleConditionalStack.clearSkipInfo();
422   else
423     CurPPLexer->pushConditionalLevel(IfTokenLoc, /*isSkipping*/ false,
424                                      FoundNonSkipPortion, FoundElse);
425 
426   // Enter raw mode to disable identifier lookup (and thus macro expansion),
427   // disabling warnings, etc.
428   CurPPLexer->LexingRawMode = true;
429   Token Tok;
430   if (auto SkipLength =
431           getSkippedRangeForExcludedConditionalBlock(HashTokenLoc)) {
432     // Skip to the next '#endif' / '#else' / '#elif'.
433     CurLexer->skipOver(*SkipLength);
434   }
435   while (true) {
436     CurLexer->Lex(Tok);
437 
438     if (Tok.is(tok::code_completion)) {
439       if (CodeComplete)
440         CodeComplete->CodeCompleteInConditionalExclusion();
441       setCodeCompletionReached();
442       continue;
443     }
444 
445     // If this is the end of the buffer, we have an error.
446     if (Tok.is(tok::eof)) {
447       // We don't emit errors for unterminated conditionals here,
448       // Lexer::LexEndOfFile can do that properly.
449       // Just return and let the caller lex after this #include.
450       if (PreambleConditionalStack.isRecording())
451         PreambleConditionalStack.SkipInfo.emplace(
452             HashTokenLoc, IfTokenLoc, FoundNonSkipPortion, FoundElse, ElseLoc);
453       break;
454     }
455 
456     // If this token is not a preprocessor directive, just skip it.
457     if (Tok.isNot(tok::hash) || !Tok.isAtStartOfLine())
458       continue;
459 
460     // We just parsed a # character at the start of a line, so we're in
461     // directive mode.  Tell the lexer this so any newlines we see will be
462     // converted into an EOD token (this terminates the macro).
463     CurPPLexer->ParsingPreprocessorDirective = true;
464     if (CurLexer) CurLexer->SetKeepWhitespaceMode(false);
465 
466 
467     // Read the next token, the directive flavor.
468     LexUnexpandedToken(Tok);
469 
470     // If this isn't an identifier directive (e.g. is "# 1\n" or "#\n", or
471     // something bogus), skip it.
472     if (Tok.isNot(tok::raw_identifier)) {
473       CurPPLexer->ParsingPreprocessorDirective = false;
474       // Restore comment saving mode.
475       if (CurLexer) CurLexer->resetExtendedTokenMode();
476       continue;
477     }
478 
479     // If the first letter isn't i or e, it isn't intesting to us.  We know that
480     // this is safe in the face of spelling differences, because there is no way
481     // to spell an i/e in a strange way that is another letter.  Skipping this
482     // allows us to avoid looking up the identifier info for #define/#undef and
483     // other common directives.
484     StringRef RI = Tok.getRawIdentifier();
485 
486     char FirstChar = RI[0];
487     if (FirstChar >= 'a' && FirstChar <= 'z' &&
488         FirstChar != 'i' && FirstChar != 'e') {
489       CurPPLexer->ParsingPreprocessorDirective = false;
490       // Restore comment saving mode.
491       if (CurLexer) CurLexer->resetExtendedTokenMode();
492       continue;
493     }
494 
495     // Get the identifier name without trigraphs or embedded newlines.  Note
496     // that we can't use Tok.getIdentifierInfo() because its lookup is disabled
497     // when skipping.
498     char DirectiveBuf[20];
499     StringRef Directive;
500     if (!Tok.needsCleaning() && RI.size() < 20) {
501       Directive = RI;
502     } else {
503       std::string DirectiveStr = getSpelling(Tok);
504       size_t IdLen = DirectiveStr.size();
505       if (IdLen >= 20) {
506         CurPPLexer->ParsingPreprocessorDirective = false;
507         // Restore comment saving mode.
508         if (CurLexer) CurLexer->resetExtendedTokenMode();
509         continue;
510       }
511       memcpy(DirectiveBuf, &DirectiveStr[0], IdLen);
512       Directive = StringRef(DirectiveBuf, IdLen);
513     }
514 
515     if (Directive.startswith("if")) {
516       StringRef Sub = Directive.substr(2);
517       if (Sub.empty() ||   // "if"
518           Sub == "def" ||   // "ifdef"
519           Sub == "ndef") {  // "ifndef"
520         // We know the entire #if/#ifdef/#ifndef block will be skipped, don't
521         // bother parsing the condition.
522         DiscardUntilEndOfDirective();
523         CurPPLexer->pushConditionalLevel(Tok.getLocation(), /*wasskipping*/true,
524                                        /*foundnonskip*/false,
525                                        /*foundelse*/false);
526       }
527     } else if (Directive[0] == 'e') {
528       StringRef Sub = Directive.substr(1);
529       if (Sub == "ndif") {  // "endif"
530         PPConditionalInfo CondInfo;
531         CondInfo.WasSkipping = true; // Silence bogus warning.
532         bool InCond = CurPPLexer->popConditionalLevel(CondInfo);
533         (void)InCond;  // Silence warning in no-asserts mode.
534         assert(!InCond && "Can't be skipping if not in a conditional!");
535 
536         // If we popped the outermost skipping block, we're done skipping!
537         if (!CondInfo.WasSkipping) {
538           // Restore the value of LexingRawMode so that trailing comments
539           // are handled correctly, if we've reached the outermost block.
540           CurPPLexer->LexingRawMode = false;
541           CheckEndOfDirective("endif");
542           CurPPLexer->LexingRawMode = true;
543           if (Callbacks)
544             Callbacks->Endif(Tok.getLocation(), CondInfo.IfLoc);
545           break;
546         } else {
547           DiscardUntilEndOfDirective();
548         }
549       } else if (Sub == "lse") { // "else".
550         // #else directive in a skipping conditional.  If not in some other
551         // skipping conditional, and if #else hasn't already been seen, enter it
552         // as a non-skipping conditional.
553         PPConditionalInfo &CondInfo = CurPPLexer->peekConditionalLevel();
554 
555         // If this is a #else with a #else before it, report the error.
556         if (CondInfo.FoundElse) Diag(Tok, diag::pp_err_else_after_else);
557 
558         // Note that we've seen a #else in this conditional.
559         CondInfo.FoundElse = true;
560 
561         // If the conditional is at the top level, and the #if block wasn't
562         // entered, enter the #else block now.
563         if (!CondInfo.WasSkipping && !CondInfo.FoundNonSkip) {
564           CondInfo.FoundNonSkip = true;
565           // Restore the value of LexingRawMode so that trailing comments
566           // are handled correctly.
567           CurPPLexer->LexingRawMode = false;
568           CheckEndOfDirective("else");
569           CurPPLexer->LexingRawMode = true;
570           if (Callbacks)
571             Callbacks->Else(Tok.getLocation(), CondInfo.IfLoc);
572           break;
573         } else {
574           DiscardUntilEndOfDirective();  // C99 6.10p4.
575         }
576       } else if (Sub == "lif") {  // "elif".
577         PPConditionalInfo &CondInfo = CurPPLexer->peekConditionalLevel();
578 
579         // If this is a #elif with a #else before it, report the error.
580         if (CondInfo.FoundElse) Diag(Tok, diag::pp_err_elif_after_else);
581 
582         // If this is in a skipping block or if we're already handled this #if
583         // block, don't bother parsing the condition.
584         if (CondInfo.WasSkipping || CondInfo.FoundNonSkip) {
585           DiscardUntilEndOfDirective();
586         } else {
587           // Restore the value of LexingRawMode so that identifiers are
588           // looked up, etc, inside the #elif expression.
589           assert(CurPPLexer->LexingRawMode && "We have to be skipping here!");
590           CurPPLexer->LexingRawMode = false;
591           IdentifierInfo *IfNDefMacro = nullptr;
592           DirectiveEvalResult DER = EvaluateDirectiveExpression(IfNDefMacro);
593           const bool CondValue = DER.Conditional;
594           CurPPLexer->LexingRawMode = true;
595           if (Callbacks) {
596             Callbacks->Elif(
597                 Tok.getLocation(), DER.ExprRange,
598                 (CondValue ? PPCallbacks::CVK_True : PPCallbacks::CVK_False),
599                 CondInfo.IfLoc);
600           }
601           // If this condition is true, enter it!
602           if (CondValue) {
603             CondInfo.FoundNonSkip = true;
604             break;
605           }
606         }
607       }
608     }
609 
610     CurPPLexer->ParsingPreprocessorDirective = false;
611     // Restore comment saving mode.
612     if (CurLexer) CurLexer->resetExtendedTokenMode();
613   }
614 
615   // Finally, if we are out of the conditional (saw an #endif or ran off the end
616   // of the file, just stop skipping and return to lexing whatever came after
617   // the #if block.
618   CurPPLexer->LexingRawMode = false;
619 
620   // The last skipped range isn't actually skipped yet if it's truncated
621   // by the end of the preamble; we'll resume parsing after the preamble.
622   if (Callbacks && (Tok.isNot(tok::eof) || !isRecordingPreamble()))
623     Callbacks->SourceRangeSkipped(
624         SourceRange(HashTokenLoc, CurPPLexer->getSourceLocation()),
625         Tok.getLocation());
626 }
627 
628 Module *Preprocessor::getModuleForLocation(SourceLocation Loc) {
629   if (!SourceMgr.isInMainFile(Loc)) {
630     // Try to determine the module of the include directive.
631     // FIXME: Look into directly passing the FileEntry from LookupFile instead.
632     FileID IDOfIncl = SourceMgr.getFileID(SourceMgr.getExpansionLoc(Loc));
633     if (const FileEntry *EntryOfIncl = SourceMgr.getFileEntryForID(IDOfIncl)) {
634       // The include comes from an included file.
635       return HeaderInfo.getModuleMap()
636           .findModuleForHeader(EntryOfIncl)
637           .getModule();
638     }
639   }
640 
641   // This is either in the main file or not in a file at all. It belongs
642   // to the current module, if there is one.
643   return getLangOpts().CurrentModule.empty()
644              ? nullptr
645              : HeaderInfo.lookupModule(getLangOpts().CurrentModule);
646 }
647 
648 const FileEntry *
649 Preprocessor::getModuleHeaderToIncludeForDiagnostics(SourceLocation IncLoc,
650                                                      Module *M,
651                                                      SourceLocation Loc) {
652   assert(M && "no module to include");
653 
654   // If the context is the global module fragment of some module, we never
655   // want to return that file; instead, we want the innermost include-guarded
656   // header that it included.
657   bool InGlobalModuleFragment = M->Kind == Module::GlobalModuleFragment;
658 
659   // If we have a module import syntax, we shouldn't include a header to
660   // make a particular module visible.
661   if ((getLangOpts().ObjC || getLangOpts().CPlusPlusModules ||
662        getLangOpts().ModulesTS) &&
663       !InGlobalModuleFragment)
664     return nullptr;
665 
666   Module *TopM = M->getTopLevelModule();
667   Module *IncM = getModuleForLocation(IncLoc);
668 
669   // Walk up through the include stack, looking through textual headers of M
670   // until we hit a non-textual header that we can #include. (We assume textual
671   // headers of a module with non-textual headers aren't meant to be used to
672   // import entities from the module.)
673   auto &SM = getSourceManager();
674   while (!Loc.isInvalid() && !SM.isInMainFile(Loc)) {
675     auto ID = SM.getFileID(SM.getExpansionLoc(Loc));
676     auto *FE = SM.getFileEntryForID(ID);
677     if (!FE)
678       break;
679 
680     if (InGlobalModuleFragment) {
681       if (getHeaderSearchInfo().isFileMultipleIncludeGuarded(FE))
682         return FE;
683       Loc = SM.getIncludeLoc(ID);
684       continue;
685     }
686 
687     bool InTextualHeader = false;
688     for (auto Header : HeaderInfo.getModuleMap().findAllModulesForHeader(FE)) {
689       if (!Header.getModule()->isSubModuleOf(TopM))
690         continue;
691 
692       if (!(Header.getRole() & ModuleMap::TextualHeader)) {
693         // If this is an accessible, non-textual header of M's top-level module
694         // that transitively includes the given location and makes the
695         // corresponding module visible, this is the thing to #include.
696         if (Header.isAccessibleFrom(IncM))
697           return FE;
698 
699         // It's in a private header; we can't #include it.
700         // FIXME: If there's a public header in some module that re-exports it,
701         // then we could suggest including that, but it's not clear that's the
702         // expected way to make this entity visible.
703         continue;
704       }
705 
706       InTextualHeader = true;
707     }
708 
709     if (!InTextualHeader)
710       break;
711 
712     Loc = SM.getIncludeLoc(ID);
713   }
714 
715   return nullptr;
716 }
717 
718 Optional<FileEntryRef> Preprocessor::LookupFile(
719     SourceLocation FilenameLoc, StringRef Filename, bool isAngled,
720     const DirectoryLookup *FromDir, const FileEntry *FromFile,
721     const DirectoryLookup *&CurDir, SmallVectorImpl<char> *SearchPath,
722     SmallVectorImpl<char> *RelativePath,
723     ModuleMap::KnownHeader *SuggestedModule, bool *IsMapped,
724     bool *IsFrameworkFound, bool SkipCache) {
725   Module *RequestingModule = getModuleForLocation(FilenameLoc);
726   bool RequestingModuleIsModuleInterface = !SourceMgr.isInMainFile(FilenameLoc);
727 
728   // If the header lookup mechanism may be relative to the current inclusion
729   // stack, record the parent #includes.
730   SmallVector<std::pair<const FileEntry *, const DirectoryEntry *>, 16>
731       Includers;
732   bool BuildSystemModule = false;
733   if (!FromDir && !FromFile) {
734     FileID FID = getCurrentFileLexer()->getFileID();
735     const FileEntry *FileEnt = SourceMgr.getFileEntryForID(FID);
736 
737     // If there is no file entry associated with this file, it must be the
738     // predefines buffer or the module includes buffer. Any other file is not
739     // lexed with a normal lexer, so it won't be scanned for preprocessor
740     // directives.
741     //
742     // If we have the predefines buffer, resolve #include references (which come
743     // from the -include command line argument) from the current working
744     // directory instead of relative to the main file.
745     //
746     // If we have the module includes buffer, resolve #include references (which
747     // come from header declarations in the module map) relative to the module
748     // map file.
749     if (!FileEnt) {
750       if (FID == SourceMgr.getMainFileID() && MainFileDir) {
751         Includers.push_back(std::make_pair(nullptr, MainFileDir));
752         BuildSystemModule = getCurrentModule()->IsSystem;
753       } else if ((FileEnt =
754                     SourceMgr.getFileEntryForID(SourceMgr.getMainFileID())))
755         Includers.push_back(std::make_pair(FileEnt, *FileMgr.getDirectory(".")));
756     } else {
757       Includers.push_back(std::make_pair(FileEnt, FileEnt->getDir()));
758     }
759 
760     // MSVC searches the current include stack from top to bottom for
761     // headers included by quoted include directives.
