1 //===--- PPDirectives.cpp - Directive Handling for Preprocessor -----------===//
2 //
3 //                     The LLVM Compiler Infrastructure
4 //
5 // This file is distributed under the University of Illinois Open Source
6 // License. See LICENSE.TXT for details.
7 //
8 //===----------------------------------------------------------------------===//
9 ///
10 /// \file
11 /// \brief Implements # directive processing for the Preprocessor.
12 ///
13 //===----------------------------------------------------------------------===//
14 
15 #include "clang/Lex/Preprocessor.h"
16 #include "clang/Basic/FileManager.h"
17 #include "clang/Basic/SourceManager.h"
18 #include "clang/Lex/CodeCompletionHandler.h"
19 #include "clang/Lex/HeaderSearch.h"
20 #include "clang/Lex/HeaderSearchOptions.h"
21 #include "clang/Lex/LexDiagnostic.h"
22 #include "clang/Lex/LiteralSupport.h"
23 #include "clang/Lex/MacroInfo.h"
24 #include "clang/Lex/ModuleLoader.h"
25 #include "clang/Lex/Pragma.h"
26 #include "llvm/ADT/APInt.h"
27 #include "llvm/Support/ErrorHandling.h"
28 #include "llvm/Support/Path.h"
29 #include "llvm/Support/SaveAndRestore.h"
30 
31 using namespace clang;
32 
33 //===----------------------------------------------------------------------===//
34 // Utility Methods for Preprocessor Directive Handling.
35 //===----------------------------------------------------------------------===//
36 
37 MacroInfo *Preprocessor::AllocateMacroInfo() {
38   MacroInfoChain *MIChain = BP.Allocate<MacroInfoChain>();
39   MIChain->Next = MIChainHead;
40   MIChainHead = MIChain;
41   return &MIChain->MI;
42 }
43 
44 MacroInfo *Preprocessor::AllocateMacroInfo(SourceLocation L) {
45   MacroInfo *MI = AllocateMacroInfo();
46   new (MI) MacroInfo(L);
47   return MI;
48 }
49 
50 MacroInfo *Preprocessor::AllocateDeserializedMacroInfo(SourceLocation L,
51                                                        unsigned SubModuleID) {
52   static_assert(llvm::AlignOf<MacroInfo>::Alignment >= sizeof(SubModuleID),
53                 "alignment for MacroInfo is less than the ID");
54   DeserializedMacroInfoChain *MIChain =
55       BP.Allocate<DeserializedMacroInfoChain>();
56   MIChain->Next = DeserialMIChainHead;
57   DeserialMIChainHead = MIChain;
58 
59   MacroInfo *MI = &MIChain->MI;
60   new (MI) MacroInfo(L);
61   MI->FromASTFile = true;
62   MI->setOwningModuleID(SubModuleID);
63   return MI;
64 }
65 
66 DefMacroDirective *Preprocessor::AllocateDefMacroDirective(MacroInfo *MI,
67                                                            SourceLocation Loc) {
68   return new (BP) DefMacroDirective(MI, Loc);
69 }
70 
71 UndefMacroDirective *
72 Preprocessor::AllocateUndefMacroDirective(SourceLocation UndefLoc) {
73   return new (BP) UndefMacroDirective(UndefLoc);
74 }
75 
76 VisibilityMacroDirective *
77 Preprocessor::AllocateVisibilityMacroDirective(SourceLocation Loc,
78                                                bool isPublic) {
79   return new (BP) VisibilityMacroDirective(Loc, isPublic);
80 }
81 
82 /// \brief Read and discard all tokens remaining on the current line until
83 /// the tok::eod token is found.
84 void Preprocessor::DiscardUntilEndOfDirective() {
85   Token Tmp;
86   do {
87     LexUnexpandedToken(Tmp);
88     assert(Tmp.isNot(tok::eof) && "EOF seen while discarding directive tokens");
89   } while (Tmp.isNot(tok::eod));
90 }
91 
92 /// \brief Enumerates possible cases of #define/#undef a reserved identifier.
93 enum MacroDiag {
94   MD_NoWarn,        //> Not a reserved identifier
95   MD_KeywordDef,    //> Macro hides keyword, enabled by default
96   MD_ReservedMacro  //> #define of #undef reserved id, disabled by default
97 };
98 
99 /// \brief Checks if the specified identifier is reserved in the specified
100 /// language.
101 /// This function does not check if the identifier is a keyword.
102 static bool isReservedId(StringRef Text, const LangOptions &Lang) {
103   // C++ [macro.names], C11 7.1.3:
104   // All identifiers that begin with an underscore and either an uppercase
105   // letter or another underscore are always reserved for any use.
106   if (Text.size() >= 2 && Text[0] == '_' &&
107       (isUppercase(Text[1]) || Text[1] == '_'))
108       return true;
109   // C++ [global.names]
110   // Each name that contains a double underscore ... is reserved to the
111   // implementation for any use.
112   if (Lang.CPlusPlus) {
113     if (Text.find("__") != StringRef::npos)
114       return true;
115   }
116   return false;
117 }
118 
119 static MacroDiag shouldWarnOnMacroDef(Preprocessor &PP, IdentifierInfo *II) {
120   const LangOptions &Lang = PP.getLangOpts();
121   StringRef Text = II->getName();
122   if (isReservedId(Text, Lang))
123     return MD_ReservedMacro;
124   if (II->isKeyword(Lang))
125     return MD_KeywordDef;
126   if (Lang.CPlusPlus11 && (Text.equals("override") || Text.equals("final")))
127     return MD_KeywordDef;
128   return MD_NoWarn;
129 }
130 
131 static MacroDiag shouldWarnOnMacroUndef(Preprocessor &PP, IdentifierInfo *II) {
132   const LangOptions &Lang = PP.getLangOpts();
133   StringRef Text = II->getName();
134   // Do not warn on keyword undef.  It is generally harmless and widely used.
135   if (isReservedId(Text, Lang))
136     return MD_ReservedMacro;
137   return MD_NoWarn;
138 }
139 
140 bool Preprocessor::CheckMacroName(Token &MacroNameTok, MacroUse isDefineUndef,
141                                   bool *ShadowFlag) {
142   // Missing macro name?
143   if (MacroNameTok.is(tok::eod))
144     return Diag(MacroNameTok, diag::err_pp_missing_macro_name);
145 
146   IdentifierInfo *II = MacroNameTok.getIdentifierInfo();
147   if (!II) {
148     bool Invalid = false;
149     std::string Spelling = getSpelling(MacroNameTok, &Invalid);
150     if (Invalid)
151       return Diag(MacroNameTok, diag::err_pp_macro_not_identifier);
152     II = getIdentifierInfo(Spelling);
153 
154     if (!II->isCPlusPlusOperatorKeyword())
155       return Diag(MacroNameTok, diag::err_pp_macro_not_identifier);
156 
157     // C++ 2.5p2: Alternative tokens behave the same as its primary token
158     // except for their spellings.
159     Diag(MacroNameTok, getLangOpts().MicrosoftExt
160                            ? diag::ext_pp_operator_used_as_macro_name
161                            : diag::err_pp_operator_used_as_macro_name)
162         << II << MacroNameTok.getKind();
163 
164     // Allow #defining |and| and friends for Microsoft compatibility or
165     // recovery when legacy C headers are included in C++.
166     MacroNameTok.setIdentifierInfo(II);
167   }
168 
169   if ((isDefineUndef != MU_Other) && II->getPPKeywordID() == tok::pp_defined) {
170     // Error if defining "defined": C99 6.10.8/4, C++ [cpp.predefined]p4.
171     return Diag(MacroNameTok, diag::err_defined_macro_name);
172   }
173 
174   if (isDefineUndef == MU_Undef) {
175     auto *MI = getMacroInfo(II);
176     if (MI && MI->isBuiltinMacro()) {
177       // Warn if undefining "__LINE__" and other builtins, per C99 6.10.8/4
178       // and C++ [cpp.predefined]p4], but allow it as an extension.
179       Diag(MacroNameTok, diag::ext_pp_undef_builtin_macro);
180     }
181   }
182 
183   // If defining/undefining reserved identifier or a keyword, we need to issue
184   // a warning.
185   SourceLocation MacroNameLoc = MacroNameTok.getLocation();
186   if (ShadowFlag)
187     *ShadowFlag = false;
188   if (!SourceMgr.isInSystemHeader(MacroNameLoc) &&
189       (strcmp(SourceMgr.getBufferName(MacroNameLoc), "<built-in>") != 0)) {
190     MacroDiag D = MD_NoWarn;
191     if (isDefineUndef == MU_Define) {
192       D = shouldWarnOnMacroDef(*this, II);
193     }
194     else if (isDefineUndef == MU_Undef)
195       D = shouldWarnOnMacroUndef(*this, II);
196     if (D == MD_KeywordDef) {
197       // We do not want to warn on some patterns widely used in configuration
198       // scripts.  This requires analyzing next tokens, so do not issue warnings
199       // now, only inform caller.