762     // See: http://msdn.microsoft.com/en-us/library/36k2cdd4.aspx
763     if (LangOpts.MSVCCompat && !isAngled) {
764       for (IncludeStackInfo &ISEntry : llvm::reverse(IncludeMacroStack)) {
765         if (IsFileLexer(ISEntry))
766           if ((FileEnt = ISEntry.ThePPLexer->getFileEntry()))
767             Includers.push_back(std::make_pair(FileEnt, FileEnt->getDir()));
768       }
769     }
770   }
771 
772   CurDir = CurDirLookup;
773 
774   if (FromFile) {
775     // We're supposed to start looking from after a particular file. Search
776     // the include path until we find that file or run out of files.
777     const DirectoryLookup *TmpCurDir = CurDir;
778     const DirectoryLookup *TmpFromDir = nullptr;
779     while (Optional<FileEntryRef> FE = HeaderInfo.LookupFile(
780                Filename, FilenameLoc, isAngled, TmpFromDir, TmpCurDir,
781                Includers, SearchPath, RelativePath, RequestingModule,
782                SuggestedModule, /*IsMapped=*/nullptr,
783                /*IsFrameworkFound=*/nullptr, SkipCache)) {
784       // Keep looking as if this file did a #include_next.
785       TmpFromDir = TmpCurDir;
786       ++TmpFromDir;
787       if (&FE->getFileEntry() == FromFile) {
788         // Found it.
789         FromDir = TmpFromDir;
790         CurDir = TmpCurDir;
791         break;
792       }
793     }
794   }
795 
796   // Do a standard file entry lookup.
797   Optional<FileEntryRef> FE = HeaderInfo.LookupFile(
798       Filename, FilenameLoc, isAngled, FromDir, CurDir, Includers, SearchPath,
799       RelativePath, RequestingModule, SuggestedModule, IsMapped,
800       IsFrameworkFound, SkipCache, BuildSystemModule);
801   if (FE) {
802     if (SuggestedModule && !LangOpts.AsmPreprocessor)
803       HeaderInfo.getModuleMap().diagnoseHeaderInclusion(
804           RequestingModule, RequestingModuleIsModuleInterface, FilenameLoc,
805           Filename, &FE->getFileEntry());
806     return FE;
807   }
808 
809   const FileEntry *CurFileEnt;
810   // Otherwise, see if this is a subframework header.  If so, this is relative
811   // to one of the headers on the #include stack.  Walk the list of the current
812   // headers on the #include stack and pass them to HeaderInfo.
813   if (IsFileLexer()) {
814     if ((CurFileEnt = CurPPLexer->getFileEntry())) {
815       if (Optional<FileEntryRef> FE = HeaderInfo.LookupSubframeworkHeader(
816               Filename, CurFileEnt, SearchPath, RelativePath, RequestingModule,
817               SuggestedModule)) {
818         if (SuggestedModule && !LangOpts.AsmPreprocessor)
819           HeaderInfo.getModuleMap().diagnoseHeaderInclusion(
820               RequestingModule, RequestingModuleIsModuleInterface, FilenameLoc,
821               Filename, &FE->getFileEntry());
822         return FE;
823       }
824     }
825   }
826 
827   for (IncludeStackInfo &ISEntry : llvm::reverse(IncludeMacroStack)) {
828     if (IsFileLexer(ISEntry)) {
829       if ((CurFileEnt = ISEntry.ThePPLexer->getFileEntry())) {
830         if (Optional<FileEntryRef> FE = HeaderInfo.LookupSubframeworkHeader(
831                 Filename, CurFileEnt, SearchPath, RelativePath,
832                 RequestingModule, SuggestedModule)) {
833           if (SuggestedModule && !LangOpts.AsmPreprocessor)
834             HeaderInfo.getModuleMap().diagnoseHeaderInclusion(
835                 RequestingModule, RequestingModuleIsModuleInterface,
836                 FilenameLoc, Filename, &FE->getFileEntry());
837           return FE;
838         }
839       }
840     }
841   }
842 
843   // Otherwise, we really couldn't find the file.
844   return None;
845 }
846 
847 //===----------------------------------------------------------------------===//
848 // Preprocessor Directive Handling.
849 //===----------------------------------------------------------------------===//
850 
851 class Preprocessor::ResetMacroExpansionHelper {
852 public:
853   ResetMacroExpansionHelper(Preprocessor *pp)
854     : PP(pp), save(pp->DisableMacroExpansion) {
855     if (pp->MacroExpansionInDirectivesOverride)
856       pp->DisableMacroExpansion = false;
857   }
858 
859   ~ResetMacroExpansionHelper() {
860     PP->DisableMacroExpansion = save;
861   }
862 
863 private:
864   Preprocessor *PP;
865   bool save;
866 };
867 
868 /// Process a directive while looking for the through header or a #pragma
869 /// hdrstop. The following directives are handled:
870 /// #include (to check if it is the through header)
871 /// #define (to warn about macros that don't match the PCH)
872 /// #pragma (to check for pragma hdrstop).
873 /// All other directives are completely discarded.
874 void Preprocessor::HandleSkippedDirectiveWhileUsingPCH(Token &Result,
875                                                        SourceLocation HashLoc) {
876   if (const IdentifierInfo *II = Result.getIdentifierInfo()) {
877     if (II->getPPKeywordID() == tok::pp_define) {
878       return HandleDefineDirective(Result,
879                                    /*ImmediatelyAfterHeaderGuard=*/false);
880     }
881     if (SkippingUntilPCHThroughHeader &&
882         II->getPPKeywordID() == tok::pp_include) {
883       return HandleIncludeDirective(HashLoc, Result);
884     }
885     if (SkippingUntilPragmaHdrStop && II->getPPKeywordID() == tok::pp_pragma) {
886       Lex(Result);
887       auto *II = Result.getIdentifierInfo();
888       if (II && II->getName() == "hdrstop")
889         return HandlePragmaHdrstop(Result);
890     }
891   }
892   DiscardUntilEndOfDirective();
893 }
894 
895 /// HandleDirective - This callback is invoked when the lexer sees a # token
896 /// at the start of a line.  This consumes the directive, modifies the
897 /// lexer/preprocessor state, and advances the lexer(s) so that the next token
898 /// read is the correct one.
899 void Preprocessor::HandleDirective(Token &Result) {
900   // FIXME: Traditional: # with whitespace before it not recognized by K&R?
901 
902   // We just parsed a # character at the start of a line, so we're in directive
903   // mode.  Tell the lexer this so any newlines we see will be converted into an
904   // EOD token (which terminates the directive).
905   CurPPLexer->ParsingPreprocessorDirective = true;
906   if (CurLexer) CurLexer->SetKeepWhitespaceMode(false);
907 
908   bool ImmediatelyAfterTopLevelIfndef =
909       CurPPLexer->MIOpt.getImmediatelyAfterTopLevelIfndef();
910   CurPPLexer->MIOpt.resetImmediatelyAfterTopLevelIfndef();
911 
912   ++NumDirectives;
913 
914   // We are about to read a token.  For the multiple-include optimization FA to
915   // work, we have to remember if we had read any tokens *before* this
916   // pp-directive.
917   bool ReadAnyTokensBeforeDirective =CurPPLexer->MIOpt.getHasReadAnyTokensVal();
918 
919   // Save the '#' token in case we need to return it later.
920   Token SavedHash = Result;
921 
922   // Read the next token, the directive flavor.  This isn't expanded due to
923   // C99 6.10.3p8.
924   LexUnexpandedToken(Result);
925 
926   // C99 6.10.3p11: Is this preprocessor directive in macro invocation?  e.g.:
927   //   #define A(x) #x
928   //   A(abc
929   //     #warning blah
930   //   def)
931   // If so, the user is relying on undefined behavior, emit a diagnostic. Do
932   // not support this for #include-like directives, since that can result in
933   // terrible diagnostics, and does not work in GCC.
934   if (InMacroArgs) {
935     if (IdentifierInfo *II = Result.getIdentifierInfo()) {
936       switch (II->getPPKeywordID()) {
937       case tok::pp_include:
938       case tok::pp_import:
939       case tok::pp_include_next:
940       case tok::pp___include_macros:
941       case tok::pp_pragma:
942         Diag(Result, diag::err_embedded_directive) << II->getName();
943         Diag(*ArgMacro, diag::note_macro_expansion_here)
944             << ArgMacro->getIdentifierInfo();
945         DiscardUntilEndOfDirective();
946         return;
947       default:
948         break;
949       }
950     }
951     Diag(Result, diag::ext_embedded_directive);
952   }
953 
954   // Temporarily enable macro expansion if set so
955   // and reset to previous state when returning from this function.
956   ResetMacroExpansionHelper helper(this);
957 
958   if (SkippingUntilPCHThroughHeader || SkippingUntilPragmaHdrStop)
959     return HandleSkippedDirectiveWhileUsingPCH(Result, SavedHash.getLocation());
960 
961   switch (Result.getKind()) {
962   case tok::eod:
963     return;   // null directive.
964   case tok::code_completion:
965     if (CodeComplete)
966       CodeComplete->CodeCompleteDirective(
967                                     CurPPLexer->getConditionalStackDepth() > 0);
968     setCodeCompletionReached();
969     return;
970   case tok::numeric_constant:  // # 7  GNU line marker directive.
971     if (getLangOpts().AsmPreprocessor)
972       break;  // # 4 is not a preprocessor directive in .S files.
973     return HandleDigitDirective(Result);
974   default:
975     IdentifierInfo *II = Result.getIdentifierInfo();
976     if (!II) break; // Not an identifier.
977 
978     // Ask what the preprocessor keyword ID is.
979     switch (II->getPPKeywordID()) {
980     default: break;
981     // C99 6.10.1 - Conditional Inclusion.
982     case tok::pp_if:
983       return HandleIfDirective(Result, SavedHash, ReadAnyTokensBeforeDirective);
984     case tok::pp_ifdef:
985       return HandleIfdefDirective(Result, SavedHash, false,
986                                   true /*not valid for miopt*/);
987     case tok::pp_ifndef:
988       return HandleIfdefDirective(Result, SavedHash, true,
989                                   ReadAnyTokensBeforeDirective);
990     case tok::pp_elif:
991       return HandleElifDirective(Result, SavedHash);
992     case tok::pp_else:
993       return HandleElseDirective(Result, SavedHash);
994     case tok::pp_endif:
995       return HandleEndifDirective(Result);
996 
997     // C99 6.10.2 - Source File Inclusion.
998     case tok::pp_include:
999       // Handle #include.
1000       return HandleIncludeDirective(SavedHash.getLocation(), Result);
1001     case tok::pp___include_macros:
1002       // Handle -imacros.
1003       return HandleIncludeMacrosDirective(SavedHash.getLocation(), Result);
1004 
1005     // C99 6.10.3 - Macro Replacement.
1006     case tok::pp_define:
1007       return HandleDefineDirective(Result, ImmediatelyAfterTopLevelIfndef);
1008     case tok::pp_undef:
1009       return HandleUndefDirective();
1010 
1011     // C99 6.10.4 - Line Control.
1012     case tok::pp_line:
1013       return HandleLineDirective();
1014 
1015     // C99 6.10.5 - Error Directive.
1016     case tok::pp_error:
1017       return HandleUserDiagnosticDirective(Result, false);
1018 
1019     // C99 6.10.6 - Pragma Directive.
1020     case tok::pp_pragma:
1021       return HandlePragmaDirective({PIK_HashPragma, SavedHash.getLocation()});
1022 
1023     // GNU Extensions.
1024     case tok::pp_import:
1025       return HandleImportDirective(SavedHash.getLocation(), Result);
1026     case tok::pp_include_next:
1027       return HandleIncludeNextDirective(SavedHash.getLocation(), Result);
1028 
1029     case tok::pp_warning:
1030       Diag(Result, diag::ext_pp_warning_directive);
1031       return HandleUserDiagnosticDirective(Result, true);
1032     case tok::pp_ident:
1033       return HandleIdentSCCSDirective(Result);
1034     case tok::pp_sccs:
1035       return HandleIdentSCCSDirective(Result);
1036     case tok::pp_assert:
1037       //isExtension = true;  // FIXME: implement #assert
1038       break;
1039     case tok::pp_unassert:
1040       //isExtension = true;  // FIXME: implement #unassert
1041       break;
1042 
1043     case tok::pp___public_macro:
1044       if (getLangOpts().Modules)
1045         return HandleMacroPublicDirective(Result);
1046       break;
1047 
1048     case tok::pp___private_macro:
1049       if (getLangOpts().Modules)
1050         return HandleMacroPrivateDirective();
1051       break;
1052     }
1053     break;
1054   }
1055 
1056   // If this is a .S file, treat unknown # directives as non-preprocessor
1057   // directives.  This is important because # may be a comment or introduce
1058   // various pseudo-ops.  Just return the # token and push back the following
1059   // token to be lexed next time.
1060   if (getLangOpts().AsmPreprocessor) {
1061     auto Toks = std::make_unique<Token[]>(2);
1062     // Return the # and the token after it.
1063     Toks[0] = SavedHash;
1064     Toks[1] = Result;
1065 
1066     // If the second token is a hashhash token, then we need to translate it to
1067     // unknown so the token lexer doesn't try to perform token pasting.
1068     if (Result.is(tok::hashhash))
1069       Toks[1].setKind(tok::unknown);
1070 
1071     // Enter this token stream so that we re-lex the tokens.  Make sure to
1072     // enable macro expansion, in case the token after the # is an identifier
1073     // that is expanded.
1074     EnterTokenStream(std::move(Toks), 2, false, /*IsReinject*/false);
1075     return;
1076   }
1077 
1078   // If we reached here, the preprocessing token is not valid!
1079   Diag(Result, diag::err_pp_invalid_directive);
1080 
1081   // Read the rest of the PP line.
1082   DiscardUntilEndOfDirective();
1083 
1084   // Okay, we're done parsing the directive.
1085 }
1086 
1087 /// GetLineValue - Convert a numeric token into an unsigned value, emitting
1088 /// Diagnostic DiagID if it is invalid, and returning the value in Val.
1089 static bool GetLineValue(Token &DigitTok, unsigned &Val,
1090                          unsigned DiagID, Preprocessor &PP,
1091                          bool IsGNULineDirective=false) {
1092   if (DigitTok.isNot(tok::numeric_constant)) {
1093     PP.Diag(DigitTok, DiagID);
1094 
1095     if (DigitTok.isNot(tok::eod))
1096       PP.DiscardUntilEndOfDirective();
1097     return true;
1098   }
1099 
1100   SmallString<64> IntegerBuffer;
1101   IntegerBuffer.resize(DigitTok.getLength());
1102   const char *DigitTokBegin = &IntegerBuffer[0];
1103   bool Invalid = false;
1104   unsigned ActualLength = PP.getSpelling(DigitTok, DigitTokBegin, &Invalid);
1105   if (Invalid)
1106     return true;
1107 
1108   // Verify that we have a simple digit-sequence, and compute the value.  This
1109   // is always a simple digit string computed in decimal, so we do this manually
1110   // here.
1111   Val = 0;
1112   for (unsigned i = 0; i != ActualLength; ++i) {
1113     // C++1y [lex.fcon]p1:
1114     //   Optional separating single quotes in a digit-sequence are ignored
1115     if (DigitTokBegin[i] == '\'')
1116       continue;
1117 
1118     if (!isDigit(DigitTokBegin[i])) {
1119       PP.Diag(PP.AdvanceToTokenCharacter(DigitTok.getLocation(), i),
1120               diag::err_pp_line_digit_sequence) << IsGNULineDirective;
1121       PP.DiscardUntilEndOfDirective();
1122       return true;
1123     }
1124 
1125     unsigned NextVal = Val*10+(DigitTokBegin[i]-'0');
1126     if (NextVal < Val) { // overflow.