200       if (ShadowFlag)
201         *ShadowFlag = true;
202     }
203     if (D == MD_ReservedMacro)
204       Diag(MacroNameTok, diag::warn_pp_macro_is_reserved_id);
205   }
206 
207   // Okay, we got a good identifier.
208   return false;
209 }
210 
211 /// \brief Lex and validate a macro name, which occurs after a
212 /// \#define or \#undef.
213 ///
214 /// This sets the token kind to eod and discards the rest of the macro line if
215 /// the macro name is invalid.
216 ///
217 /// \param MacroNameTok Token that is expected to be a macro name.
218 /// \param isDefineUndef Context in which macro is used.
219 /// \param ShadowFlag Points to a flag that is set if macro shadows a keyword.
220 void Preprocessor::ReadMacroName(Token &MacroNameTok, MacroUse isDefineUndef,
221                                  bool *ShadowFlag) {
222   // Read the token, don't allow macro expansion on it.
223   LexUnexpandedToken(MacroNameTok);
224 
225   if (MacroNameTok.is(tok::code_completion)) {
226     if (CodeComplete)
227       CodeComplete->CodeCompleteMacroName(isDefineUndef == MU_Define);
228     setCodeCompletionReached();
229     LexUnexpandedToken(MacroNameTok);
230   }
231 
232   if (!CheckMacroName(MacroNameTok, isDefineUndef, ShadowFlag))
233     return;
234 
235   // Invalid macro name, read and discard the rest of the line and set the
236   // token kind to tok::eod if necessary.
237   if (MacroNameTok.isNot(tok::eod)) {
238     MacroNameTok.setKind(tok::eod);
239     DiscardUntilEndOfDirective();
240   }
241 }
242 
243 /// \brief Ensure that the next token is a tok::eod token.
244 ///
245 /// If not, emit a diagnostic and consume up until the eod.  If EnableMacros is
246 /// true, then we consider macros that expand to zero tokens as being ok.
247 void Preprocessor::CheckEndOfDirective(const char *DirType, bool EnableMacros) {
248   Token Tmp;
249   // Lex unexpanded tokens for most directives: macros might expand to zero
250   // tokens, causing us to miss diagnosing invalid lines.  Some directives (like
251   // #line) allow empty macros.
252   if (EnableMacros)
253     Lex(Tmp);
254   else
255     LexUnexpandedToken(Tmp);
256 
257   // There should be no tokens after the directive, but we allow them as an
258   // extension.
259   while (Tmp.is(tok::comment))  // Skip comments in -C mode.
260     LexUnexpandedToken(Tmp);
261 
262   if (Tmp.isNot(tok::eod)) {
263     // Add a fixit in GNU/C99/C++ mode.  Don't offer a fixit for strict-C89,
264     // or if this is a macro-style preprocessing directive, because it is more
265     // trouble than it is worth to insert /**/ and check that there is no /**/
266     // in the range also.
267     FixItHint Hint;
268     if ((LangOpts.GNUMode || LangOpts.C99 || LangOpts.CPlusPlus) &&
269         !CurTokenLexer)
270       Hint = FixItHint::CreateInsertion(Tmp.getLocation(),"//");
271     Diag(Tmp, diag::ext_pp_extra_tokens_at_eol) << DirType << Hint;
272     DiscardUntilEndOfDirective();
273   }
274 }
275 
276 /// SkipExcludedConditionalBlock - We just read a \#if or related directive and
277 /// decided that the subsequent tokens are in the \#if'd out portion of the
278 /// file.  Lex the rest of the file, until we see an \#endif.  If
279 /// FoundNonSkipPortion is true, then we have already emitted code for part of
280 /// this \#if directive, so \#else/\#elif blocks should never be entered.
281 /// If ElseOk is true, then \#else directives are ok, if not, then we have
282 /// already seen one so a \#else directive is a duplicate.  When this returns,
283 /// the caller can lex the first valid token.
284 void Preprocessor::SkipExcludedConditionalBlock(SourceLocation IfTokenLoc,
285                                                 bool FoundNonSkipPortion,
286                                                 bool FoundElse,
287                                                 SourceLocation ElseLoc) {
288   ++NumSkipped;
289   assert(!CurTokenLexer && CurPPLexer && "Lexing a macro, not a file?");
290 
291   CurPPLexer->pushConditionalLevel(IfTokenLoc, /*isSkipping*/false,
292                                  FoundNonSkipPortion, FoundElse);
293 
294   if (CurPTHLexer) {
295     PTHSkipExcludedConditionalBlock();
296     return;
297   }
298 
299   // Enter raw mode to disable identifier lookup (and thus macro expansion),
300   // disabling warnings, etc.
301   CurPPLexer->LexingRawMode = true;
302   Token Tok;
303   while (1) {
304     CurLexer->Lex(Tok);
305 
306     if (Tok.is(tok::code_completion)) {
307       if (CodeComplete)
308         CodeComplete->CodeCompleteInConditionalExclusion();
309       setCodeCompletionReached();
310       continue;
311     }
312 
313     // If this is the end of the buffer, we have an error.
314     if (Tok.is(tok::eof)) {
315       // Emit errors for each unterminated conditional on the stack, including
316       // the current one.
317       while (!CurPPLexer->ConditionalStack.empty()) {
318         if (CurLexer->getFileLoc() != CodeCompletionFileLoc)
319           Diag(CurPPLexer->ConditionalStack.back().IfLoc,
320                diag::err_pp_unterminated_conditional);
321         CurPPLexer->ConditionalStack.pop_back();
322       }
323 
324       // Just return and let the caller lex after this #include.
325       break;
326     }
327 
328     // If this token is not a preprocessor directive, just skip it.
329     if (Tok.isNot(tok::hash) || !Tok.isAtStartOfLine())
330       continue;
331 
332     // We just parsed a # character at the start of a line, so we're in
333     // directive mode.  Tell the lexer this so any newlines we see will be
334     // converted into an EOD token (this terminates the macro).
335     CurPPLexer->ParsingPreprocessorDirective = true;
336     if (CurLexer) CurLexer->SetKeepWhitespaceMode(false);
337 
338 
339     // Read the next token, the directive flavor.
340     LexUnexpandedToken(Tok);
341 
342     // If this isn't an identifier directive (e.g. is "# 1\n" or "#\n", or
343     // something bogus), skip it.
344     if (Tok.isNot(tok::raw_identifier)) {
345       CurPPLexer->ParsingPreprocessorDirective = false;
346       // Restore comment saving mode.
347       if (CurLexer) CurLexer->resetExtendedTokenMode();
348       continue;
349     }
350 
351     // If the first letter isn't i or e, it isn't intesting to us.  We know that
352     // this is safe in the face of spelling differences, because there is no way
353     // to spell an i/e in a strange way that is another letter.  Skipping this
354     // allows us to avoid looking up the identifier info for #define/#undef and
355     // other common directives.
356     StringRef RI = Tok.getRawIdentifier();
357 
358     char FirstChar = RI[0];
359     if (FirstChar >= 'a' && FirstChar <= 'z' &&
360         FirstChar != 'i' && FirstChar != 'e') {
361       CurPPLexer->ParsingPreprocessorDirective = false;
362       // Restore comment saving mode.
363       if (CurLexer) CurLexer->resetExtendedTokenMode();
364       continue;
365     }
366 
367     // Get the identifier name without trigraphs or embedded newlines.  Note
368     // that we can't use Tok.getIdentifierInfo() because its lookup is disabled
369     // when skipping.
370     char DirectiveBuf[20];
371     StringRef Directive;
372     if (!Tok.needsCleaning() && RI.size() < 20) {
373       Directive = RI;
374     } else {
375       std::string DirectiveStr = getSpelling(Tok);
376       unsigned IdLen = DirectiveStr.size();
377       if (IdLen >= 20) {
378         CurPPLexer->ParsingPreprocessorDirective = false;
379         // Restore comment saving mode.
380         if (CurLexer) CurLexer->resetExtendedTokenMode();
381         continue;
382       }
383       memcpy(DirectiveBuf, &DirectiveStr[0], IdLen);
384       Directive = StringRef(DirectiveBuf, IdLen);
385     }
386 
387     if (Directive.startswith("if")) {
388       StringRef Sub = Directive.substr(2);
389       if (Sub.empty() ||   // "if"
390           Sub == "def" ||   // "ifdef"
391           Sub == "ndef") {  // "ifndef"
392         // We know the entire #if/#ifdef/#ifndef block will be skipped, don't
393         // bother parsing the condition.
394         DiscardUntilEndOfDirective();
395         CurPPLexer->pushConditionalLevel(Tok.getLocation(), /*wasskipping*/true,
396                                        /*foundnonskip*/false,
397                                        /*foundelse*/false);
398       }
399     } else if (Directive[0] == 'e') {
400       StringRef Sub = Directive.substr(1);
401       if (Sub == "ndif") {  // "endif"
402         PPConditionalInfo CondInfo;
403         CondInfo.WasSkipping = true; // Silence bogus warning.
404         bool InCond = CurPPLexer->popConditionalLevel(CondInfo);
405         (void)InCond;  // Silence warning in no-asserts mode.
406         assert(!InCond && "Can't be skipping if not in a conditional!");
407 
408         // If we popped the outermost skipping block, we're done skipping!
409         if (!CondInfo.WasSkipping) {
410           // Restore the value of LexingRawMode so that trailing comments
411           // are handled correctly, if we've reached the outermost block.
412           CurPPLexer->LexingRawMode = false;
413           CheckEndOfDirective("endif");
414           CurPPLexer->LexingRawMode = true;
415           if (Callbacks)
416             Callbacks->Endif(Tok.getLocation(), CondInfo.IfLoc);
417           break;
418         } else {
419           DiscardUntilEndOfDirective();
420         }
421       } else if (Sub == "lse") { // "else".
422         // #else directive in a skipping conditional.  If not in some other
423         // skipping conditional, and if #else hasn't already been seen, enter it
424         // as a non-skipping conditional.
425         PPConditionalInfo &CondInfo = CurPPLexer->peekConditionalLevel();
426 
427         // If this is a #else with a #else before it, report the error.
428         if (CondInfo.FoundElse) Diag(Tok, diag::pp_err_else_after_else);
429 
430         // Note that we've seen a #else in this conditional.
431         CondInfo.FoundElse = true;
432 
433         // If the conditional is at the top level, and the #if block wasn't
434         // entered, enter the #else block now.
435         if (!CondInfo.WasSkipping && !CondInfo.FoundNonSkip) {
436           CondInfo.FoundNonSkip = true;
437           // Restore the value of LexingRawMode so that trailing comments
438           // are handled correctly.
439           CurPPLexer->LexingRawMode = false;
440           CheckEndOfDirective("else");
441           CurPPLexer->LexingRawMode = true;
442           if (Callbacks)
443             Callbacks->Else(Tok.getLocation(), CondInfo.IfLoc);
444           break;
445         } else {
446           DiscardUntilEndOfDirective();  // C99 6.10p4.
447         }
448       } else if (Sub == "lif") {  // "elif".
449         PPConditionalInfo &CondInfo = CurPPLexer->peekConditionalLevel();
450 
451         // If this is a #elif with a #else before it, report the error.
452         if (CondInfo.FoundElse) Diag(Tok, diag::pp_err_elif_after_else);
453 
454         // If this is in a skipping block or if we're already handled this #if
455         // block, don't bother parsing the condition.
456         if (CondInfo.WasSkipping || CondInfo.FoundNonSkip) {
457           DiscardUntilEndOfDirective();
458         } else {
459           const SourceLocation CondBegin = CurPPLexer->getSourceLocation();
460           // Restore the value of LexingRawMode so that identifiers are
461           // looked up, etc, inside the #elif expression.
462           assert(CurPPLexer->LexingRawMode && "We have to be skipping here!");
463           CurPPLexer->LexingRawMode = false;
464           IdentifierInfo *IfNDefMacro = nullptr;
465           const bool CondValue = EvaluateDirectiveExpression(IfNDefMacro);
466           CurPPLexer->LexingRawMode = true;
467           if (Callbacks) {
468             const SourceLocation CondEnd = CurPPLexer->getSourceLocation();
469             Callbacks->Elif(Tok.getLocation(),
470                             SourceRange(CondBegin, CondEnd),
471                             (CondValue ? PPCallbacks::CVK_True : PPCallbacks::CVK_False), CondInfo.IfLoc);
472           }
473           // If this condition is true, enter it!
474           if (CondValue) {
475             CondInfo.FoundNonSkip = true;
476             break;
477           }
478         }
479       }
480     }
481 
482     CurPPLexer->ParsingPreprocessorDirective = false;
483     // Restore comment saving mode.
484     if (CurLexer) CurLexer->resetExtendedTokenMode();
485   }
486 
487   // Finally, if we are out of the conditional (saw an #endif or ran off the end
488   // of the file, just stop skipping and return to lexing whatever came after
489   // the #if block.
490   CurPPLexer->LexingRawMode = false;
491 
492   if (Callbacks) {
493     SourceLocation BeginLoc = ElseLoc.isValid() ? ElseLoc : IfTokenLoc;
494     Callbacks->SourceRangeSkipped(SourceRange(BeginLoc, Tok.getLocation()));
495   }
496 }
497 
498 void Preprocessor::PTHSkipExcludedConditionalBlock() {
499   while (1) {
500     assert(CurPTHLexer);
501     assert(CurPTHLexer->LexingRawMode == false);
502 
503     // Skip to the next '#else', '#elif', or #endif.
504     if (CurPTHLexer->SkipBlock()) {
505       // We have reached an #endif.  Both the '#' and 'endif' tokens
506       // have been consumed by the PTHLexer.  Just pop off the condition level.
507       PPConditionalInfo CondInfo;
508       bool InCond = CurPTHLexer->popConditionalLevel(CondInfo);
509       (void)InCond;  // Silence warning in no-asserts mode.
510       assert(!InCond && "Can't be skipping if not in a conditional!");
511       break;
512     }
513 
514     // We have reached a '#else' or '#elif'.  Lex the next token to get
515     // the directive flavor.
516     Token Tok;
517     LexUnexpandedToken(Tok);
518 
519     // We can actually look up the IdentifierInfo here since we aren't in
520     // raw mode.
521     tok::PPKeywordKind K = Tok.getIdentifierInfo()->getPPKeywordID();
522 
523     if (K == tok::pp_else) {
524       // #else: Enter the else condition.  We aren't in a nested condition
525       //  since we skip those. We're always in the one matching the last
526       //  blocked we skipped.
527       PPConditionalInfo &CondInfo = CurPTHLexer->peekConditionalLevel();
528       // Note that we've seen a #else in this conditional.
529       CondInfo.FoundElse = true;
530 
531       // If the #if block wasn't entered then enter the #else block now.
532       if (!CondInfo.FoundNonSkip) {
533         CondInfo.FoundNonSkip = true;
534 
535         // Scan until the eod token.
536         CurPTHLexer->ParsingPreprocessorDirective = true;
537         DiscardUntilEndOfDirective();
538         CurPTHLexer->ParsingPreprocessorDirective = false;
539 
540         break;
541       }
542 
543       // Otherwise skip this block.
544       continue;
545     }
546 
547     assert(K == tok::pp_elif);
548     PPConditionalInfo &CondInfo = CurPTHLexer->peekConditionalLevel();
549 
550     // If this is a #elif with a #else before it, report the error.
551     if (CondInfo.FoundElse)
552       Diag(Tok, diag::pp_err_elif_after_else);
553 
554     // If this is in a skipping block or if we're already handled this #if
555     // block, don't bother parsing the condition.  We just skip this block.
556     if (CondInfo.FoundNonSkip)
557       continue;
558 
559     // Evaluate the condition of the #elif.
560     IdentifierInfo *IfNDefMacro = nullptr;
561     CurPTHLexer->ParsingPreprocessorDirective = true;
562     bool ShouldEnter = EvaluateDirectiveExpression(IfNDefMacro);
563     CurPTHLexer->ParsingPreprocessorDirective = false;
564 
565     // If this condition is true, enter it!
566     if (ShouldEnter) {
567       CondInfo.FoundNonSkip = true;
568       break;
569     }
570 
571     // Otherwise, skip this block and go to the next one.
572   }
573 }
574 
575 Module *Preprocessor::getModuleForLocation(SourceLocation Loc) {
576   if (!SourceMgr.isInMainFile(Loc)) {
577     // Try to determine the module of the include directive.
578     // FIXME: Look into directly passing the FileEntry from LookupFile instead.
579     FileID IDOfIncl = SourceMgr.getFileID(SourceMgr.getExpansionLoc(Loc));
580     if (const FileEntry *EntryOfIncl = SourceMgr.getFileEntryForID(IDOfIncl)) {
581       // The include comes from an included file.
582       return HeaderInfo.getModuleMap()
583           .findModuleForHeader(EntryOfIncl)
584           .getModule();
585     }
586   }
587 
588   // This is either in the main file or not in a file at all. It belongs
589   // to the current module, if there is one.