1127       PP.Diag(DigitTok, DiagID);
1128       PP.DiscardUntilEndOfDirective();
1129       return true;
1130     }
1131     Val = NextVal;
1132   }
1133 
1134   if (DigitTokBegin[0] == '0' && Val)
1135     PP.Diag(DigitTok.getLocation(), diag::warn_pp_line_decimal)
1136       << IsGNULineDirective;
1137 
1138   return false;
1139 }
1140 
1141 /// Handle a \#line directive: C99 6.10.4.
1142 ///
1143 /// The two acceptable forms are:
1144 /// \verbatim
1145 ///   # line digit-sequence
1146 ///   # line digit-sequence "s-char-sequence"
1147 /// \endverbatim
1148 void Preprocessor::HandleLineDirective() {
1149   // Read the line # and string argument.  Per C99 6.10.4p5, these tokens are
1150   // expanded.
1151   Token DigitTok;
1152   Lex(DigitTok);
1153 
1154   // Validate the number and convert it to an unsigned.
1155   unsigned LineNo;
1156   if (GetLineValue(DigitTok, LineNo, diag::err_pp_line_requires_integer,*this))
1157     return;
1158 
1159   if (LineNo == 0)
1160     Diag(DigitTok, diag::ext_pp_line_zero);
1161 
1162   // Enforce C99 6.10.4p3: "The digit sequence shall not specify ... a
1163   // number greater than 2147483647".  C90 requires that the line # be <= 32767.
1164   unsigned LineLimit = 32768U;
1165   if (LangOpts.C99 || LangOpts.CPlusPlus11)
1166     LineLimit = 2147483648U;
1167   if (LineNo >= LineLimit)
1168     Diag(DigitTok, diag::ext_pp_line_too_big) << LineLimit;
1169   else if (LangOpts.CPlusPlus11 && LineNo >= 32768U)
1170     Diag(DigitTok, diag::warn_cxx98_compat_pp_line_too_big);
1171 
1172   int FilenameID = -1;
1173   Token StrTok;
1174   Lex(StrTok);
1175 
1176   // If the StrTok is "eod", then it wasn't present.  Otherwise, it must be a
1177   // string followed by eod.
1178   if (StrTok.is(tok::eod))
1179     ; // ok
1180   else if (StrTok.isNot(tok::string_literal)) {
1181     Diag(StrTok, diag::err_pp_line_invalid_filename);
1182     DiscardUntilEndOfDirective();
1183     return;
1184   } else if (StrTok.hasUDSuffix()) {
1185     Diag(StrTok, diag::err_invalid_string_udl);
1186     DiscardUntilEndOfDirective();
1187     return;
1188   } else {
1189     // Parse and validate the string, converting it into a unique ID.
1190     StringLiteralParser Literal(StrTok, *this);
1191     assert(Literal.isAscii() && "Didn't allow wide strings in");
1192     if (Literal.hadError) {
1193       DiscardUntilEndOfDirective();
1194       return;
1195     }
1196     if (Literal.Pascal) {
1197       Diag(StrTok, diag::err_pp_linemarker_invalid_filename);
1198       DiscardUntilEndOfDirective();
1199       return;
1200     }
1201     FilenameID = SourceMgr.getLineTableFilenameID(Literal.GetString());
1202 
1203     // Verify that there is nothing after the string, other than EOD.  Because
1204     // of C99 6.10.4p5, macros that expand to empty tokens are ok.
1205     CheckEndOfDirective("line", true);
1206   }
1207 
1208   // Take the file kind of the file containing the #line directive. #line
1209   // directives are often used for generated sources from the same codebase, so
1210   // the new file should generally be classified the same way as the current
1211   // file. This is visible in GCC's pre-processed output, which rewrites #line
1212   // to GNU line markers.
1213   SrcMgr::CharacteristicKind FileKind =
1214       SourceMgr.getFileCharacteristic(DigitTok.getLocation());
1215 
1216   SourceMgr.AddLineNote(DigitTok.getLocation(), LineNo, FilenameID, false,
1217                         false, FileKind);
1218 
1219   if (Callbacks)
1220     Callbacks->FileChanged(CurPPLexer->getSourceLocation(),
1221                            PPCallbacks::RenameFile, FileKind);
1222 }
1223 
1224 /// ReadLineMarkerFlags - Parse and validate any flags at the end of a GNU line
1225 /// marker directive.
1226 static bool ReadLineMarkerFlags(bool &IsFileEntry, bool &IsFileExit,
1227                                 SrcMgr::CharacteristicKind &FileKind,
1228                                 Preprocessor &PP) {
1229   unsigned FlagVal;
1230   Token FlagTok;
1231   PP.Lex(FlagTok);
1232   if (FlagTok.is(tok::eod)) return false;
1233   if (GetLineValue(FlagTok, FlagVal, diag::err_pp_linemarker_invalid_flag, PP))
1234     return true;
1235 
1236   if (FlagVal == 1) {
1237     IsFileEntry = true;
1238 
1239     PP.Lex(FlagTok);
1240     if (FlagTok.is(tok::eod)) return false;
1241     if (GetLineValue(FlagTok, FlagVal, diag::err_pp_linemarker_invalid_flag,PP))
1242       return true;
1243   } else if (FlagVal == 2) {
1244     IsFileExit = true;
1245 
1246     SourceManager &SM = PP.getSourceManager();
1247     // If we are leaving the current presumed file, check to make sure the
1248     // presumed include stack isn't empty!
1249     FileID CurFileID =
1250       SM.getDecomposedExpansionLoc(FlagTok.getLocation()).first;
1251     PresumedLoc PLoc = SM.getPresumedLoc(FlagTok.getLocation());
1252     if (PLoc.isInvalid())
1253       return true;
1254 
1255     // If there is no include loc (main file) or if the include loc is in a
1256     // different physical file, then we aren't in a "1" line marker flag region.
1257     SourceLocation IncLoc = PLoc.getIncludeLoc();
1258     if (IncLoc.isInvalid() ||
1259         SM.getDecomposedExpansionLoc(IncLoc).first != CurFileID) {
1260       PP.Diag(FlagTok, diag::err_pp_linemarker_invalid_pop);
1261       PP.DiscardUntilEndOfDirective();
1262       return true;
1263     }
1264 
1265     PP.Lex(FlagTok);
1266     if (FlagTok.is(tok::eod)) return false;
1267     if (GetLineValue(FlagTok, FlagVal, diag::err_pp_linemarker_invalid_flag,PP))
1268       return true;
1269   }
1270 
1271   // We must have 3 if there are still flags.
1272   if (FlagVal != 3) {
1273     PP.Diag(FlagTok, diag::err_pp_linemarker_invalid_flag);
1274     PP.DiscardUntilEndOfDirective();
1275     return true;
1276   }
1277 
1278   FileKind = SrcMgr::C_System;
1279 
1280   PP.Lex(FlagTok);
1281   if (FlagTok.is(tok::eod)) return false;
1282   if (GetLineValue(FlagTok, FlagVal, diag::err_pp_linemarker_invalid_flag, PP))
1283     return true;
1284 
1285   // We must have 4 if there is yet another flag.
1286   if (FlagVal != 4) {
1287     PP.Diag(FlagTok, diag::err_pp_linemarker_invalid_flag);
1288     PP.DiscardUntilEndOfDirective();
1289     return true;
1290   }
1291 
1292   FileKind = SrcMgr::C_ExternCSystem;
1293 
1294   PP.Lex(FlagTok);
1295   if (FlagTok.is(tok::eod)) return false;
1296 
1297   // There are no more valid flags here.
1298   PP.Diag(FlagTok, diag::err_pp_linemarker_invalid_flag);
1299   PP.DiscardUntilEndOfDirective();
1300   return true;
1301 }
1302 
1303 /// HandleDigitDirective - Handle a GNU line marker directive, whose syntax is
1304 /// one of the following forms:
1305 ///
1306 ///     # 42
1307 ///     # 42 "file" ('1' | '2')?
1308 ///     # 42 "file" ('1' | '2')? '3' '4'?
1309 ///
1310 void Preprocessor::HandleDigitDirective(Token &DigitTok) {
1311   // Validate the number and convert it to an unsigned.  GNU does not have a
1312   // line # limit other than it fit in 32-bits.
1313   unsigned LineNo;
1314   if (GetLineValue(DigitTok, LineNo, diag::err_pp_linemarker_requires_integer,
1315                    *this, true))
1316     return;
1317 
1318   Token StrTok;
1319   Lex(StrTok);
1320 
1321   bool IsFileEntry = false, IsFileExit = false;
1322   int FilenameID = -1;
1323   SrcMgr::CharacteristicKind FileKind = SrcMgr::C_User;
1324 
1325   // If the StrTok is "eod", then it wasn't present.  Otherwise, it must be a
1326   // string followed by eod.
1327   if (StrTok.is(tok::eod)) {
1328     // Treat this like "#line NN", which doesn't change file characteristics.
1329     FileKind = SourceMgr.getFileCharacteristic(DigitTok.getLocation());
1330   } else if (StrTok.isNot(tok::string_literal)) {
1331     Diag(StrTok, diag::err_pp_linemarker_invalid_filename);
1332     DiscardUntilEndOfDirective();
1333     return;
1334   } else if (StrTok.hasUDSuffix()) {
1335     Diag(StrTok, diag::err_invalid_string_udl);
1336     DiscardUntilEndOfDirective();
1337     return;
1338   } else {
1339     // Parse and validate the string, converting it into a unique ID.
1340     StringLiteralParser Literal(StrTok, *this);
1341     assert(Literal.isAscii() && "Didn't allow wide strings in");
1342     if (Literal.hadError) {
1343       DiscardUntilEndOfDirective();
1344       return;
1345     }
1346     if (Literal.Pascal) {
1347       Diag(StrTok, diag::err_pp_linemarker_invalid_filename);
1348       DiscardUntilEndOfDirective();
1349       return;
1350     }
1351     FilenameID = SourceMgr.getLineTableFilenameID(Literal.GetString());
1352 
1353     // If a filename was present, read any flags that are present.
1354     if (ReadLineMarkerFlags(IsFileEntry, IsFileExit, FileKind, *this))
1355       return;
1356   }
1357 
1358   // Create a line note with this information.
1359   SourceMgr.AddLineNote(DigitTok.getLocation(), LineNo, FilenameID, IsFileEntry,
1360                         IsFileExit, FileKind);
1361 
1362   // If the preprocessor has callbacks installed, notify them of the #line
1363   // change.  This is used so that the line marker comes out in -E mode for
1364   // example.
1365   if (Callbacks) {
1366     PPCallbacks::FileChangeReason Reason = PPCallbacks::RenameFile;
1367     if (IsFileEntry)
1368       Reason = PPCallbacks::EnterFile;
1369     else if (IsFileExit)
1370       Reason = PPCallbacks::ExitFile;
1371 
1372     Callbacks->FileChanged(CurPPLexer->getSourceLocation(), Reason, FileKind);
1373   }
1374 }
1375 
1376 /// HandleUserDiagnosticDirective - Handle a #warning or #error directive.
1377 ///
1378 void Preprocessor::HandleUserDiagnosticDirective(Token &Tok,
1379                                                  bool isWarning) {
1380   // Read the rest of the line raw.  We do this because we don't want macros
1381   // to be expanded and we don't require that the tokens be valid preprocessing
1382   // tokens.  For example, this is allowed: "#warning `   'foo".  GCC does
1383   // collapse multiple consecutive white space between tokens, but this isn't
1384   // specified by the standard.
1385   SmallString<128> Message;
1386   CurLexer->ReadToEndOfLine(&Message);
1387 
1388   // Find the first non-whitespace character, so that we can make the
1389   // diagnostic more succinct.
1390   StringRef Msg = StringRef(Message).ltrim(' ');
1391 
1392   if (isWarning)
1393     Diag(Tok, diag::pp_hash_warning) << Msg;
1394   else
1395     Diag(Tok, diag::err_pp_hash_error) << Msg;
1396 }
1397 
1398 /// HandleIdentSCCSDirective - Handle a #ident/#sccs directive.
1399 ///
1400 void Preprocessor::HandleIdentSCCSDirective(Token &Tok) {
1401   // Yes, this directive is an extension.
1402   Diag(Tok, diag::ext_pp_ident_directive);
1403 
1404   // Read the string argument.
1405   Token StrTok;
1406   Lex(StrTok);
1407 
1408   // If the token kind isn't a string, it's a malformed directive.
1409   if (StrTok.isNot(tok::string_literal) &&
1410       StrTok.isNot(tok::wide_string_literal)) {
1411     Diag(StrTok, diag::err_pp_malformed_ident);
1412     if (StrTok.isNot(tok::eod))
1413       DiscardUntilEndOfDirective();
1414     return;
1415   }
1416 
1417   if (StrTok.hasUDSuffix()) {
1418     Diag(StrTok, diag::err_invalid_string_udl);
1419     DiscardUntilEndOfDirective();
1420     return;
1421   }
1422 
1423   // Verify that there is nothing after the string, other than EOD.
1424   CheckEndOfDirective("ident");
1425 
1426   if (Callbacks) {
1427     bool Invalid = false;
1428     std::string Str = getSpelling(StrTok, &Invalid);
1429     if (!Invalid)
1430       Callbacks->Ident(Tok.getLocation(), Str);
1431   }
1432 }
1433 
1434 /// Handle a #public directive.
1435 void Preprocessor::HandleMacroPublicDirective(Token &Tok) {
1436   Token MacroNameTok;
1437   ReadMacroName(MacroNameTok, MU_Undef);
1438 
1439   // Error reading macro name?  If so, diagnostic already issued.
1440   if (MacroNameTok.is(tok::eod))
1441     return;
1442 
1443   // Check to see if this is the last token on the #__public_macro line.
1444   CheckEndOfDirective("__public_macro");
1445 
1446   IdentifierInfo *II = MacroNameTok.getIdentifierInfo();
1447   // Okay, we finally have a valid identifier to undef.
1448   MacroDirective *MD = getLocalMacroDirective(II);
1449 
1450   // If the macro is not defined, this is an error.
1451   if (!MD) {
1452     Diag(MacroNameTok, diag::err_pp_visibility_non_macro) << II;
1453     return;
1454   }
1455 
1456   // Note that this macro has now been exported.
1457   appendMacroDirective(II, AllocateVisibilityMacroDirective(
1458                                 MacroNameTok.getLocation(), /*isPublic=*/true));
1459 }
1460 
1461 /// Handle a #private directive.
1462 void Preprocessor::HandleMacroPrivateDirective() {
1463   Token MacroNameTok;
1464   ReadMacroName(MacroNameTok, MU_Undef);
1465 
1466   // Error reading macro name?  If so, diagnostic already issued.
1467   if (MacroNameTok.is(tok::eod))
1468     return;
1469 
1470   // Check to see if this is the last token on the #__private_macro line.