590   return getLangOpts().CurrentModule.empty()
591              ? nullptr
592              : HeaderInfo.lookupModule(getLangOpts().CurrentModule);
593 }
594 
595 Module *Preprocessor::getModuleContainingLocation(SourceLocation Loc) {
596   return HeaderInfo.getModuleMap().inferModuleFromLocation(
597       FullSourceLoc(Loc, SourceMgr));
598 }
599 
600 const FileEntry *
601 Preprocessor::getModuleHeaderToIncludeForDiagnostics(SourceLocation IncLoc,
602                                                      SourceLocation Loc) {
603   // If we have a module import syntax, we shouldn't include a header to
604   // make a particular module visible.
605   if (getLangOpts().ObjC2)
606     return nullptr;
607 
608   // Figure out which module we'd want to import.
609   Module *M = getModuleContainingLocation(Loc);
610   if (!M)
611     return nullptr;
612 
613   Module *TopM = M->getTopLevelModule();
614   Module *IncM = getModuleForLocation(IncLoc);
615 
616   // Walk up through the include stack, looking through textual headers of M
617   // until we hit a non-textual header that we can #include. (We assume textual
618   // headers of a module with non-textual headers aren't meant to be used to
619   // import entities from the module.)
620   auto &SM = getSourceManager();
621   while (!Loc.isInvalid() && !SM.isInMainFile(Loc)) {
622     auto ID = SM.getFileID(SM.getExpansionLoc(Loc));
623     auto *FE = SM.getFileEntryForID(ID);
624 
625     bool InTextualHeader = false;
626     for (auto Header : HeaderInfo.getModuleMap().findAllModulesForHeader(FE)) {
627       if (!Header.getModule()->isSubModuleOf(TopM))
628         continue;
629 
630       if (!(Header.getRole() & ModuleMap::TextualHeader)) {
631         // If this is an accessible, non-textual header of M's top-level module
632         // that transitively includes the given location and makes the
633         // corresponding module visible, this is the thing to #include.
634         if (Header.isAccessibleFrom(IncM))
635           return FE;
636 
637         // It's in a private header; we can't #include it.
638         // FIXME: If there's a public header in some module that re-exports it,
639         // then we could suggest including that, but it's not clear that's the
640         // expected way to make this entity visible.
641         continue;
642       }
643 
644       InTextualHeader = true;
645     }
646 
647     if (!InTextualHeader)
648       break;
649 
650     Loc = SM.getIncludeLoc(ID);
651   }
652 
653   return nullptr;
654 }
655 
656 const FileEntry *Preprocessor::LookupFile(
657     SourceLocation FilenameLoc,
658     StringRef Filename,
659     bool isAngled,
660     const DirectoryLookup *FromDir,
661     const FileEntry *FromFile,
662     const DirectoryLookup *&CurDir,
663     SmallVectorImpl<char> *SearchPath,
664     SmallVectorImpl<char> *RelativePath,
665     ModuleMap::KnownHeader *SuggestedModule,
666     bool SkipCache) {
667   Module *RequestingModule = getModuleForLocation(FilenameLoc);
668   bool RequestingModuleIsModuleInterface = !SourceMgr.isInMainFile(FilenameLoc);
669 
670   // If the header lookup mechanism may be relative to the current inclusion
671   // stack, record the parent #includes.
672   SmallVector<std::pair<const FileEntry *, const DirectoryEntry *>, 16>
673       Includers;
674   if (!FromDir && !FromFile) {
675     FileID FID = getCurrentFileLexer()->getFileID();
676     const FileEntry *FileEnt = SourceMgr.getFileEntryForID(FID);
677 
678     // If there is no file entry associated with this file, it must be the
679     // predefines buffer or the module includes buffer. Any other file is not
680     // lexed with a normal lexer, so it won't be scanned for preprocessor
681     // directives.
682     //
683     // If we have the predefines buffer, resolve #include references (which come
684     // from the -include command line argument) from the current working
685     // directory instead of relative to the main file.
686     //
687     // If we have the module includes buffer, resolve #include references (which
688     // come from header declarations in the module map) relative to the module
689     // map file.
690     if (!FileEnt) {
691       if (FID == SourceMgr.getMainFileID() && MainFileDir)
692         Includers.push_back(std::make_pair(nullptr, MainFileDir));
693       else if ((FileEnt =
694                     SourceMgr.getFileEntryForID(SourceMgr.getMainFileID())))
695         Includers.push_back(std::make_pair(FileEnt, FileMgr.getDirectory(".")));
696     } else {
697       Includers.push_back(std::make_pair(FileEnt, FileEnt->getDir()));
698     }
699 
700     // MSVC searches the current include stack from top to bottom for
701     // headers included by quoted include directives.
702     // See: http://msdn.microsoft.com/en-us/library/36k2cdd4.aspx
703     if (LangOpts.MSVCCompat && !isAngled) {
704       for (unsigned i = 0, e = IncludeMacroStack.size(); i != e; ++i) {
705         IncludeStackInfo &ISEntry = IncludeMacroStack[e - i - 1];
706         if (IsFileLexer(ISEntry))
707           if ((FileEnt = ISEntry.ThePPLexer->getFileEntry()))
708             Includers.push_back(std::make_pair(FileEnt, FileEnt->getDir()));
709       }
710     }
711   }
712 
713   CurDir = CurDirLookup;
714 
715   if (FromFile) {
716     // We're supposed to start looking from after a particular file. Search
717     // the include path until we find that file or run out of files.
718     const DirectoryLookup *TmpCurDir = CurDir;
719     const DirectoryLookup *TmpFromDir = nullptr;
720     while (const FileEntry *FE = HeaderInfo.LookupFile(
721                Filename, FilenameLoc, isAngled, TmpFromDir, TmpCurDir,
722                Includers, SearchPath, RelativePath, RequestingModule,
723                SuggestedModule, SkipCache)) {
724       // Keep looking as if this file did a #include_next.
725       TmpFromDir = TmpCurDir;
726       ++TmpFromDir;
727       if (FE == FromFile) {
728         // Found it.
729         FromDir = TmpFromDir;
730         CurDir = TmpCurDir;
731         break;
732       }
733     }
734   }
735 
736   // Do a standard file entry lookup.
737   const FileEntry *FE = HeaderInfo.LookupFile(
738       Filename, FilenameLoc, isAngled, FromDir, CurDir, Includers, SearchPath,
739       RelativePath, RequestingModule, SuggestedModule, SkipCache);
740   if (FE) {
741     if (SuggestedModule && !LangOpts.AsmPreprocessor)
742       HeaderInfo.getModuleMap().diagnoseHeaderInclusion(
743           RequestingModule, RequestingModuleIsModuleInterface, FilenameLoc,
744           Filename, FE);
745     return FE;
746   }
747 
748   const FileEntry *CurFileEnt;
749   // Otherwise, see if this is a subframework header.  If so, this is relative
750   // to one of the headers on the #include stack.  Walk the list of the current
751   // headers on the #include stack and pass them to HeaderInfo.
752   if (IsFileLexer()) {
753     if ((CurFileEnt = CurPPLexer->getFileEntry())) {
754       if ((FE = HeaderInfo.LookupSubframeworkHeader(Filename, CurFileEnt,
755                                                     SearchPath, RelativePath,
756                                                     RequestingModule,
757                                                     SuggestedModule))) {
758         if (SuggestedModule && !LangOpts.AsmPreprocessor)
759           HeaderInfo.getModuleMap().diagnoseHeaderInclusion(
760               RequestingModule, RequestingModuleIsModuleInterface, FilenameLoc,
761               Filename, FE);
762         return FE;
763       }
764     }
765   }
766 
767   for (unsigned i = 0, e = IncludeMacroStack.size(); i != e; ++i) {
768     IncludeStackInfo &ISEntry = IncludeMacroStack[e-i-1];
769     if (IsFileLexer(ISEntry)) {
770       if ((CurFileEnt = ISEntry.ThePPLexer->getFileEntry())) {
771         if ((FE = HeaderInfo.LookupSubframeworkHeader(
772                 Filename, CurFileEnt, SearchPath, RelativePath,
773                 RequestingModule, SuggestedModule))) {
774           if (SuggestedModule && !LangOpts.AsmPreprocessor)
775             HeaderInfo.getModuleMap().diagnoseHeaderInclusion(
776                 RequestingModule, RequestingModuleIsModuleInterface,
777                 FilenameLoc, Filename, FE);
778           return FE;
779         }
780       }
781     }
782   }
783 
784   // Otherwise, we really couldn't find the file.
785   return nullptr;
786 }
787 
788 //===----------------------------------------------------------------------===//
789 // Preprocessor Directive Handling.
790 //===----------------------------------------------------------------------===//
791 
792 class Preprocessor::ResetMacroExpansionHelper {
793 public:
794   ResetMacroExpansionHelper(Preprocessor *pp)
795     : PP(pp), save(pp->DisableMacroExpansion) {
796     if (pp->MacroExpansionInDirectivesOverride)
797       pp->DisableMacroExpansion = false;
798   }
799 
800   ~ResetMacroExpansionHelper() {
801     PP->DisableMacroExpansion = save;
802   }
803 
804 private:
805   Preprocessor *PP;
806   bool save;
807 };
808 
809 /// HandleDirective - This callback is invoked when the lexer sees a # token
810 /// at the start of a line.  This consumes the directive, modifies the
811 /// lexer/preprocessor state, and advances the lexer(s) so that the next token
812 /// read is the correct one.
813 void Preprocessor::HandleDirective(Token &Result) {
814   // FIXME: Traditional: # with whitespace before it not recognized by K&R?
815 
816   // We just parsed a # character at the start of a line, so we're in directive
817   // mode.  Tell the lexer this so any newlines we see will be converted into an
818   // EOD token (which terminates the directive).
819   CurPPLexer->ParsingPreprocessorDirective = true;
820   if (CurLexer) CurLexer->SetKeepWhitespaceMode(false);
821 
822   bool ImmediatelyAfterTopLevelIfndef =
823       CurPPLexer->MIOpt.getImmediatelyAfterTopLevelIfndef();
824   CurPPLexer->MIOpt.resetImmediatelyAfterTopLevelIfndef();
825 
826   ++NumDirectives;
827 
828   // We are about to read a token.  For the multiple-include optimization FA to
829   // work, we have to remember if we had read any tokens *before* this
830   // pp-directive.
831   bool ReadAnyTokensBeforeDirective =CurPPLexer->MIOpt.getHasReadAnyTokensVal();
832 
833   // Save the '#' token in case we need to return it later.
834   Token SavedHash = Result;
835 
836   // Read the next token, the directive flavor.  This isn't expanded due to
837   // C99 6.10.3p8.
838   LexUnexpandedToken(Result);
839 
840   // C99 6.10.3p11: Is this preprocessor directive in macro invocation?  e.g.:
841   //   #define A(x) #x
842   //   A(abc
843   //     #warning blah
844   //   def)
845   // If so, the user is relying on undefined behavior, emit a diagnostic. Do
846   // not support this for #include-like directives, since that can result in
847   // terrible diagnostics, and does not work in GCC.
848   if (InMacroArgs) {
849     if (IdentifierInfo *II = Result.getIdentifierInfo()) {
850       switch (II->getPPKeywordID()) {
851       case tok::pp_include:
852       case tok::pp_import:
853       case tok::pp_include_next:
854       case tok::pp___include_macros:
855       case tok::pp_pragma:
856         Diag(Result, diag::err_embedded_directive) << II->getName();
857         DiscardUntilEndOfDirective();
858         return;
859       default:
860         break;
861       }
862     }
863     Diag(Result, diag::ext_embedded_directive);
864   }
865 
866   // Temporarily enable macro expansion if set so
867   // and reset to previous state when returning from this function.
868   ResetMacroExpansionHelper helper(this);
869 
870   switch (Result.getKind()) {
871   case tok::eod:
872     return;   // null directive.
873   case tok::code_completion:
874     if (CodeComplete)
875       CodeComplete->CodeCompleteDirective(
876                                     CurPPLexer->getConditionalStackDepth() > 0);
877     setCodeCompletionReached();
878     return;
879   case tok::numeric_constant:  // # 7  GNU line marker directive.
880     if (getLangOpts().AsmPreprocessor)
881       break;  // # 4 is not a preprocessor directive in .S files.
882     return HandleDigitDirective(Result);
883   default:
884     IdentifierInfo *II = Result.getIdentifierInfo();
885     if (!II) break; // Not an identifier.
886 
887     // Ask what the preprocessor keyword ID is.
888     switch (II->getPPKeywordID()) {
889     default: break;
890     // C99 6.10.1 - Conditional Inclusion.
891     case tok::pp_if:
892       return HandleIfDirective(Result, ReadAnyTokensBeforeDirective);
893     case tok::pp_ifdef:
894       return HandleIfdefDirective(Result, false, true/*not valid for miopt*/);
895     case tok::pp_ifndef:
896       return HandleIfdefDirective(Result, true, ReadAnyTokensBeforeDirective);
897     case tok::pp_elif:
898       return HandleElifDirective(Result);
899     case tok::pp_else:
900       return HandleElseDirective(Result);
901     case tok::pp_endif:
902       return HandleEndifDirective(Result);
903 
904     // C99 6.10.2 - Source File Inclusion.
905     case tok::pp_include:
906       // Handle #include.
907       return HandleIncludeDirective(SavedHash.getLocation(), Result);
908     case tok::pp___include_macros:
909       // Handle -imacros.
910       return HandleIncludeMacrosDirective(SavedHash.getLocation(), Result);
911 
912     // C99 6.10.3 - Macro Replacement.
913     case tok::pp_define:
914       return HandleDefineDirective(Result, ImmediatelyAfterTopLevelIfndef);
915     case tok::pp_undef:
916       return HandleUndefDirective(Result);
917 
918     // C99 6.10.4 - Line Control.
919     case tok::pp_line:
920       return HandleLineDirective(Result);
921 
922     // C99 6.10.5 - Error Directive.
923     case tok::pp_error:
924       return HandleUserDiagnosticDirective(Result, false);
925 
926     // C99 6.10.6 - Pragma Directive.
927     case tok::pp_pragma:
928       return HandlePragmaDirective(SavedHash.getLocation(), PIK_HashPragma);
929 
930     // GNU Extensions.
931     case tok::pp_import:
932       return HandleImportDirective(SavedHash.getLocation(), Result);
933     case tok::pp_include_next:
934       return HandleIncludeNextDirective(SavedHash.getLocation(), Result);
935 
936     case tok::pp_warning:
937       Diag(Result, diag::ext_pp_warning_directive);
938       return HandleUserDiagnosticDirective(Result, true);
939     case tok::pp_ident:
940       return HandleIdentSCCSDirective(Result);
941     case tok::pp_sccs:
942       return HandleIdentSCCSDirective(Result);
943     case tok::pp_assert:
944       //isExtension = true;  // FIXME: implement #assert
945       break;
946     case tok::pp_unassert:
947       //isExtension = true;  // FIXME: implement #unassert
948       break;
949 
950     case tok::pp___public_macro:
951       if (getLangOpts().Modules)
952         return HandleMacroPublicDirective(Result);
953       break;
954 
955     case tok::pp___private_macro:
956       if (getLangOpts().Modules)
957         return HandleMacroPrivateDirective(Result);
958       break;
959     }
960     break;
961   }
962 
963   // If this is a .S file, treat unknown # directives as non-preprocessor
964   // directives.  This is important because # may be a comment or introduce
965   // various pseudo-ops.  Just return the # token and push back the following
966   // token to be lexed next time.
967   if (getLangOpts().AsmPreprocessor) {
968     auto Toks = llvm::make_unique<Token[]>(2);
969     // Return the # and the token after it.
970     Toks[0] = SavedHash;
971     Toks[1] = Result;
972 
973     // If the second token is a hashhash token, then we need to translate it to
974     // unknown so the token lexer doesn't try to perform token pasting.
975     if (Result.is(tok::hashhash))
976       Toks[1].setKind(tok::unknown);
977 
978     // Enter this token stream so that we re-lex the tokens.  Make sure to
979     // enable macro expansion, in case the token after the # is an identifier
980     // that is expanded.
981     EnterTokenStream(std::move(Toks), 2, false);
982     return;
983   }
984 
985   // If we reached here, the preprocessing token is not valid!
986   Diag(Result, diag::err_pp_invalid_directive);
987 
988   // Read the rest of the PP line.
989   DiscardUntilEndOfDirective();
990 
991   // Okay, we're done parsing the directive.
992 }
993 
994 /// GetLineValue - Convert a numeric token into an unsigned value, emitting
995 /// Diagnostic DiagID if it is invalid, and returning the value in Val.
996 static bool GetLineValue(Token &DigitTok, unsigned &Val,
997                          unsigned DiagID, Preprocessor &PP,
998                          bool IsGNULineDirective=false) {
999   if (DigitTok.isNot(tok::numeric_constant)) {
1000     PP.Diag(DigitTok, DiagID);
1001 
1002     if (DigitTok.isNot(tok::eod))
1003       PP.DiscardUntilEndOfDirective();
1004     return true;
1005   }
1006 
1007   SmallString<64> IntegerBuffer;
1008   IntegerBuffer.resize(DigitTok.getLength());
1009   const char *DigitTokBegin = &IntegerBuffer[0];
1010   bool Invalid = false;
1011   unsigned ActualLength = PP.getSpelling(DigitTok, DigitTokBegin, &Invalid);
1012   if (Invalid)
1013     return true;
1014 
1015   // Verify that we have a simple digit-sequence, and compute the value.  This
1016   // is always a simple digit string computed in decimal, so we do this manually
1017   // here.