1471   CheckEndOfDirective("__private_macro");
1472 
1473   IdentifierInfo *II = MacroNameTok.getIdentifierInfo();
1474   // Okay, we finally have a valid identifier to undef.
1475   MacroDirective *MD = getLocalMacroDirective(II);
1476 
1477   // If the macro is not defined, this is an error.
1478   if (!MD) {
1479     Diag(MacroNameTok, diag::err_pp_visibility_non_macro) << II;
1480     return;
1481   }
1482 
1483   // Note that this macro has now been marked private.
1484   appendMacroDirective(II, AllocateVisibilityMacroDirective(
1485                                MacroNameTok.getLocation(), /*isPublic=*/false));
1486 }
1487 
1488 //===----------------------------------------------------------------------===//
1489 // Preprocessor Include Directive Handling.
1490 //===----------------------------------------------------------------------===//
1491 
1492 /// GetIncludeFilenameSpelling - Turn the specified lexer token into a fully
1493 /// checked and spelled filename, e.g. as an operand of \#include. This returns
1494 /// true if the input filename was in <>'s or false if it were in ""'s.  The
1495 /// caller is expected to provide a buffer that is large enough to hold the
1496 /// spelling of the filename, but is also expected to handle the case when
1497 /// this method decides to use a different buffer.
1498 bool Preprocessor::GetIncludeFilenameSpelling(SourceLocation Loc,
1499                                               StringRef &Buffer) {
1500   // Get the text form of the filename.
1501   assert(!Buffer.empty() && "Can't have tokens with empty spellings!");
1502 
1503   // FIXME: Consider warning on some of the cases described in C11 6.4.7/3 and
1504   // C++20 [lex.header]/2:
1505   //
1506   // If `"`, `'`, `\`, `/*`, or `//` appears in a header-name, then
1507   //   in C: behavior is undefined
1508   //   in C++: program is conditionally-supported with implementation-defined
1509   //           semantics
1510 
1511   // Make sure the filename is <x> or "x".
1512   bool isAngled;
1513   if (Buffer[0] == '<') {
1514     if (Buffer.back() != '>') {
1515       Diag(Loc, diag::err_pp_expects_filename);
1516       Buffer = StringRef();
1517       return true;
1518     }
1519     isAngled = true;
1520   } else if (Buffer[0] == '"') {
1521     if (Buffer.back() != '"') {
1522       Diag(Loc, diag::err_pp_expects_filename);
1523       Buffer = StringRef();
1524       return true;
1525     }
1526     isAngled = false;
1527   } else {
1528     Diag(Loc, diag::err_pp_expects_filename);
1529     Buffer = StringRef();
1530     return true;
1531   }
1532 
1533   // Diagnose #include "" as invalid.
1534   if (Buffer.size() <= 2) {
1535     Diag(Loc, diag::err_pp_empty_filename);
1536     Buffer = StringRef();
1537     return true;
1538   }
1539 
1540   // Skip the brackets.
1541   Buffer = Buffer.substr(1, Buffer.size()-2);
1542   return isAngled;
1543 }
1544 
1545 /// Push a token onto the token stream containing an annotation.
1546 void Preprocessor::EnterAnnotationToken(SourceRange Range,
1547                                         tok::TokenKind Kind,
1548                                         void *AnnotationVal) {
1549   // FIXME: Produce this as the current token directly, rather than
1550   // allocating a new token for it.
1551   auto Tok = std::make_unique<Token[]>(1);
1552   Tok[0].startToken();
1553   Tok[0].setKind(Kind);
1554   Tok[0].setLocation(Range.getBegin());
1555   Tok[0].setAnnotationEndLoc(Range.getEnd());
1556   Tok[0].setAnnotationValue(AnnotationVal);
1557   EnterTokenStream(std::move(Tok), 1, true, /*IsReinject*/ false);
1558 }
1559 
1560 /// Produce a diagnostic informing the user that a #include or similar
1561 /// was implicitly treated as a module import.
1562 static void diagnoseAutoModuleImport(
1563     Preprocessor &PP, SourceLocation HashLoc, Token &IncludeTok,
1564     ArrayRef<std::pair<IdentifierInfo *, SourceLocation>> Path,
1565     SourceLocation PathEnd) {
1566   StringRef ImportKeyword;
1567   if (PP.getLangOpts().ObjC)
1568     ImportKeyword = "@import";
1569   else if (PP.getLangOpts().ModulesTS || PP.getLangOpts().CPlusPlusModules)
1570     ImportKeyword = "import";
1571   else
1572     return; // no import syntax available
1573 
1574   SmallString<128> PathString;
1575   for (size_t I = 0, N = Path.size(); I != N; ++I) {
1576     if (I)
1577       PathString += '.';
1578     PathString += Path[I].first->getName();
1579   }
1580   int IncludeKind = 0;
1581 
1582   switch (IncludeTok.getIdentifierInfo()->getPPKeywordID()) {
1583   case tok::pp_include:
1584     IncludeKind = 0;
1585     break;
1586 
1587   case tok::pp_import:
1588     IncludeKind = 1;
1589     break;
1590 
1591   case tok::pp_include_next:
1592     IncludeKind = 2;
1593     break;
1594 
1595   case tok::pp___include_macros:
1596     IncludeKind = 3;
1597     break;
1598 
1599   default:
1600     llvm_unreachable("unknown include directive kind");
1601   }
1602 
1603   CharSourceRange ReplaceRange(SourceRange(HashLoc, PathEnd),
1604                                /*IsTokenRange=*/false);
1605   PP.Diag(HashLoc, diag::warn_auto_module_import)
1606       << IncludeKind << PathString
1607       << FixItHint::CreateReplacement(
1608              ReplaceRange, (ImportKeyword + " " + PathString + ";").str());
1609 }
1610 
1611 // Given a vector of path components and a string containing the real
1612 // path to the file, build a properly-cased replacement in the vector,
1613 // and return true if the replacement should be suggested.
1614 static bool trySimplifyPath(SmallVectorImpl<StringRef> &Components,
1615                             StringRef RealPathName) {
1616   auto RealPathComponentIter = llvm::sys::path::rbegin(RealPathName);
1617   auto RealPathComponentEnd = llvm::sys::path::rend(RealPathName);
1618   int Cnt = 0;
1619   bool SuggestReplacement = false;
1620   // Below is a best-effort to handle ".." in paths. It is admittedly
1621   // not 100% correct in the presence of symlinks.
1622   for (auto &Component : llvm::reverse(Components)) {
1623     if ("." == Component) {
1624     } else if (".." == Component) {
1625       ++Cnt;
1626     } else if (Cnt) {
1627       --Cnt;
1628     } else if (RealPathComponentIter != RealPathComponentEnd) {
1629       if (Component != *RealPathComponentIter) {
1630         // If these path components differ by more than just case, then we
1631         // may be looking at symlinked paths. Bail on this diagnostic to avoid
1632         // noisy false positives.
1633         SuggestReplacement = RealPathComponentIter->equals_lower(Component);
1634         if (!SuggestReplacement)
1635           break;
1636         Component = *RealPathComponentIter;
1637       }
1638       ++RealPathComponentIter;
1639     }
1640   }
1641   return SuggestReplacement;
1642 }
1643 
1644 bool Preprocessor::checkModuleIsAvailable(const LangOptions &LangOpts,
1645                                           const TargetInfo &TargetInfo,
1646                                           DiagnosticsEngine &Diags, Module *M) {
1647   Module::Requirement Requirement;
1648   Module::UnresolvedHeaderDirective MissingHeader;
1649   Module *ShadowingModule = nullptr;
1650   if (M->isAvailable(LangOpts, TargetInfo, Requirement, MissingHeader,
1651                      ShadowingModule))
1652     return false;
1653 
1654   if (MissingHeader.FileNameLoc.isValid()) {
1655     Diags.Report(MissingHeader.FileNameLoc, diag::err_module_header_missing)
1656         << MissingHeader.IsUmbrella << MissingHeader.FileName;
1657   } else if (ShadowingModule) {
1658     Diags.Report(M->DefinitionLoc, diag::err_module_shadowed) << M->Name;
1659     Diags.Report(ShadowingModule->DefinitionLoc,
1660                  diag::note_previous_definition);
1661   } else {
1662     // FIXME: Track the location at which the requirement was specified, and
1663     // use it here.
1664     Diags.Report(M->DefinitionLoc, diag::err_module_unavailable)
1665         << M->getFullModuleName() << Requirement.second << Requirement.first;
1666   }
1667   return true;
1668 }
1669 
1670 /// HandleIncludeDirective - The "\#include" tokens have just been read, read
1671 /// the file to be included from the lexer, then include it!  This is a common
1672 /// routine with functionality shared between \#include, \#include_next and
1673 /// \#import.  LookupFrom is set when this is a \#include_next directive, it
1674 /// specifies the file to start searching from.
1675 void Preprocessor::HandleIncludeDirective(SourceLocation HashLoc,
1676                                           Token &IncludeTok,
1677                                           const DirectoryLookup *LookupFrom,
1678                                           const FileEntry *LookupFromFile) {
1679   Token FilenameTok;
1680   if (LexHeaderName(FilenameTok))
1681     return;
1682 
1683   if (FilenameTok.isNot(tok::header_name)) {
1684     Diag(FilenameTok.getLocation(), diag::err_pp_expects_filename);
1685     if (FilenameTok.isNot(tok::eod))
1686       DiscardUntilEndOfDirective();
1687     return;
1688   }
1689 
1690   // Verify that there is nothing after the filename, other than EOD.  Note
1691   // that we allow macros that expand to nothing after the filename, because
1692   // this falls into the category of "#include pp-tokens new-line" specified
1693   // in C99 6.10.2p4.
1694   SourceLocation EndLoc =
1695       CheckEndOfDirective(IncludeTok.getIdentifierInfo()->getNameStart(), true);
1696 
1697   auto Action = HandleHeaderIncludeOrImport(HashLoc, IncludeTok, FilenameTok,
1698                                             EndLoc, LookupFrom, LookupFromFile);
1699   switch (Action.Kind) {
1700   case ImportAction::None:
1701   case ImportAction::SkippedModuleImport:
1702     break;
1703   case ImportAction::ModuleBegin:
1704     EnterAnnotationToken(SourceRange(HashLoc, EndLoc),
1705                          tok::annot_module_begin, Action.ModuleForHeader);
1706     break;
1707   case ImportAction::ModuleImport:
1708     EnterAnnotationToken(SourceRange(HashLoc, EndLoc),
1709                          tok::annot_module_include, Action.ModuleForHeader);
1710     break;
1711   case ImportAction::Failure:
1712     assert(TheModuleLoader.HadFatalFailure &&
1713            "This should be an early exit only to a fatal error");
1714     TheModuleLoader.HadFatalFailure = true;
1715     IncludeTok.setKind(tok::eof);
1716     CurLexer->cutOffLexing();
1717     return;
1718   }
1719 }
1720 
1721 Optional<FileEntryRef> Preprocessor::LookupHeaderIncludeOrImport(
1722     const DirectoryLookup *&CurDir, StringRef Filename,
1723     SourceLocation FilenameLoc, CharSourceRange FilenameRange,
1724     const Token &FilenameTok, bool &IsFrameworkFound, bool IsImportDecl,
1725     bool &IsMapped, const DirectoryLookup *LookupFrom,
1726     const FileEntry *LookupFromFile, StringRef LookupFilename,
1727     SmallVectorImpl<char> &RelativePath, SmallVectorImpl<char> &SearchPath,
1728     ModuleMap::KnownHeader &SuggestedModule, bool isAngled) {
1729   Optional<FileEntryRef> File = LookupFile(
1730       FilenameLoc, LookupFilename,
1731       isAngled, LookupFrom, LookupFromFile, CurDir,
1732       Callbacks ? &SearchPath : nullptr, Callbacks ? &RelativePath : nullptr,
1733       &SuggestedModule, &IsMapped, &IsFrameworkFound);
1734   if (File)
1735     return File;
1736 
1737   if (Callbacks) {
1738     // Give the clients a chance to recover.
1739     SmallString<128> RecoveryPath;
1740     if (Callbacks->FileNotFound(Filename, RecoveryPath)) {
1741       if (auto DE = FileMgr.getOptionalDirectoryRef(RecoveryPath)) {
1742         // Add the recovery path to the list of search paths.
1743         DirectoryLookup DL(*DE, SrcMgr::C_User, false);
1744         HeaderInfo.AddSearchPath(DL, isAngled);
1745 
1746         // Try the lookup again, skipping the cache.
1747         Optional<FileEntryRef> File = LookupFile(
1748             FilenameLoc,
1749             LookupFilename, isAngled,
1750             LookupFrom, LookupFromFile, CurDir, nullptr, nullptr,
1751             &SuggestedModule, &IsMapped, /*IsFrameworkFound=*/nullptr,
1752             /*SkipCache*/ true);
1753         if (File)
1754           return File;
1755       }
1756     }
1757   }
1758 
1759   if (SuppressIncludeNotFoundError)
1760     return None;
1761 
1762   // If the file could not be located and it was included via angle
1763   // brackets, we can attempt a lookup as though it were a quoted path to
1764   // provide the user with a possible fixit.
1765   if (isAngled) {
1766     Optional<FileEntryRef> File = LookupFile(
1767         FilenameLoc, LookupFilename,
1768         false, LookupFrom, LookupFromFile, CurDir,
1769         Callbacks ? &SearchPath : nullptr, Callbacks ? &RelativePath : nullptr,
1770         &SuggestedModule, &IsMapped,
1771         /*IsFrameworkFound=*/nullptr);
1772     if (File) {
1773       Diag(FilenameTok, diag::err_pp_file_not_found_angled_include_not_fatal)
1774           << Filename << IsImportDecl
1775           << FixItHint::CreateReplacement(FilenameRange,
1776                                           "\"" + Filename.str() + "\"");
1777       return File;
1778     }
1779   }
1780 
1781   // Check for likely typos due to leading or trailing non-isAlphanumeric
1782   // characters
1783   StringRef OriginalFilename = Filename;
1784   if (LangOpts.SpellChecking) {
1785     // A heuristic to correct a typo file name by removing leading and
1786     // trailing non-isAlphanumeric characters.
1787     auto CorrectTypoFilename = [](llvm::StringRef Filename) {
1788       Filename = Filename.drop_until(isAlphanumeric);
1789       while (!Filename.empty() && !isAlphanumeric(Filename.back())) {
1790         Filename = Filename.drop_back();
1791       }
1792       return Filename;
1793     };
1794     StringRef TypoCorrectionName = CorrectTypoFilename(Filename);
1795 
1796 #ifndef _WIN32
1797     // Normalize slashes when compiling with -fms-extensions on non-Windows.
1798     // This is unnecessary on Windows since the filesystem there handles
1799     // backslashes.