1018   Val = 0;
1019   for (unsigned i = 0; i != ActualLength; ++i) {
1020     // C++1y [lex.fcon]p1:
1021     //   Optional separating single quotes in a digit-sequence are ignored
1022     if (DigitTokBegin[i] == '\'')
1023       continue;
1024 
1025     if (!isDigit(DigitTokBegin[i])) {
1026       PP.Diag(PP.AdvanceToTokenCharacter(DigitTok.getLocation(), i),
1027               diag::err_pp_line_digit_sequence) << IsGNULineDirective;
1028       PP.DiscardUntilEndOfDirective();
1029       return true;
1030     }
1031 
1032     unsigned NextVal = Val*10+(DigitTokBegin[i]-'0');
1033     if (NextVal < Val) { // overflow.
1034       PP.Diag(DigitTok, DiagID);
1035       PP.DiscardUntilEndOfDirective();
1036       return true;
1037     }
1038     Val = NextVal;
1039   }
1040 
1041   if (DigitTokBegin[0] == '0' && Val)
1042     PP.Diag(DigitTok.getLocation(), diag::warn_pp_line_decimal)
1043       << IsGNULineDirective;
1044 
1045   return false;
1046 }
1047 
1048 /// \brief Handle a \#line directive: C99 6.10.4.
1049 ///
1050 /// The two acceptable forms are:
1051 /// \verbatim
1052 ///   # line digit-sequence
1053 ///   # line digit-sequence "s-char-sequence"
1054 /// \endverbatim
1055 void Preprocessor::HandleLineDirective(Token &Tok) {
1056   // Read the line # and string argument.  Per C99 6.10.4p5, these tokens are
1057   // expanded.
1058   Token DigitTok;
1059   Lex(DigitTok);
1060 
1061   // Validate the number and convert it to an unsigned.
1062   unsigned LineNo;
1063   if (GetLineValue(DigitTok, LineNo, diag::err_pp_line_requires_integer,*this))
1064     return;
1065 
1066   if (LineNo == 0)
1067     Diag(DigitTok, diag::ext_pp_line_zero);
1068 
1069   // Enforce C99 6.10.4p3: "The digit sequence shall not specify ... a
1070   // number greater than 2147483647".  C90 requires that the line # be <= 32767.
1071   unsigned LineLimit = 32768U;
1072   if (LangOpts.C99 || LangOpts.CPlusPlus11)
1073     LineLimit = 2147483648U;
1074   if (LineNo >= LineLimit)
1075     Diag(DigitTok, diag::ext_pp_line_too_big) << LineLimit;
1076   else if (LangOpts.CPlusPlus11 && LineNo >= 32768U)
1077     Diag(DigitTok, diag::warn_cxx98_compat_pp_line_too_big);
1078 
1079   int FilenameID = -1;
1080   Token StrTok;
1081   Lex(StrTok);
1082 
1083   // If the StrTok is "eod", then it wasn't present.  Otherwise, it must be a
1084   // string followed by eod.
1085   if (StrTok.is(tok::eod))
1086     ; // ok
1087   else if (StrTok.isNot(tok::string_literal)) {
1088     Diag(StrTok, diag::err_pp_line_invalid_filename);
1089     return DiscardUntilEndOfDirective();
1090   } else if (StrTok.hasUDSuffix()) {
1091     Diag(StrTok, diag::err_invalid_string_udl);
1092     return DiscardUntilEndOfDirective();
1093   } else {
1094     // Parse and validate the string, converting it into a unique ID.
1095     StringLiteralParser Literal(StrTok, *this);
1096     assert(Literal.isAscii() && "Didn't allow wide strings in");
1097     if (Literal.hadError)
1098       return DiscardUntilEndOfDirective();
1099     if (Literal.Pascal) {
1100       Diag(StrTok, diag::err_pp_linemarker_invalid_filename);
1101       return DiscardUntilEndOfDirective();
1102     }
1103     FilenameID = SourceMgr.getLineTableFilenameID(Literal.GetString());
1104 
1105     // Verify that there is nothing after the string, other than EOD.  Because
1106     // of C99 6.10.4p5, macros that expand to empty tokens are ok.
1107     CheckEndOfDirective("line", true);
1108   }
1109 
1110   SourceMgr.AddLineNote(DigitTok.getLocation(), LineNo, FilenameID);
1111 
1112   if (Callbacks)
1113     Callbacks->FileChanged(CurPPLexer->getSourceLocation(),
1114                            PPCallbacks::RenameFile,
1115                            SrcMgr::C_User);
1116 }
1117 
1118 /// ReadLineMarkerFlags - Parse and validate any flags at the end of a GNU line
1119 /// marker directive.
1120 static bool ReadLineMarkerFlags(bool &IsFileEntry, bool &IsFileExit,
1121                                 bool &IsSystemHeader, bool &IsExternCHeader,
1122                                 Preprocessor &PP) {
1123   unsigned FlagVal;
1124   Token FlagTok;
1125   PP.Lex(FlagTok);
1126   if (FlagTok.is(tok::eod)) return false;
1127   if (GetLineValue(FlagTok, FlagVal, diag::err_pp_linemarker_invalid_flag, PP))
1128     return true;
1129 
1130   if (FlagVal == 1) {
1131     IsFileEntry = true;
1132 
1133     PP.Lex(FlagTok);
1134     if (FlagTok.is(tok::eod)) return false;
1135     if (GetLineValue(FlagTok, FlagVal, diag::err_pp_linemarker_invalid_flag,PP))
1136       return true;
1137   } else if (FlagVal == 2) {
1138     IsFileExit = true;
1139 
1140     SourceManager &SM = PP.getSourceManager();
1141     // If we are leaving the current presumed file, check to make sure the
1142     // presumed include stack isn't empty!
1143     FileID CurFileID =
1144       SM.getDecomposedExpansionLoc(FlagTok.getLocation()).first;
1145     PresumedLoc PLoc = SM.getPresumedLoc(FlagTok.getLocation());
1146     if (PLoc.isInvalid())
1147       return true;
1148 
1149     // If there is no include loc (main file) or if the include loc is in a
1150     // different physical file, then we aren't in a "1" line marker flag region.
1151     SourceLocation IncLoc = PLoc.getIncludeLoc();
1152     if (IncLoc.isInvalid() ||
1153         SM.getDecomposedExpansionLoc(IncLoc).first != CurFileID) {
1154       PP.Diag(FlagTok, diag::err_pp_linemarker_invalid_pop);
1155       PP.DiscardUntilEndOfDirective();
1156       return true;
1157     }
1158 
1159     PP.Lex(FlagTok);
1160     if (FlagTok.is(tok::eod)) return false;
1161     if (GetLineValue(FlagTok, FlagVal, diag::err_pp_linemarker_invalid_flag,PP))
1162       return true;
1163   }
1164 
1165   // We must have 3 if there are still flags.
1166   if (FlagVal != 3) {
1167     PP.Diag(FlagTok, diag::err_pp_linemarker_invalid_flag);
1168     PP.DiscardUntilEndOfDirective();
1169     return true;
1170   }
1171 
1172   IsSystemHeader = true;
1173 
1174   PP.Lex(FlagTok);
1175   if (FlagTok.is(tok::eod)) return false;
1176   if (GetLineValue(FlagTok, FlagVal, diag::err_pp_linemarker_invalid_flag, PP))
1177     return true;
1178 
1179   // We must have 4 if there is yet another flag.
1180   if (FlagVal != 4) {
1181     PP.Diag(FlagTok, diag::err_pp_linemarker_invalid_flag);
1182     PP.DiscardUntilEndOfDirective();
1183     return true;
1184   }
1185 
1186   IsExternCHeader = true;
1187 
1188   PP.Lex(FlagTok);
1189   if (FlagTok.is(tok::eod)) return false;
1190 
1191   // There are no more valid flags here.
1192   PP.Diag(FlagTok, diag::err_pp_linemarker_invalid_flag);
1193   PP.DiscardUntilEndOfDirective();
1194   return true;
1195 }
1196 
1197 /// HandleDigitDirective - Handle a GNU line marker directive, whose syntax is
1198 /// one of the following forms:
1199 ///
1200 ///     # 42
1201 ///     # 42 "file" ('1' | '2')?
1202 ///     # 42 "file" ('1' | '2')? '3' '4'?
1203 ///
1204 void Preprocessor::HandleDigitDirective(Token &DigitTok) {
1205   // Validate the number and convert it to an unsigned.  GNU does not have a
1206   // line # limit other than it fit in 32-bits.
1207   unsigned LineNo;
1208   if (GetLineValue(DigitTok, LineNo, diag::err_pp_linemarker_requires_integer,
1209                    *this, true))
1210     return;
1211 
1212   Token StrTok;
1213   Lex(StrTok);
1214 
1215   bool IsFileEntry = false, IsFileExit = false;
1216   bool IsSystemHeader = false, IsExternCHeader = false;
1217   int FilenameID = -1;
1218 
1219   // If the StrTok is "eod", then it wasn't present.  Otherwise, it must be a
1220   // string followed by eod.
1221   if (StrTok.is(tok::eod))
1222     ; // ok
1223   else if (StrTok.isNot(tok::string_literal)) {
1224     Diag(StrTok, diag::err_pp_linemarker_invalid_filename);
1225     return DiscardUntilEndOfDirective();
1226   } else if (StrTok.hasUDSuffix()) {
1227     Diag(StrTok, diag::err_invalid_string_udl);
1228     return DiscardUntilEndOfDirective();
1229   } else {
1230     // Parse and validate the string, converting it into a unique ID.
1231     StringLiteralParser Literal(StrTok, *this);
1232     assert(Literal.isAscii() && "Didn't allow wide strings in");
1233     if (Literal.hadError)
1234       return DiscardUntilEndOfDirective();
1235     if (Literal.Pascal) {
1236       Diag(StrTok, diag::err_pp_linemarker_invalid_filename);
1237       return DiscardUntilEndOfDirective();
1238     }
1239     FilenameID = SourceMgr.getLineTableFilenameID(Literal.GetString());
1240 
1241     // If a filename was present, read any flags that are present.
1242     if (ReadLineMarkerFlags(IsFileEntry, IsFileExit,
1243                             IsSystemHeader, IsExternCHeader, *this))
1244       return;
1245   }
1246 
1247   // Create a line note with this information.
1248   SourceMgr.AddLineNote(DigitTok.getLocation(), LineNo, FilenameID,
1249                         IsFileEntry, IsFileExit,
1250                         IsSystemHeader, IsExternCHeader);
1251 
1252   // If the preprocessor has callbacks installed, notify them of the #line
1253   // change.  This is used so that the line marker comes out in -E mode for
1254   // example.
1255   if (Callbacks) {
1256     PPCallbacks::FileChangeReason Reason = PPCallbacks::RenameFile;
1257     if (IsFileEntry)
1258       Reason = PPCallbacks::EnterFile;
1259     else if (IsFileExit)
1260       Reason = PPCallbacks::ExitFile;
1261     SrcMgr::CharacteristicKind FileKind = SrcMgr::C_User;
1262     if (IsExternCHeader)
1263       FileKind = SrcMgr::C_ExternCSystem;
1264     else if (IsSystemHeader)
1265       FileKind = SrcMgr::C_System;
1266 
1267     Callbacks->FileChanged(CurPPLexer->getSourceLocation(), Reason, FileKind);
1268   }
1269 }
1270 
1271 /// HandleUserDiagnosticDirective - Handle a #warning or #error directive.
1272 ///
1273 void Preprocessor::HandleUserDiagnosticDirective(Token &Tok,
1274                                                  bool isWarning) {
1275   // PTH doesn't emit #warning or #error directives.
1276   if (CurPTHLexer)
1277     return CurPTHLexer->DiscardToEndOfLine();
1278 
1279   // Read the rest of the line raw.  We do this because we don't want macros
1280   // to be expanded and we don't require that the tokens be valid preprocessing
1281   // tokens.  For example, this is allowed: "#warning `   'foo".  GCC does
1282   // collapse multiple consequtive white space between tokens, but this isn't
1283   // specified by the standard.
1284   SmallString<128> Message;
1285   CurLexer->ReadToEndOfLine(&Message);
1286 
1287   // Find the first non-whitespace character, so that we can make the
1288   // diagnostic more succinct.
1289   StringRef Msg = StringRef(Message).ltrim(' ');
1290 
1291   if (isWarning)
1292     Diag(Tok, diag::pp_hash_warning) << Msg;
1293   else
1294     Diag(Tok, diag::err_pp_hash_error) << Msg;
1295 }
1296 
1297 /// HandleIdentSCCSDirective - Handle a #ident/#sccs directive.
1298 ///
1299 void Preprocessor::HandleIdentSCCSDirective(Token &Tok) {
1300   // Yes, this directive is an extension.
1301   Diag(Tok, diag::ext_pp_ident_directive);
1302 
1303   // Read the string argument.
1304   Token StrTok;
1305   Lex(StrTok);
1306 
1307   // If the token kind isn't a string, it's a malformed directive.
1308   if (StrTok.isNot(tok::string_literal) &&
1309       StrTok.isNot(tok::wide_string_literal)) {
1310     Diag(StrTok, diag::err_pp_malformed_ident);
1311     if (StrTok.isNot(tok::eod))
1312       DiscardUntilEndOfDirective();
1313     return;
1314   }
1315 
1316   if (StrTok.hasUDSuffix()) {
1317     Diag(StrTok, diag::err_invalid_string_udl);
1318     return DiscardUntilEndOfDirective();
1319   }
1320 
1321   // Verify that there is nothing after the string, other than EOD.
1322   CheckEndOfDirective("ident");
1323 
1324   if (Callbacks) {
1325     bool Invalid = false;
1326     std::string Str = getSpelling(StrTok, &Invalid);
1327     if (!Invalid)
1328       Callbacks->Ident(Tok.getLocation(), Str);
1329   }
1330 }
1331 
1332 /// \brief Handle a #public directive.
1333 void Preprocessor::HandleMacroPublicDirective(Token &Tok) {
1334   Token MacroNameTok;
1335   ReadMacroName(MacroNameTok, MU_Undef);
1336 
1337   // Error reading macro name?  If so, diagnostic already issued.
1338   if (MacroNameTok.is(tok::eod))
1339     return;
1340 
1341   // Check to see if this is the last token on the #__public_macro line.
1342   CheckEndOfDirective("__public_macro");
1343 
1344   IdentifierInfo *II = MacroNameTok.getIdentifierInfo();
1345   // Okay, we finally have a valid identifier to undef.
1346   MacroDirective *MD = getLocalMacroDirective(II);
1347 
1348   // If the macro is not defined, this is an error.
1349   if (!MD) {
1350     Diag(MacroNameTok, diag::err_pp_visibility_non_macro) << II;
1351     return;
1352   }
1353 
1354   // Note that this macro has now been exported.
1355   appendMacroDirective(II, AllocateVisibilityMacroDirective(
1356                                 MacroNameTok.getLocation(), /*IsPublic=*/true));
1357 }
1358 
1359 /// \brief Handle a #private directive.
1360 void Preprocessor::HandleMacroPrivateDirective(Token &Tok) {
1361   Token MacroNameTok;
1362   ReadMacroName(MacroNameTok, MU_Undef);
1363 
1364   // Error reading macro name?  If so, diagnostic already issued.
1365   if (MacroNameTok.is(tok::eod))
1366     return;
1367 
1368   // Check to see if this is the last token on the #__private_macro line.
1369   CheckEndOfDirective("__private_macro");
1370 
1371   IdentifierInfo *II = MacroNameTok.getIdentifierInfo();
1372   // Okay, we finally have a valid identifier to undef.
1373   MacroDirective *MD = getLocalMacroDirective(II);
1374 
1375   // If the macro is not defined, this is an error.
1376   if (!MD) {
1377     Diag(MacroNameTok, diag::err_pp_visibility_non_macro) << II;
1378     return;
1379   }
1380 
1381   // Note that this macro has now been marked private.
1382   appendMacroDirective(II, AllocateVisibilityMacroDirective(
1383                                MacroNameTok.getLocation(), /*IsPublic=*/false));
1384 }
1385 
1386 //===----------------------------------------------------------------------===//
1387 // Preprocessor Include Directive Handling.
1388 //===----------------------------------------------------------------------===//
1389 
1390 /// GetIncludeFilenameSpelling - Turn the specified lexer token into a fully
1391 /// checked and spelled filename, e.g. as an operand of \#include. This returns
1392 /// true if the input filename was in <>'s or false if it were in ""'s.  The
1393 /// caller is expected to provide a buffer that is large enough to hold the
1394 /// spelling of the filename, but is also expected to handle the case when
1395 /// this method decides to use a different buffer.
1396 bool Preprocessor::GetIncludeFilenameSpelling(SourceLocation Loc,
1397                                               StringRef &Buffer) {
1398   // Get the text form of the filename.
1399   assert(!Buffer.empty() && "Can't have tokens with empty spellings!");
1400 
1401   // Make sure the filename is <x> or "x".