1800     SmallString<128> NormalizedTypoCorrectionPath;
1801     if (LangOpts.MicrosoftExt) {
1802       NormalizedTypoCorrectionPath = TypoCorrectionName;
1803       llvm::sys::path::native(NormalizedTypoCorrectionPath);
1804       TypoCorrectionName = NormalizedTypoCorrectionPath;
1805     }
1806 #endif
1807 
1808     Optional<FileEntryRef> File = LookupFile(
1809         FilenameLoc, TypoCorrectionName, isAngled, LookupFrom, LookupFromFile,
1810         CurDir, Callbacks ? &SearchPath : nullptr,
1811         Callbacks ? &RelativePath : nullptr, &SuggestedModule, &IsMapped,
1812         /*IsFrameworkFound=*/nullptr);
1813     if (File) {
1814       auto Hint =
1815           isAngled ? FixItHint::CreateReplacement(
1816                          FilenameRange, "<" + TypoCorrectionName.str() + ">")
1817                    : FixItHint::CreateReplacement(
1818                          FilenameRange, "\"" + TypoCorrectionName.str() + "\"");
1819       Diag(FilenameTok, diag::err_pp_file_not_found_typo_not_fatal)
1820           << OriginalFilename << TypoCorrectionName << Hint;
1821       // We found the file, so set the Filename to the name after typo
1822       // correction.
1823       Filename = TypoCorrectionName;
1824       return File;
1825     }
1826   }
1827 
1828   // If the file is still not found, just go with the vanilla diagnostic
1829   assert(!File.hasValue() && "expected missing file");
1830   Diag(FilenameTok, diag::err_pp_file_not_found)
1831       << OriginalFilename << FilenameRange;
1832   if (IsFrameworkFound) {
1833     size_t SlashPos = OriginalFilename.find('/');
1834     assert(SlashPos != StringRef::npos &&
1835            "Include with framework name should have '/' in the filename");
1836     StringRef FrameworkName = OriginalFilename.substr(0, SlashPos);
1837     FrameworkCacheEntry &CacheEntry =
1838         HeaderInfo.LookupFrameworkCache(FrameworkName);
1839     assert(CacheEntry.Directory && "Found framework should be in cache");
1840     Diag(FilenameTok, diag::note_pp_framework_without_header)
1841         << OriginalFilename.substr(SlashPos + 1) << FrameworkName
1842         << CacheEntry.Directory->getName();
1843   }
1844 
1845   return None;
1846 }
1847 
1848 /// Handle either a #include-like directive or an import declaration that names
1849 /// a header file.
1850 ///
1851 /// \param HashLoc The location of the '#' token for an include, or
1852 ///        SourceLocation() for an import declaration.
1853 /// \param IncludeTok The include / include_next / import token.
1854 /// \param FilenameTok The header-name token.
1855 /// \param EndLoc The location at which any imported macros become visible.
1856 /// \param LookupFrom For #include_next, the starting directory for the
1857 ///        directory lookup.
1858 /// \param LookupFromFile For #include_next, the starting file for the directory
1859 ///        lookup.
1860 Preprocessor::ImportAction Preprocessor::HandleHeaderIncludeOrImport(
1861     SourceLocation HashLoc, Token &IncludeTok, Token &FilenameTok,
1862     SourceLocation EndLoc, const DirectoryLookup *LookupFrom,
1863     const FileEntry *LookupFromFile) {
1864   SmallString<128> FilenameBuffer;
1865   StringRef Filename = getSpelling(FilenameTok, FilenameBuffer);
1866   SourceLocation CharEnd = FilenameTok.getEndLoc();
1867 
1868   CharSourceRange FilenameRange
1869     = CharSourceRange::getCharRange(FilenameTok.getLocation(), CharEnd);
1870   StringRef OriginalFilename = Filename;
1871   bool isAngled =
1872     GetIncludeFilenameSpelling(FilenameTok.getLocation(), Filename);
1873 
1874   // If GetIncludeFilenameSpelling set the start ptr to null, there was an
1875   // error.
1876   if (Filename.empty())
1877     return {ImportAction::None};
1878 
1879   bool IsImportDecl = HashLoc.isInvalid();
1880   SourceLocation StartLoc = IsImportDecl ? IncludeTok.getLocation() : HashLoc;
1881 
1882   // Complain about attempts to #include files in an audit pragma.
1883   if (PragmaARCCFCodeAuditedInfo.second.isValid()) {
1884     Diag(StartLoc, diag::err_pp_include_in_arc_cf_code_audited) << IsImportDecl;
1885     Diag(PragmaARCCFCodeAuditedInfo.second, diag::note_pragma_entered_here);
1886 
1887     // Immediately leave the pragma.
1888     PragmaARCCFCodeAuditedInfo = {nullptr, SourceLocation()};
1889   }
1890 
1891   // Complain about attempts to #include files in an assume-nonnull pragma.
1892   if (PragmaAssumeNonNullLoc.isValid()) {
1893     Diag(StartLoc, diag::err_pp_include_in_assume_nonnull) << IsImportDecl;
1894     Diag(PragmaAssumeNonNullLoc, diag::note_pragma_entered_here);
1895 
1896     // Immediately leave the pragma.
1897     PragmaAssumeNonNullLoc = SourceLocation();
1898   }
1899 
1900   if (HeaderInfo.HasIncludeAliasMap()) {
1901     // Map the filename with the brackets still attached.  If the name doesn't
1902     // map to anything, fall back on the filename we've already gotten the
1903     // spelling for.
1904     StringRef NewName = HeaderInfo.MapHeaderToIncludeAlias(OriginalFilename);
1905     if (!NewName.empty())
1906       Filename = NewName;
1907   }
1908 
1909   // Search include directories.
1910   bool IsMapped = false;
1911   bool IsFrameworkFound = false;
1912   const DirectoryLookup *CurDir;
1913   SmallString<1024> SearchPath;
1914   SmallString<1024> RelativePath;
1915   // We get the raw path only if we have 'Callbacks' to which we later pass
1916   // the path.
1917   ModuleMap::KnownHeader SuggestedModule;
1918   SourceLocation FilenameLoc = FilenameTok.getLocation();
1919   StringRef LookupFilename = Filename;
1920 
1921 #ifndef _WIN32
1922   // Normalize slashes when compiling with -fms-extensions on non-Windows. This
1923   // is unnecessary on Windows since the filesystem there handles backslashes.
1924   SmallString<128> NormalizedPath;
1925   if (LangOpts.MicrosoftExt) {
1926     NormalizedPath = Filename.str();
1927     llvm::sys::path::native(NormalizedPath);
1928     LookupFilename = NormalizedPath;
1929   }
1930 #endif
1931 
1932   Optional<FileEntryRef> File = LookupHeaderIncludeOrImport(
1933       CurDir, Filename, FilenameLoc, FilenameRange, FilenameTok,
1934       IsFrameworkFound, IsImportDecl, IsMapped, LookupFrom, LookupFromFile,
1935       LookupFilename, RelativePath, SearchPath, SuggestedModule, isAngled);
1936 
1937   if (usingPCHWithThroughHeader() && SkippingUntilPCHThroughHeader) {
1938     if (File && isPCHThroughHeader(&File->getFileEntry()))
1939       SkippingUntilPCHThroughHeader = false;
1940     return {ImportAction::None};
1941   }
1942 
1943   // Check for circular inclusion of the main file.
1944   // We can't generate a consistent preamble with regard to the conditional
1945   // stack if the main file is included again as due to the preamble bounds
1946   // some directives (e.g. #endif of a header guard) will never be seen.
1947   // Since this will lead to confusing errors, avoid the inclusion.
1948   if (File && PreambleConditionalStack.isRecording() &&
1949       SourceMgr.translateFile(&File->getFileEntry()) ==
1950           SourceMgr.getMainFileID()) {
1951     Diag(FilenameTok.getLocation(),
1952          diag::err_pp_including_mainfile_in_preamble);
1953     return {ImportAction::None};
1954   }
1955 
1956   // Should we enter the source file? Set to Skip if either the source file is
1957   // known to have no effect beyond its effect on module visibility -- that is,
1958   // if it's got an include guard that is already defined, set to Import if it
1959   // is a modular header we've already built and should import.
1960   enum { Enter, Import, Skip, IncludeLimitReached } Action = Enter;
1961 
1962   if (PPOpts->SingleFileParseMode)
1963     Action = IncludeLimitReached;
1964 
1965   // If we've reached the max allowed include depth, it is usually due to an
1966   // include cycle. Don't enter already processed files again as it can lead to
1967   // reaching the max allowed include depth again.
1968   if (Action == Enter && HasReachedMaxIncludeDepth && File &&
1969       HeaderInfo.getFileInfo(&File->getFileEntry()).NumIncludes)
1970     Action = IncludeLimitReached;
1971 
1972   // Determine whether we should try to import the module for this #include, if
1973   // there is one. Don't do so if precompiled module support is disabled or we
1974   // are processing this module textually (because we're building the module).
1975   if (Action == Enter && File && SuggestedModule && getLangOpts().Modules &&
1976       !isForModuleBuilding(SuggestedModule.getModule(),
1977                            getLangOpts().CurrentModule,
1978                            getLangOpts().ModuleName)) {
1979     // If this include corresponds to a module but that module is
1980     // unavailable, diagnose the situation and bail out.
1981     // FIXME: Remove this; loadModule does the same check (but produces
1982     // slightly worse diagnostics).
1983     if (checkModuleIsAvailable(getLangOpts(), getTargetInfo(), getDiagnostics(),
1984                                SuggestedModule.getModule())) {
1985       Diag(FilenameTok.getLocation(),
1986            diag::note_implicit_top_level_module_import_here)
1987           << SuggestedModule.getModule()->getTopLevelModuleName();
1988       return {ImportAction::None};
1989     }
1990 
1991     // Compute the module access path corresponding to this module.
1992     // FIXME: Should we have a second loadModule() overload to avoid this
1993     // extra lookup step?
1994     SmallVector<std::pair<IdentifierInfo *, SourceLocation>, 2> Path;
1995     for (Module *Mod = SuggestedModule.getModule(); Mod; Mod = Mod->Parent)
1996       Path.push_back(std::make_pair(getIdentifierInfo(Mod->Name),
1997                                     FilenameTok.getLocation()));
1998     std::reverse(Path.begin(), Path.end());
1999 
2000     // Warn that we're replacing the include/import with a module import.
2001     if (!IsImportDecl)
2002       diagnoseAutoModuleImport(*this, StartLoc, IncludeTok, Path, CharEnd);
2003 
2004     // Load the module to import its macros. We'll make the declarations
2005     // visible when the parser gets here.
2006     // FIXME: Pass SuggestedModule in here rather than converting it to a path
2007     // and making the module loader convert it back again.
2008     ModuleLoadResult Imported = TheModuleLoader.loadModule(
2009         IncludeTok.getLocation(), Path, Module::Hidden,
2010         /*IsInclusionDirective=*/true);
2011     assert((Imported == nullptr || Imported == SuggestedModule.getModule()) &&
2012            "the imported module is different than the suggested one");
2013 
2014     if (Imported) {
2015       Action = Import;
2016     } else if (Imported.isMissingExpected()) {
2017       // We failed to find a submodule that we assumed would exist (because it
2018       // was in the directory of an umbrella header, for instance), but no
2019       // actual module containing it exists (because the umbrella header is
2020       // incomplete).  Treat this as a textual inclusion.
2021       SuggestedModule = ModuleMap::KnownHeader();
2022     } else if (Imported.isConfigMismatch()) {
2023       // On a configuration mismatch, enter the header textually. We still know
2024       // that it's part of the corresponding module.
2025     } else {
2026       // We hit an error processing the import. Bail out.
2027       if (hadModuleLoaderFatalFailure()) {
2028         // With a fatal failure in the module loader, we abort parsing.
2029         Token &Result = IncludeTok;
2030         assert(CurLexer && "#include but no current lexer set!");
2031         Result.startToken();
2032         CurLexer->FormTokenWithChars(Result, CurLexer->BufferEnd, tok::eof);
2033         CurLexer->cutOffLexing();
2034       }
2035       return {ImportAction::None};
2036     }
2037   }
2038 
2039   // The #included file will be considered to be a system header if either it is
2040   // in a system include directory, or if the #includer is a system include
2041   // header.
2042   SrcMgr::CharacteristicKind FileCharacter =
2043       SourceMgr.getFileCharacteristic(FilenameTok.getLocation());
2044   if (File)
2045     FileCharacter = std::max(HeaderInfo.getFileDirFlavor(&File->getFileEntry()),
2046                              FileCharacter);
2047 
2048   // If this is a '#import' or an import-declaration, don't re-enter the file.
2049   //
2050   // FIXME: If we have a suggested module for a '#include', and we've already
2051   // visited this file, don't bother entering it again. We know it has no
2052   // further effect.
2053   bool EnterOnce =
2054       IsImportDecl ||
2055       IncludeTok.getIdentifierInfo()->getPPKeywordID() == tok::pp_import;
2056 
2057   // Ask HeaderInfo if we should enter this #include file.  If not, #including
2058   // this file will have no effect.
2059   if (Action == Enter && File &&
2060       !HeaderInfo.ShouldEnterIncludeFile(*this, &File->getFileEntry(),
2061                                          EnterOnce, getLangOpts().Modules,
2062                                          SuggestedModule.getModule())) {
2063     // Even if we've already preprocessed this header once and know that we
2064     // don't need to see its contents again, we still need to import it if it's
2065     // modular because we might not have imported it from this submodule before.
2066     //
2067     // FIXME: We don't do this when compiling a PCH because the AST
2068     // serialization layer can't cope with it. This means we get local
2069     // submodule visibility semantics wrong in that case.
2070     Action = (SuggestedModule && !getLangOpts().CompilingPCH) ? Import : Skip;
2071   }
2072 
2073   if (Callbacks && !IsImportDecl) {
2074     // Notify the callback object that we've seen an inclusion directive.
2075     // FIXME: Use a different callback for a pp-import?
2076     Callbacks->InclusionDirective(
2077         HashLoc, IncludeTok, LookupFilename, isAngled, FilenameRange,
2078         File ? &File->getFileEntry() : nullptr, SearchPath, RelativePath,
2079         Action == Import ? SuggestedModule.getModule() : nullptr,
2080         FileCharacter);
2081     if (Action == Skip && File)
2082       Callbacks->FileSkipped(*File, FilenameTok, FileCharacter);
2083   }
2084 
2085   if (!File)
2086     return {ImportAction::None};
2087 
2088   // If this is a C++20 pp-import declaration, diagnose if we didn't find any
2089   // module corresponding to the named header.
2090   if (IsImportDecl && !SuggestedModule) {
2091     Diag(FilenameTok, diag::err_header_import_not_header_unit)
2092       << OriginalFilename << File->getName();
2093     return {ImportAction::None};
2094   }
2095 
2096   // Issue a diagnostic if the name of the file on disk has a different case
2097   // than the one we're about to open.
2098   const bool CheckIncludePathPortability =
2099       !IsMapped && !File->getFileEntry().tryGetRealPathName().empty();
2100 
2101   if (CheckIncludePathPortability) {
2102     StringRef Name = LookupFilename;
2103     StringRef RealPathName = File->getFileEntry().tryGetRealPathName();
2104     SmallVector<StringRef, 16> Components(llvm::sys::path::begin(Name),
2105                                           llvm::sys::path::end(Name));
2106 
2107     if (trySimplifyPath(Components, RealPathName)) {
2108       SmallString<128> Path;
2109       Path.reserve(Name.size()+2);
2110       Path.push_back(isAngled ? '<' : '"');
2111       bool isLeadingSeparator = llvm::sys::path::is_absolute(Name);
2112       for (auto Component : Components) {
2113         if (isLeadingSeparator)
2114           isLeadingSeparator = false;
2115         else
2116           Path.append(Component);
2117         // Append the separator the user used, or the close quote
2118         Path.push_back(
2119           Path.size() <= Filename.size() ? Filename[Path.size()-1] :
2120             (isAngled ? '>' : '"'));
2121       }
2122       // For user files and known standard headers, by default we issue a diagnostic.
2123       // For other system headers, we don't. They can be controlled separately.
2124       auto DiagId = (FileCharacter == SrcMgr::C_User || warnByDefaultOnWrongCase(Name)) ?
2125           diag::pp_nonportable_path : diag::pp_nonportable_system_path;
2126       Diag(FilenameTok, DiagId) << Path <<
2127         FixItHint::CreateReplacement(FilenameRange, Path);
2128     }
2129   }
2130 
2131   switch (Action) {
2132   case Skip:
2133     // If we don't need to enter the file, stop now.
2134     if (Module *M = SuggestedModule.getModule())
2135       return {ImportAction::SkippedModuleImport, M};
2136     return {ImportAction::None};
2137 
2138   case IncludeLimitReached:
2139     // If we reached our include limit and don't want to enter any more files,
2140     // don't go any further.
2141     return {ImportAction::None};
2142 
2143   case Import: {
2144     // If this is a module import, make it visible if needed.
2145     Module *M = SuggestedModule.getModule();
2146     assert(M && "no module to import");
2147 
2148     makeModuleVisible(M, EndLoc);
2149 
2150     if (IncludeTok.getIdentifierInfo()->getPPKeywordID() ==
2151         tok::pp___include_macros)
2152       return {ImportAction::None};
2153 
2154     return {ImportAction::ModuleImport, M};
2155   }
2156 
2157   case Enter:
2158     break;
2159   }
2160 
2161   // Check that we don't have infinite #include recursion.
2162   if (IncludeMacroStack.size() == MaxAllowedIncludeStackDepth-1) {
2163     Diag(FilenameTok, diag::err_pp_include_too_deep);
2164     HasReachedMaxIncludeDepth = true;
2165     return {ImportAction::None};
2166   }
2167 
2168   // Look up the file, create a File ID for it.
2169   SourceLocation IncludePos = FilenameTok.getLocation();
2170   // If the filename string was the result of macro expansions, set the include
2171   // position on the file where it will be included and after the expansions.
2172   if (IncludePos.isMacroID())
2173     IncludePos = SourceMgr.getExpansionRange(IncludePos).getEnd();
2174   FileID FID = SourceMgr.createFileID(*File, IncludePos, FileCharacter);
2175   if (!FID.isValid()) {
2176     TheModuleLoader.HadFatalFailure = true;
2177     return ImportAction::Failure;
2178   }
2179 
2180   // If all is good, enter the new file!
2181   if (EnterSourceFile(FID, CurDir, FilenameTok.getLocation()))
2182     return {ImportAction::None};
2183 
2184   // Determine if we're switching to building a new submodule, and which one.
2185   if (auto *M = SuggestedModule.getModule()) {
2186     if (M->getTopLevelModule()->ShadowingModule) {
2187       // We are building a submodule that belongs to a shadowed module. This
2188       // means we find header files in the shadowed module.
2189       Diag(M->DefinitionLoc, diag::err_module_build_shadowed_submodule)
2190         << M->getFullModuleName();
2191       Diag(M->getTopLevelModule()->ShadowingModule->DefinitionLoc,
2192            diag::note_previous_definition);
2193       return {ImportAction::None};
2194     }
2195     // When building a pch, -fmodule-name tells the compiler to textually
2196     // include headers in the specified module. We are not building the
2197     // specified module.
2198     //
2199     // FIXME: This is the wrong way to handle this. We should produce a PCH
2200     // that behaves the same as the header would behave in a compilation using
2201     // that PCH, which means we should enter the submodule. We need to teach
2202     // the AST serialization layer to deal with the resulting AST.
2203     if (getLangOpts().CompilingPCH &&
2204         isForModuleBuilding(M, getLangOpts().CurrentModule,
2205                             getLangOpts().ModuleName))
2206       return {ImportAction::None};
2207 
2208     assert(!CurLexerSubmodule && "should not have marked this as a module yet");
2209     CurLexerSubmodule = M;
2210 
2211     // Let the macro handling code know that any future macros are within
2212     // the new submodule.
2213     EnterSubmodule(M, EndLoc, /*ForPragma*/false);
2214 
2215     // Let the parser know that any future declarations are within the new
2216     // submodule.
2217     // FIXME: There's no point doing this if we're handling a #__include_macros
2218     // directive.
2219     return {ImportAction::ModuleBegin, M};
2220   }
2221 
2222   assert(!IsImportDecl && "failed to diagnose missing module for import decl");
2223   return {ImportAction::None};
2224 }
2225 
2226 /// HandleIncludeNextDirective - Implements \#include_next.
2227 ///
2228 void Preprocessor::HandleIncludeNextDirective(SourceLocation HashLoc,
2229                                               Token &IncludeNextTok) {
2230   Diag(IncludeNextTok, diag::ext_pp_include_next_directive);
2231 
2232   // #include_next is like #include, except that we start searching after
2233   // the current found directory.  If we can't do this, issue a
2234   // diagnostic.
2235   const DirectoryLookup *Lookup = CurDirLookup;
2236   const FileEntry *LookupFromFile = nullptr;
2237   if (isInPrimaryFile() && LangOpts.IsHeaderFile) {
2238     // If the main file is a header, then it's either for PCH/AST generation,
2239     // or libclang opened it. Either way, handle it as a normal include below
2240     // and do not complain about include_next.
2241   } else if (isInPrimaryFile()) {
2242     Lookup = nullptr;
2243     Diag(IncludeNextTok, diag::pp_include_next_in_primary);
2244   } else if (CurLexerSubmodule) {
2245     // Start looking up in the directory *after* the one in which the current
2246     // file would be found, if any.
2247     assert(CurPPLexer && "#include_next directive in macro?");
2248     LookupFromFile = CurPPLexer->getFileEntry();
2249     Lookup = nullptr;
2250   } else if (!Lookup) {
2251     // The current file was not found by walking the include path. Either it
2252     // is the primary file (handled above), or it was found by absolute path,
2253     // or it was found relative to such a file.
2254     // FIXME: Track enough information so we know which case we're in.
2255     Diag(IncludeNextTok, diag::pp_include_next_absolute_path);
2256   } else {
2257     // Start looking up in the next directory.
2258     ++Lookup;
2259   }
2260 
2261   return HandleIncludeDirective(HashLoc, IncludeNextTok, Lookup,
2262                                 LookupFromFile);
2263 }
2264 
2265 /// HandleMicrosoftImportDirective - Implements \#import for Microsoft Mode
2266 void Preprocessor::HandleMicrosoftImportDirective(Token &Tok) {
2267   // The Microsoft #import directive takes a type library and generates header
2268   // files from it, and includes those.  This is beyond the scope of what clang
2269   // does, so we ignore it and error out.  However, #import can optionally have
2270   // trailing attributes that span multiple lines.  We're going to eat those
2271   // so we can continue processing from there.
2272   Diag(Tok, diag::err_pp_import_directive_ms );
2273 
2274   // Read tokens until we get to the end of the directive.  Note that the
2275   // directive can be split over multiple lines using the backslash character.
2276   DiscardUntilEndOfDirective();
2277 }
2278 
2279 /// HandleImportDirective - Implements \#import.
2280 ///
2281 void Preprocessor::HandleImportDirective(SourceLocation HashLoc,
2282                                          Token &ImportTok) {
2283   if (!LangOpts.ObjC) {  // #import is standard for ObjC.
2284     if (LangOpts.MSVCCompat)
2285       return HandleMicrosoftImportDirective(ImportTok);
2286     Diag(ImportTok, diag::ext_pp_import_directive);
2287   }
2288   return HandleIncludeDirective(HashLoc, ImportTok);
2289 }
2290 
2291 /// HandleIncludeMacrosDirective - The -imacros command line option turns into a
2292 /// pseudo directive in the predefines buffer.  This handles it by sucking all
2293 /// tokens through the preprocessor and discarding them (only keeping the side
2294 /// effects on the preprocessor).
2295 void Preprocessor::HandleIncludeMacrosDirective(SourceLocation HashLoc,
2296                                                 Token &IncludeMacrosTok) {
2297   // This directive should only occur in the predefines buffer.  If not, emit an
2298   // error and reject it.
2299   SourceLocation Loc = IncludeMacrosTok.getLocation();
2300   if (SourceMgr.getBufferName(Loc) != "<built-in>") {
2301     Diag(IncludeMacrosTok.getLocation(),
2302          diag::pp_include_macros_out_of_predefines);
2303     DiscardUntilEndOfDirective();
2304     return;
2305   }
2306 
2307   // Treat this as a normal #include for checking purposes.  If this is
2308   // successful, it will push a new lexer onto the include stack.
2309   HandleIncludeDirective(HashLoc, IncludeMacrosTok);
2310 
2311   Token TmpTok;
2312   do {
2313     Lex(TmpTok);
2314     assert(TmpTok.isNot(tok::eof) && "Didn't find end of -imacros!");
2315   } while (TmpTok.isNot(tok::hashhash));
2316 }
2317 
2318 //===----------------------------------------------------------------------===//
2319 // Preprocessor Macro Directive Handling.
2320 //===----------------------------------------------------------------------===//
2321 
2322 /// ReadMacroParameterList - The ( starting a parameter list of a macro
2323 /// definition has just been read.  Lex the rest of the parameters and the
2324 /// closing ), updating MI with what we learn.  Return true if an error occurs
2325 /// parsing the param list.
2326 bool Preprocessor::ReadMacroParameterList(MacroInfo *MI, Token &Tok) {
2327   SmallVector<IdentifierInfo*, 32> Parameters;
2328 
2329   while (true) {
2330     LexUnexpandedToken(Tok);
2331     switch (Tok.getKind()) {
2332     case tok::r_paren:
2333       // Found the end of the parameter list.
2334       if (Parameters.empty())  // #define FOO()
2335         return false;
2336       // Otherwise we have #define FOO(A,)
2337       Diag(Tok, diag::err_pp_expected_ident_in_arg_list);
2338       return true;
2339     case tok::ellipsis:  // #define X(... -> C99 varargs
2340       if (!LangOpts.C99)
2341         Diag(Tok, LangOpts.CPlusPlus11 ?
2342              diag::warn_cxx98_compat_variadic_macro :
2343              diag::ext_variadic_macro);
2344 
2345       // OpenCL v1.2 s6.9.e: variadic macros are not supported.
2346       if (LangOpts.OpenCL) {
2347         Diag(Tok, diag::ext_pp_opencl_variadic_macros);
2348       }
2349 
2350       // Lex the token after the identifier.
2351       LexUnexpandedToken(Tok);
2352       if (Tok.isNot(tok::r_paren)) {
2353         Diag(Tok, diag::err_pp_missing_rparen_in_macro_def);
2354         return true;
2355       }
2356       // Add the __VA_ARGS__ identifier as a parameter.
2357       Parameters.push_back(Ident__VA_ARGS__);
2358       MI->setIsC99Varargs();
2359       MI->setParameterList(Parameters, BP);
2360       return false;
2361     case tok::eod:  // #define X(
2362       Diag(Tok, diag::err_pp_missing_rparen_in_macro_def);
2363       return true;
2364     default:
2365       // Handle keywords and identifiers here to accept things like
2366       // #define Foo(for) for.
2367       IdentifierInfo *II = Tok.getIdentifierInfo();
2368       if (!II) {
2369         // #define X(1
2370         Diag(Tok, diag::err_pp_invalid_tok_in_arg_list);
2371         return true;
2372       }
2373 
2374       // If this is already used as a parameter, it is used multiple times (e.g.
2375       // #define X(A,A.
2376       if (llvm::find(Parameters, II) != Parameters.end()) { // C99 6.10.3p6
2377         Diag(Tok, diag::err_pp_duplicate_name_in_arg_list) << II;
2378         return true;
2379       }
2380 
2381       // Add the parameter to the macro info.
2382       Parameters.push_back(II);
2383 
2384       // Lex the token after the identifier.
2385       LexUnexpandedToken(Tok);
2386 
2387       switch (Tok.getKind()) {
2388       default:          // #define X(A B
2389         Diag(Tok, diag::err_pp_expected_comma_in_arg_list);
2390         return true;
2391       case tok::r_paren: // #define X(A)
2392         MI->setParameterList(Parameters, BP);
2393         return false;
2394       case tok::comma:  // #define X(A,
2395         break;
2396       case tok::ellipsis:  // #define X(A... -> GCC extension
2397         // Diagnose extension.
2398         Diag(Tok, diag::ext_named_variadic_macro);
2399 
2400         // Lex the token after the identifier.
2401         LexUnexpandedToken(Tok);
2402         if (Tok.isNot(tok::r_paren)) {
2403           Diag(Tok, diag::err_pp_missing_rparen_in_macro_def);
2404           return true;
2405         }
2406 
2407         MI->setIsGNUVarargs();
2408         MI->setParameterList(Parameters, BP);
2409         return false;
2410       }
2411     }
2412   }
2413 }
2414 
2415 static bool isConfigurationPattern(Token &MacroName, MacroInfo *MI,
2416                                    const LangOptions &LOptions) {
2417   if (MI->getNumTokens() == 1) {
2418     const Token &Value = MI->getReplacementToken(0);
2419 
2420     // Macro that is identity, like '#define inline inline' is a valid pattern.
2421     if (MacroName.getKind() == Value.getKind())
2422       return true;
2423 
2424     // Macro that maps a keyword to the same keyword decorated with leading/
2425     // trailing underscores is a valid pattern:
2426     //    #define inline __inline
2427     //    #define inline __inline__
2428     //    #define inline _inline (in MS compatibility mode)
2429     StringRef MacroText = MacroName.getIdentifierInfo()->getName();
2430     if (IdentifierInfo *II = Value.getIdentifierInfo()) {
2431       if (!II->isKeyword(LOptions))
2432         return false;
2433       StringRef ValueText = II->getName();
2434       StringRef TrimmedValue = ValueText;
2435       if (!ValueText.startswith("__")) {
2436         if (ValueText.startswith("_"))
2437           TrimmedValue = TrimmedValue.drop_front(1);
2438         else
2439           return false;
2440       } else {
2441         TrimmedValue = TrimmedValue.drop_front(2);
2442         if (TrimmedValue.endswith("__"))
2443           TrimmedValue = TrimmedValue.drop_back(2);
2444       }
2445       return TrimmedValue.equals(MacroText);
2446     } else {
2447       return false;
2448     }
2449   }
2450 
2451   // #define inline
2452   return MacroName.isOneOf(tok::kw_extern, tok::kw_inline, tok::kw_static,
2453                            tok::kw_const) &&
2454          MI->getNumTokens() == 0;
2455 }
2456 
2457 // ReadOptionalMacroParameterListAndBody - This consumes all (i.e. the
2458 // entire line) of the macro's tokens and adds them to MacroInfo, and while
2459 // doing so performs certain validity checks including (but not limited to):
2460 //   - # (stringization) is followed by a macro parameter
2461 //
2462 //  Returns a nullptr if an invalid sequence of tokens is encountered or returns
2463 //  a pointer to a MacroInfo object.
2464 
2465 MacroInfo *Preprocessor::ReadOptionalMacroParameterListAndBody(
2466     const Token &MacroNameTok, const bool ImmediatelyAfterHeaderGuard) {
2467 
2468   Token LastTok = MacroNameTok;
2469   // Create the new macro.
2470   MacroInfo *const MI = AllocateMacroInfo(MacroNameTok.getLocation());
2471 
2472   Token Tok;
2473   LexUnexpandedToken(Tok);
2474 
2475   // Ensure we consume the rest of the macro body if errors occur.
2476   auto _ = llvm::make_scope_exit([&]() {
2477     // The flag indicates if we are still waiting for 'eod'.
2478     if (CurLexer->ParsingPreprocessorDirective)
2479       DiscardUntilEndOfDirective();
2480   });
2481 
2482   // Used to un-poison and then re-poison identifiers of the __VA_ARGS__ ilk
2483   // within their appropriate context.
2484   VariadicMacroScopeGuard VariadicMacroScopeGuard(*this);
2485 
2486   // If this is a function-like macro definition, parse the argument list,
2487   // marking each of the identifiers as being used as macro arguments.  Also,
2488   // check other constraints on the first token of the macro body.
2489   if (Tok.is(tok::eod)) {
2490     if (ImmediatelyAfterHeaderGuard) {
2491       // Save this macro information since it may part of a header guard.
2492       CurPPLexer->MIOpt.SetDefinedMacro(MacroNameTok.getIdentifierInfo(),
2493                                         MacroNameTok.getLocation());
2494     }
2495     // If there is no body to this macro, we have no special handling here.
2496   } else if (Tok.hasLeadingSpace()) {
2497     // This is a normal token with leading space.  Clear the leading space
2498     // marker on the first token to get proper expansion.
2499     Tok.clearFlag(Token::LeadingSpace);
2500   } else if (Tok.is(tok::l_paren)) {
2501     // This is a function-like macro definition.  Read the argument list.
2502     MI->setIsFunctionLike();
2503     if (ReadMacroParameterList(MI, LastTok))
2504       return nullptr;
2505 
2506     // If this is a definition of an ISO C/C++ variadic function-like macro (not
2507     // using the GNU named varargs extension) inform our variadic scope guard
2508     // which un-poisons and re-poisons certain identifiers (e.g. __VA_ARGS__)
2509     // allowed only within the definition of a variadic macro.
2510 
2511     if (MI->isC99Varargs()) {
2512       VariadicMacroScopeGuard.enterScope();
2513     }
2514 
2515     // Read the first token after the arg list for down below.
2516     LexUnexpandedToken(Tok);
2517   } else if (LangOpts.C99 || LangOpts.CPlusPlus11) {
2518     // C99 requires whitespace between the macro definition and the body.  Emit
2519     // a diagnostic for something like "#define X+".
2520     Diag(Tok, diag::ext_c99_whitespace_required_after_macro_name);
2521   } else {
2522     // C90 6.8 TC1 says: "In the definition of an object-like macro, if the
2523     // first character of a replacement list is not a character required by
2524     // subclause 5.2.1, then there shall be white-space separation between the
2525     // identifier and the replacement list.".  5.2.1 lists this set:
2526     //   "A-Za-z0-9!"#%&'()*+,_./:;<=>?[\]^_{|}~" as well as whitespace, which
2527     // is irrelevant here.
2528     bool isInvalid = false;
2529     if (Tok.is(tok::at)) // @ is not in the list above.
2530       isInvalid = true;
2531     else if (Tok.is(tok::unknown)) {
2532       // If we have an unknown token, it is something strange like "`".  Since
2533       // all of valid characters would have lexed into a single character
2534       // token of some sort, we know this is not a valid case.
2535       isInvalid = true;
2536     }
2537     if (isInvalid)
2538       Diag(Tok, diag::ext_missing_whitespace_after_macro_name);
2539     else
2540       Diag(Tok, diag::warn_missing_whitespace_after_macro_name);
2541   }
2542 
2543   if (!Tok.is(tok::eod))
2544     LastTok = Tok;
2545 
2546   // Read the rest of the macro body.
2547   if (MI->isObjectLike()) {
2548     // Object-like macros are very simple, just read their body.
2549     while (Tok.isNot(tok::eod)) {
2550       LastTok = Tok;
2551       MI->AddTokenToBody(Tok);
2552       // Get the next token of the macro.
2553       LexUnexpandedToken(Tok);
2554     }
2555   } else {
2556     // Otherwise, read the body of a function-like macro.  While we are at it,
2557     // check C99 6.10.3.2p1: ensure that # operators are followed by macro
2558     // parameters in function-like macro expansions.
2559 
2560     VAOptDefinitionContext VAOCtx(*this);
2561 
2562     while (Tok.isNot(tok::eod)) {
2563       LastTok = Tok;
2564 
2565       if (!Tok.isOneOf(tok::hash, tok::hashat, tok::hashhash)) {
2566         MI->AddTokenToBody(Tok);
2567 
2568         if (VAOCtx.isVAOptToken(Tok)) {
2569           // If we're already within a VAOPT, emit an error.
2570           if (VAOCtx.isInVAOpt()) {
2571             Diag(Tok, diag::err_pp_vaopt_nested_use);
2572             return nullptr;
2573           }
2574           // Ensure VAOPT is followed by a '(' .
2575           LexUnexpandedToken(Tok);
2576           if (Tok.isNot(tok::l_paren)) {
2577             Diag(Tok, diag::err_pp_missing_lparen_in_vaopt_use);
2578             return nullptr;
2579           }
2580           MI->AddTokenToBody(Tok);
2581           VAOCtx.sawVAOptFollowedByOpeningParens(Tok.getLocation());
2582           LexUnexpandedToken(Tok);
2583           if (Tok.is(tok::hashhash)) {
2584             Diag(Tok, diag::err_vaopt_paste_at_start);
2585             return nullptr;
2586           }
2587           continue;
2588         } else if (VAOCtx.isInVAOpt()) {
2589           if (Tok.is(tok::r_paren)) {
2590             if (VAOCtx.sawClosingParen()) {
2591               const unsigned NumTokens = MI->getNumTokens();
2592               assert(NumTokens >= 3 && "Must have seen at least __VA_OPT__( "
2593                                        "and a subsequent tok::r_paren");
2594               if (MI->getReplacementToken(NumTokens - 2).is(tok::hashhash)) {
2595                 Diag(Tok, diag::err_vaopt_paste_at_end);
2596                 return nullptr;
2597               }
2598             }
2599           } else if (Tok.is(tok::l_paren)) {
2600             VAOCtx.sawOpeningParen(Tok.getLocation());
2601           }
2602         }
2603         // Get the next token of the macro.
2604         LexUnexpandedToken(Tok);
2605         continue;
2606       }
2607 
2608       // If we're in -traditional mode, then we should ignore stringification
2609       // and token pasting. Mark the tokens as unknown so as not to confuse
2610       // things.
2611       if (getLangOpts().TraditionalCPP) {
2612         Tok.setKind(tok::unknown);
2613         MI->AddTokenToBody(Tok);
2614 
2615         // Get the next token of the macro.
2616         LexUnexpandedToken(Tok);
2617         continue;
2618       }
2619 
2620       if (Tok.is(tok::hashhash)) {
2621         // If we see token pasting, check if it looks like the gcc comma
2622         // pasting extension.  We'll use this information to suppress
2623         // diagnostics later on.
2624 
2625         // Get the next token of the macro.
2626         LexUnexpandedToken(Tok);
2627 
2628         if (Tok.is(tok::eod)) {
2629           MI->AddTokenToBody(LastTok);
2630           break;
2631         }
2632 
2633         unsigned NumTokens = MI->getNumTokens();
2634         if (NumTokens && Tok.getIdentifierInfo() == Ident__VA_ARGS__ &&
2635             MI->getReplacementToken(NumTokens-1).is(tok::comma))
2636           MI->setHasCommaPasting();
2637 
2638         // Things look ok, add the '##' token to the macro.
2639         MI->AddTokenToBody(LastTok);
2640         continue;
2641       }
2642 
2643       // Our Token is a stringization operator.
2644       // Get the next token of the macro.
2645       LexUnexpandedToken(Tok);
2646 
2647       // Check for a valid macro arg identifier or __VA_OPT__.
2648       if (!VAOCtx.isVAOptToken(Tok) &&
2649           (Tok.getIdentifierInfo() == nullptr ||
2650            MI->getParameterNum(Tok.getIdentifierInfo()) == -1)) {
2651 
2652         // If this is assembler-with-cpp mode, we accept random gibberish after
2653         // the '#' because '#' is often a comment character.  However, change
2654         // the kind of the token to tok::unknown so that the preprocessor isn't
2655         // confused.
2656         if (getLangOpts().AsmPreprocessor && Tok.isNot(tok::eod)) {
2657           LastTok.setKind(tok::unknown);
2658           MI->AddTokenToBody(LastTok);
2659           continue;
2660         } else {
2661           Diag(Tok, diag::err_pp_stringize_not_parameter)
2662             << LastTok.is(tok::hashat);
2663           return nullptr;
2664         }
2665       }
2666 
2667       // Things look ok, add the '#' and param name tokens to the macro.
2668       MI->AddTokenToBody(LastTok);
2669 
2670       // If the token following '#' is VAOPT, let the next iteration handle it
2671       // and check it for correctness, otherwise add the token and prime the
2672       // loop with the next one.
2673       if (!VAOCtx.isVAOptToken(Tok)) {
2674         MI->AddTokenToBody(Tok);
2675         LastTok = Tok;
2676 
2677         // Get the next token of the macro.
2678         LexUnexpandedToken(Tok);
2679       }
2680     }
2681     if (VAOCtx.isInVAOpt()) {
2682       assert(Tok.is(tok::eod) && "Must be at End Of preprocessing Directive");
2683       Diag(Tok, diag::err_pp_expected_after)
2684         << LastTok.getKind() << tok::r_paren;
2685       Diag(VAOCtx.getUnmatchedOpeningParenLoc(), diag::note_matching) << tok::l_paren;
2686       return nullptr;
2687     }
2688   }
2689   MI->setDefinitionEndLoc(LastTok.getLocation());
2690   return MI;
2691 }
2692 /// HandleDefineDirective - Implements \#define.  This consumes the entire macro
2693 /// line then lets the caller lex the next real token.
2694 void Preprocessor::HandleDefineDirective(
2695     Token &DefineTok, const bool ImmediatelyAfterHeaderGuard) {
2696   ++NumDefined;
2697 
2698   Token MacroNameTok;
2699   bool MacroShadowsKeyword;
2700   ReadMacroName(MacroNameTok, MU_Define, &MacroShadowsKeyword);
2701 
2702   // Error reading macro name?  If so, diagnostic already issued.
2703   if (MacroNameTok.is(tok::eod))
2704     return;
2705 
2706   // If we are supposed to keep comments in #defines, reenable comment saving
2707   // mode.
2708   if (CurLexer) CurLexer->SetCommentRetentionState(KeepMacroComments);
2709 
2710   MacroInfo *const MI = ReadOptionalMacroParameterListAndBody(
2711       MacroNameTok, ImmediatelyAfterHeaderGuard);
2712 
2713   if (!MI) return;
2714 
2715   if (MacroShadowsKeyword &&
2716       !isConfigurationPattern(MacroNameTok, MI, getLangOpts())) {
2717     Diag(MacroNameTok, diag::warn_pp_macro_hides_keyword);
2718   }
2719   // Check that there is no paste (##) operator at the beginning or end of the
2720   // replacement list.
2721   unsigned NumTokens = MI->getNumTokens();
2722   if (NumTokens != 0) {
2723     if (MI->getReplacementToken(0).is(tok::hashhash)) {
2724       Diag(MI->getReplacementToken(0), diag::err_paste_at_start);
2725       return;
2726     }
2727     if (MI->getReplacementToken(NumTokens-1).is(tok::hashhash)) {
2728       Diag(MI->getReplacementToken(NumTokens-1), diag::err_paste_at_end);
2729       return;
2730     }
2731   }
2732 
2733   // When skipping just warn about macros that do not match.
2734   if (SkippingUntilPCHThroughHeader) {
2735     const MacroInfo *OtherMI = getMacroInfo(MacroNameTok.getIdentifierInfo());
2736     if (!OtherMI || !MI->isIdenticalTo(*OtherMI, *this,
2737                              /*Syntactic=*/LangOpts.MicrosoftExt))
2738       Diag(MI->getDefinitionLoc(), diag::warn_pp_macro_def_mismatch_with_pch)
2739           << MacroNameTok.getIdentifierInfo();
2740     // Issue the diagnostic but allow the change if msvc extensions are enabled
2741     if (!LangOpts.MicrosoftExt)
2742       return;
2743   }
2744 
2745   // Finally, if this identifier already had a macro defined for it, verify that
2746   // the macro bodies are identical, and issue diagnostics if they are not.
2747   if (const MacroInfo *OtherMI=getMacroInfo(MacroNameTok.getIdentifierInfo())) {
2748     // In Objective-C, ignore attempts to directly redefine the builtin
2749     // definitions of the ownership qualifiers.  It's still possible to
2750     // #undef them.
2751     auto isObjCProtectedMacro = [](const IdentifierInfo *II) -> bool {
2752       return II->isStr("__strong") ||
2753              II->isStr("__weak") ||
2754              II->isStr("__unsafe_unretained") ||
2755              II->isStr("__autoreleasing");
2756     };
2757    if (getLangOpts().ObjC &&
2758         SourceMgr.getFileID(OtherMI->getDefinitionLoc())
2759           == getPredefinesFileID() &&
2760         isObjCProtectedMacro(MacroNameTok.getIdentifierInfo())) {
2761       // Warn if it changes the tokens.
2762       if ((!getDiagnostics().getSuppressSystemWarnings() ||
2763            !SourceMgr.isInSystemHeader(DefineTok.getLocation())) &&
2764           !MI->isIdenticalTo(*OtherMI, *this,
2765                              /*Syntactic=*/LangOpts.MicrosoftExt)) {
2766         Diag(MI->getDefinitionLoc(), diag::warn_pp_objc_macro_redef_ignored);
2767       }
2768       assert(!OtherMI->isWarnIfUnused());
2769       return;
2770     }
2771 
2772     // It is very common for system headers to have tons of macro redefinitions
2773     // and for warnings to be disabled in system headers.  If this is the case,
2774     // then don't bother calling MacroInfo::isIdenticalTo.
2775     if (!getDiagnostics().getSuppressSystemWarnings() ||
2776         !SourceMgr.isInSystemHeader(DefineTok.getLocation())) {
2777       if (!OtherMI->isUsed() && OtherMI->isWarnIfUnused())
2778         Diag(OtherMI->getDefinitionLoc(), diag::pp_macro_not_used);
2779 
2780       // Warn if defining "__LINE__" and other builtins, per C99 6.10.8/4 and
2781       // C++ [cpp.predefined]p4, but allow it as an extension.
2782       if (OtherMI->isBuiltinMacro())
2783         Diag(MacroNameTok, diag::ext_pp_redef_builtin_macro);
2784       // Macros must be identical.  This means all tokens and whitespace
2785       // separation must be the same.  C99 6.10.3p2.
2786       else if (!OtherMI->isAllowRedefinitionsWithoutWarning() &&
2787                !MI->isIdenticalTo(*OtherMI, *this, /*Syntactic=*/LangOpts.MicrosoftExt)) {
2788         Diag(MI->getDefinitionLoc(), diag::ext_pp_macro_redef)
2789           << MacroNameTok.getIdentifierInfo();
2790         Diag(OtherMI->getDefinitionLoc(), diag::note_previous_definition);
2791       }
2792     }
2793     if (OtherMI->isWarnIfUnused())
2794       WarnUnusedMacroLocs.erase(OtherMI->getDefinitionLoc());
2795   }
2796 
2797   DefMacroDirective *MD =
2798       appendDefMacroDirective(MacroNameTok.getIdentifierInfo(), MI);
2799 
2800   assert(!MI->isUsed());
2801   // If we need warning for not using the macro, add its location in the
2802   // warn-because-unused-macro set. If it gets used it will be removed from set.
2803   if (getSourceManager().isInMainFile(MI->getDefinitionLoc()) &&
2804       !Diags->isIgnored(diag::pp_macro_not_used, MI->getDefinitionLoc()) &&
2805       !MacroExpansionInDirectivesOverride) {
2806     MI->setIsWarnIfUnused(true);
2807     WarnUnusedMacroLocs.insert(MI->getDefinitionLoc());
2808   }
2809 
2810   // If the callbacks want to know, tell them about the macro definition.
2811   if (Callbacks)
2812     Callbacks->MacroDefined(MacroNameTok, MD);
2813 }
2814 
2815 /// HandleUndefDirective - Implements \#undef.
2816 ///
2817 void Preprocessor::HandleUndefDirective() {
2818   ++NumUndefined;
2819 
2820   Token MacroNameTok;
2821   ReadMacroName(MacroNameTok, MU_Undef);
2822 
2823   // Error reading macro name?  If so, diagnostic already issued.
2824   if (MacroNameTok.is(tok::eod))
2825     return;
2826 
2827   // Check to see if this is the last token on the #undef line.
2828   CheckEndOfDirective("undef");
2829 
2830   // Okay, we have a valid identifier to undef.
2831   auto *II = MacroNameTok.getIdentifierInfo();
2832   auto MD = getMacroDefinition(II);
2833   UndefMacroDirective *Undef = nullptr;
2834 
2835   // If the macro is not defined, this is a noop undef.
2836   if (const MacroInfo *MI = MD.getMacroInfo()) {
2837     if (!MI->isUsed() && MI->isWarnIfUnused())
2838       Diag(MI->getDefinitionLoc(), diag::pp_macro_not_used);
2839 
2840     if (MI->isWarnIfUnused())
2841       WarnUnusedMacroLocs.erase(MI->getDefinitionLoc());
2842 
2843     Undef = AllocateUndefMacroDirective(MacroNameTok.getLocation());
2844   }
2845 
2846   // If the callbacks want to know, tell them about the macro #undef.
2847   // Note: no matter if the macro was defined or not.
2848   if (Callbacks)
2849     Callbacks->MacroUndefined(MacroNameTok, MD, Undef);
2850 
2851   if (Undef)
2852     appendMacroDirective(II, Undef);
2853 }
2854 
2855 //===----------------------------------------------------------------------===//
2856 // Preprocessor Conditional Directive Handling.
2857 //===----------------------------------------------------------------------===//
2858 
2859 /// HandleIfdefDirective - Implements the \#ifdef/\#ifndef directive.  isIfndef
2860 /// is true when this is a \#ifndef directive.  ReadAnyTokensBeforeDirective is
2861 /// true if any tokens have been returned or pp-directives activated before this
2862 /// \#ifndef has been lexed.
2863 ///
2864 void Preprocessor::HandleIfdefDirective(Token &Result,
2865                                         const Token &HashToken,
2866                                         bool isIfndef,
2867                                         bool ReadAnyTokensBeforeDirective) {
2868   ++NumIf;
2869   Token DirectiveTok = Result;
2870 
2871   Token MacroNameTok;
2872   ReadMacroName(MacroNameTok);
2873 
2874   // Error reading macro name?  If so, diagnostic already issued.
2875   if (MacroNameTok.is(tok::eod)) {
2876     // Skip code until we get to #endif.  This helps with recovery by not
2877     // emitting an error when the #endif is reached.
2878     SkipExcludedConditionalBlock(HashToken.getLocation(),
2879                                  DirectiveTok.getLocation(),
2880                                  /*Foundnonskip*/ false, /*FoundElse*/ false);
2881     return;
2882   }
2883 
2884   // Check to see if this is the last token on the #if[n]def line.
2885   CheckEndOfDirective(isIfndef ? "ifndef" : "ifdef");
2886 
2887   IdentifierInfo *MII = MacroNameTok.getIdentifierInfo();
2888   auto MD = getMacroDefinition(MII);
2889   MacroInfo *MI = MD.getMacroInfo();
2890 
2891   if (CurPPLexer->getConditionalStackDepth() == 0) {
2892     // If the start of a top-level #ifdef and if the macro is not defined,
2893     // inform MIOpt that this might be the start of a proper include guard.
2894     // Otherwise it is some other form of unknown conditional which we can't
2895     // handle.
2896     if (!ReadAnyTokensBeforeDirective && !MI) {
2897       assert(isIfndef && "#ifdef shouldn't reach here");
2898       CurPPLexer->MIOpt.EnterTopLevelIfndef(MII, MacroNameTok.getLocation());
2899     } else
2900       CurPPLexer->MIOpt.EnterTopLevelConditional();
2901   }
2902 
2903   // If there is a macro, process it.
2904   if (MI)  // Mark it used.
2905     markMacroAsUsed(MI);
2906 
2907   if (Callbacks) {
2908     if (isIfndef)
2909       Callbacks->Ifndef(DirectiveTok.getLocation(), MacroNameTok, MD);
2910     else
2911       Callbacks->Ifdef(DirectiveTok.getLocation(), MacroNameTok, MD);
2912   }
2913 
2914   bool RetainExcludedCB = PPOpts->RetainExcludedConditionalBlocks &&
2915     getSourceManager().isInMainFile(DirectiveTok.getLocation());
2916 
2917   // Should we include the stuff contained by this directive?
2918   if (PPOpts->SingleFileParseMode && !MI) {
2919     // In 'single-file-parse mode' undefined identifiers trigger parsing of all
2920     // the directive blocks.
2921     CurPPLexer->pushConditionalLevel(DirectiveTok.getLocation(),
2922                                      /*wasskip*/false, /*foundnonskip*/false,
2923                                      /*foundelse*/false);
2924   } else if (!MI == isIfndef || RetainExcludedCB) {
2925     // Yes, remember that we are inside a conditional, then lex the next token.
2926     CurPPLexer->pushConditionalLevel(DirectiveTok.getLocation(),
2927                                      /*wasskip*/false, /*foundnonskip*/true,
2928                                      /*foundelse*/false);
2929   } else {
2930     // No, skip the contents of this block.
2931     SkipExcludedConditionalBlock(HashToken.getLocation(),
2932                                  DirectiveTok.getLocation(),
2933                                  /*Foundnonskip*/ false,
2934                                  /*FoundElse*/ false);
2935   }
2936 }
2937 
2938 /// HandleIfDirective - Implements the \#if directive.
2939 ///
2940 void Preprocessor::HandleIfDirective(Token &IfToken,
2941                                      const Token &HashToken,
2942                                      bool ReadAnyTokensBeforeDirective) {
2943   ++NumIf;
2944 
2945   // Parse and evaluate the conditional expression.
2946   IdentifierInfo *IfNDefMacro = nullptr;
2947   const DirectiveEvalResult DER = EvaluateDirectiveExpression(IfNDefMacro);
2948   const bool ConditionalTrue = DER.Conditional;
2949 
2950   // If this condition is equivalent to #ifndef X, and if this is the first
2951   // directive seen, handle it for the multiple-include optimization.
2952   if (CurPPLexer->getConditionalStackDepth() == 0) {
2953     if (!ReadAnyTokensBeforeDirective && IfNDefMacro && ConditionalTrue)
2954       // FIXME: Pass in the location of the macro name, not the 'if' token.
2955       CurPPLexer->MIOpt.EnterTopLevelIfndef(IfNDefMacro, IfToken.getLocation());
2956     else
2957       CurPPLexer->MIOpt.EnterTopLevelConditional();
2958   }
2959 
2960   if (Callbacks)
2961     Callbacks->If(
2962         IfToken.getLocation(), DER.ExprRange,
2963         (ConditionalTrue ? PPCallbacks::CVK_True : PPCallbacks::CVK_False));
2964 
2965   bool RetainExcludedCB = PPOpts->RetainExcludedConditionalBlocks &&
2966     getSourceManager().isInMainFile(IfToken.getLocation());
2967 
2968   // Should we include the stuff contained by this directive?
2969   if (PPOpts->SingleFileParseMode && DER.IncludedUndefinedIds) {
2970     // In 'single-file-parse mode' undefined identifiers trigger parsing of all
2971     // the directive blocks.
2972     CurPPLexer->pushConditionalLevel(IfToken.getLocation(), /*wasskip*/false,
2973                                      /*foundnonskip*/false, /*foundelse*/false);
2974   } else if (ConditionalTrue || RetainExcludedCB) {
2975     // Yes, remember that we are inside a conditional, then lex the next token.
2976     CurPPLexer->pushConditionalLevel(IfToken.getLocation(), /*wasskip*/false,
2977                                    /*foundnonskip*/true, /*foundelse*/false);
2978   } else {
2979     // No, skip the contents of this block.
2980     SkipExcludedConditionalBlock(HashToken.getLocation(), IfToken.getLocation(),
2981                                  /*Foundnonskip*/ false,
2982                                  /*FoundElse*/ false);
2983   }
2984 }
2985 
2986 /// HandleEndifDirective - Implements the \#endif directive.
2987 ///
2988 void Preprocessor::HandleEndifDirective(Token &EndifToken) {
2989   ++NumEndif;
2990 
2991   // Check that this is the whole directive.
2992   CheckEndOfDirective("endif");
2993 
2994   PPConditionalInfo CondInfo;
2995   if (CurPPLexer->popConditionalLevel(CondInfo)) {
2996     // No conditionals on the stack: this is an #endif without an #if.
2997     Diag(EndifToken, diag::err_pp_endif_without_if);
2998     return;
2999   }
3000 
3001   // If this the end of a top-level #endif, inform MIOpt.
3002   if (CurPPLexer->getConditionalStackDepth() == 0)
3003     CurPPLexer->MIOpt.ExitTopLevelConditional();
3004 
3005   assert(!CondInfo.WasSkipping && !CurPPLexer->LexingRawMode &&
3006          "This code should only be reachable in the non-skipping case!");
3007 
3008   if (Callbacks)
3009     Callbacks->Endif(EndifToken.getLocation(), CondInfo.IfLoc);
3010 }
3011 
3012 /// HandleElseDirective - Implements the \#else directive.
3013 ///
3014 void Preprocessor::HandleElseDirective(Token &Result, const Token &HashToken) {
3015   ++NumElse;
3016 
3017   // #else directive in a non-skipping conditional... start skipping.
3018   CheckEndOfDirective("else");
3019 
3020   PPConditionalInfo CI;
3021   if (CurPPLexer->popConditionalLevel(CI)) {
3022     Diag(Result, diag::pp_err_else_without_if);
3023     return;
3024   }
3025 
3026   // If this is a top-level #else, inform the MIOpt.
3027   if (CurPPLexer->getConditionalStackDepth() == 0)
3028     CurPPLexer->MIOpt.EnterTopLevelConditional();
3029 
3030   // If this is a #else with a #else before it, report the error.
3031   if (CI.FoundElse) Diag(Result, diag::pp_err_else_after_else);
3032 
3033   if (Callbacks)
3034     Callbacks->Else(Result.getLocation(), CI.IfLoc);
3035 
3036   bool RetainExcludedCB = PPOpts->RetainExcludedConditionalBlocks &&
3037     getSourceManager().isInMainFile(Result.getLocation());
3038 
3039   if ((PPOpts->SingleFileParseMode && !CI.FoundNonSkip) || RetainExcludedCB) {
3040     // In 'single-file-parse mode' undefined identifiers trigger parsing of all
3041     // the directive blocks.
3042     CurPPLexer->pushConditionalLevel(CI.IfLoc, /*wasskip*/false,
3043                                      /*foundnonskip*/false, /*foundelse*/true);
3044     return;
3045   }
3046 
3047   // Finally, skip the rest of the contents of this block.
3048   SkipExcludedConditionalBlock(HashToken.getLocation(), CI.IfLoc,
3049                                /*Foundnonskip*/ true,
3050                                /*FoundElse*/ true, Result.getLocation());
3051 }
3052 
3053 /// HandleElifDirective - Implements the \#elif directive.
3054 ///
3055 void Preprocessor::HandleElifDirective(Token &ElifToken,
3056                                        const Token &HashToken) {
3057   ++NumElse;
3058 
3059   // #elif directive in a non-skipping conditional... start skipping.
3060   // We don't care what the condition is, because we will always skip it (since
3061   // the block immediately before it was included).
3062   SourceRange ConditionRange = DiscardUntilEndOfDirective();
3063 
3064   PPConditionalInfo CI;
3065   if (CurPPLexer->popConditionalLevel(CI)) {
3066     Diag(ElifToken, diag::pp_err_elif_without_if);
3067     return;
3068   }
3069 
3070   // If this is a top-level #elif, inform the MIOpt.
3071   if (CurPPLexer->getConditionalStackDepth() == 0)
3072     CurPPLexer->MIOpt.EnterTopLevelConditional();
3073 
3074   // If this is a #elif with a #else before it, report the error.
3075   if (CI.FoundElse) Diag(ElifToken, diag::pp_err_elif_after_else);
3076 
3077   if (Callbacks)
3078     Callbacks->Elif(ElifToken.getLocation(), ConditionRange,
3079                     PPCallbacks::CVK_NotEvaluated, CI.IfLoc);
3080 
3081   bool RetainExcludedCB = PPOpts->RetainExcludedConditionalBlocks &&
3082     getSourceManager().isInMainFile(ElifToken.getLocation());
3083 
3084   if ((PPOpts->SingleFileParseMode && !CI.FoundNonSkip) || RetainExcludedCB) {
3085     // In 'single-file-parse mode' undefined identifiers trigger parsing of all
3086     // the directive blocks.
3087     CurPPLexer->pushConditionalLevel(ElifToken.getLocation(), /*wasskip*/false,
3088                                      /*foundnonskip*/false, /*foundelse*/false);
3089     return;
3090   }
3091 
3092   // Finally, skip the rest of the contents of this block.
3093   SkipExcludedConditionalBlock(
3094       HashToken.getLocation(), CI.IfLoc, /*Foundnonskip*/ true,
3095       /*FoundElse*/ CI.FoundElse, ElifToken.getLocation());
3096 }
3097