1402   bool isAngled;
1403   if (Buffer[0] == '<') {
1404     if (Buffer.back() != '>') {
1405       Diag(Loc, diag::err_pp_expects_filename);
1406       Buffer = StringRef();
1407       return true;
1408     }
1409     isAngled = true;
1410   } else if (Buffer[0] == '"') {
1411     if (Buffer.back() != '"') {
1412       Diag(Loc, diag::err_pp_expects_filename);
1413       Buffer = StringRef();
1414       return true;
1415     }
1416     isAngled = false;
1417   } else {
1418     Diag(Loc, diag::err_pp_expects_filename);
1419     Buffer = StringRef();
1420     return true;
1421   }
1422 
1423   // Diagnose #include "" as invalid.
1424   if (Buffer.size() <= 2) {
1425     Diag(Loc, diag::err_pp_empty_filename);
1426     Buffer = StringRef();
1427     return true;
1428   }
1429 
1430   // Skip the brackets.
1431   Buffer = Buffer.substr(1, Buffer.size()-2);
1432   return isAngled;
1433 }
1434 
1435 // \brief Handle cases where the \#include name is expanded from a macro
1436 // as multiple tokens, which need to be glued together.
1437 //
1438 // This occurs for code like:
1439 // \code
1440 //    \#define FOO <a/b.h>
1441 //    \#include FOO
1442 // \endcode
1443 // because in this case, "<a/b.h>" is returned as 7 tokens, not one.
1444 //
1445 // This code concatenates and consumes tokens up to the '>' token.  It returns
1446 // false if the > was found, otherwise it returns true if it finds and consumes
1447 // the EOD marker.
1448 bool Preprocessor::ConcatenateIncludeName(SmallString<128> &FilenameBuffer,
1449                                           SourceLocation &End) {
1450   Token CurTok;
1451 
1452   Lex(CurTok);
1453   while (CurTok.isNot(tok::eod)) {
1454     End = CurTok.getLocation();
1455 
1456     // FIXME: Provide code completion for #includes.
1457     if (CurTok.is(tok::code_completion)) {
1458       setCodeCompletionReached();
1459       Lex(CurTok);
1460       continue;
1461     }
1462 
1463     // Append the spelling of this token to the buffer. If there was a space
1464     // before it, add it now.
1465     if (CurTok.hasLeadingSpace())
1466       FilenameBuffer.push_back(' ');
1467 
1468     // Get the spelling of the token, directly into FilenameBuffer if possible.
1469     unsigned PreAppendSize = FilenameBuffer.size();
1470     FilenameBuffer.resize(PreAppendSize+CurTok.getLength());
1471 
1472     const char *BufPtr = &FilenameBuffer[PreAppendSize];
1473     unsigned ActualLen = getSpelling(CurTok, BufPtr);
1474 
1475     // If the token was spelled somewhere else, copy it into FilenameBuffer.
1476     if (BufPtr != &FilenameBuffer[PreAppendSize])
1477       memcpy(&FilenameBuffer[PreAppendSize], BufPtr, ActualLen);
1478 
1479     // Resize FilenameBuffer to the correct size.
1480     if (CurTok.getLength() != ActualLen)
1481       FilenameBuffer.resize(PreAppendSize+ActualLen);
1482 
1483     // If we found the '>' marker, return success.
1484     if (CurTok.is(tok::greater))
1485       return false;
1486 
1487     Lex(CurTok);
1488   }
1489 
1490   // If we hit the eod marker, emit an error and return true so that the caller
1491   // knows the EOD has been read.
1492   Diag(CurTok.getLocation(), diag::err_pp_expects_filename);
1493   return true;
1494 }
1495 
1496 /// \brief Push a token onto the token stream containing an annotation.
1497 static void EnterAnnotationToken(Preprocessor &PP,
1498                                  SourceLocation Begin, SourceLocation End,
1499                                  tok::TokenKind Kind, void *AnnotationVal) {
1500   // FIXME: Produce this as the current token directly, rather than
1501   // allocating a new token for it.
1502   auto Tok = llvm::make_unique<Token[]>(1);
1503   Tok[0].startToken();
1504   Tok[0].setKind(Kind);
1505   Tok[0].setLocation(Begin);
1506   Tok[0].setAnnotationEndLoc(End);
1507   Tok[0].setAnnotationValue(AnnotationVal);
1508   PP.EnterTokenStream(std::move(Tok), 1, true);
1509 }
1510 
1511 /// \brief Produce a diagnostic informing the user that a #include or similar
1512 /// was implicitly treated as a module import.
1513 static void diagnoseAutoModuleImport(
1514     Preprocessor &PP, SourceLocation HashLoc, Token &IncludeTok,
1515     ArrayRef<std::pair<IdentifierInfo *, SourceLocation>> Path,
1516     SourceLocation PathEnd) {
1517   assert(PP.getLangOpts().ObjC2 && "no import syntax available");
1518 
1519   SmallString<128> PathString;
1520   for (unsigned I = 0, N = Path.size(); I != N; ++I) {
1521     if (I)
1522       PathString += '.';
1523     PathString += Path[I].first->getName();
1524   }
1525   int IncludeKind = 0;
1526 
1527   switch (IncludeTok.getIdentifierInfo()->getPPKeywordID()) {
1528   case tok::pp_include:
1529     IncludeKind = 0;
1530     break;
1531 
1532   case tok::pp_import:
1533     IncludeKind = 1;
1534     break;
1535 
1536   case tok::pp_include_next:
1537     IncludeKind = 2;
1538     break;
1539 
1540   case tok::pp___include_macros:
1541     IncludeKind = 3;
1542     break;
1543 
1544   default:
1545     llvm_unreachable("unknown include directive kind");
1546   }
1547 
1548   CharSourceRange ReplaceRange(SourceRange(HashLoc, PathEnd),
1549                                /*IsTokenRange=*/false);
1550   PP.Diag(HashLoc, diag::warn_auto_module_import)
1551       << IncludeKind << PathString
1552       << FixItHint::CreateReplacement(ReplaceRange,
1553                                       ("@import " + PathString + ";").str());
1554 }
1555 
1556 /// HandleIncludeDirective - The "\#include" tokens have just been read, read
1557 /// the file to be included from the lexer, then include it!  This is a common
1558 /// routine with functionality shared between \#include, \#include_next and
1559 /// \#import.  LookupFrom is set when this is a \#include_next directive, it
1560 /// specifies the file to start searching from.
1561 void Preprocessor::HandleIncludeDirective(SourceLocation HashLoc,
1562                                           Token &IncludeTok,
1563                                           const DirectoryLookup *LookupFrom,
1564                                           const FileEntry *LookupFromFile,
1565                                           bool isImport) {
1566   Token FilenameTok;
1567   CurPPLexer->LexIncludeFilename(FilenameTok);
1568 
1569   // Reserve a buffer to get the spelling.
1570   SmallString<128> FilenameBuffer;
1571   StringRef Filename;
1572   SourceLocation End;
1573   SourceLocation CharEnd; // the end of this directive, in characters
1574 
1575   switch (FilenameTok.getKind()) {
1576   case tok::eod:
1577     // If the token kind is EOD, the error has already been diagnosed.
1578     return;
1579 
1580   case tok::angle_string_literal:
1581   case tok::string_literal:
1582     Filename = getSpelling(FilenameTok, FilenameBuffer);
1583     End = FilenameTok.getLocation();
1584     CharEnd = End.getLocWithOffset(FilenameTok.getLength());
1585     break;
1586 
1587   case tok::less:
1588     // This could be a <foo/bar.h> file coming from a macro expansion.  In this
1589     // case, glue the tokens together into FilenameBuffer and interpret those.
1590     FilenameBuffer.push_back('<');
1591     if (ConcatenateIncludeName(FilenameBuffer, End))
1592       return;   // Found <eod> but no ">"?  Diagnostic already emitted.
1593     Filename = FilenameBuffer;
1594     CharEnd = End.getLocWithOffset(1);
1595     break;
1596   default:
1597     Diag(FilenameTok.getLocation(), diag::err_pp_expects_filename);
1598     DiscardUntilEndOfDirective();
1599     return;
1600   }
1601 
1602   CharSourceRange FilenameRange
1603     = CharSourceRange::getCharRange(FilenameTok.getLocation(), CharEnd);
1604   StringRef OriginalFilename = Filename;
1605   bool isAngled =
1606     GetIncludeFilenameSpelling(FilenameTok.getLocation(), Filename);
1607   // If GetIncludeFilenameSpelling set the start ptr to null, there was an
1608   // error.
1609   if (Filename.empty()) {
1610     DiscardUntilEndOfDirective();
1611     return;
1612   }
1613 
1614   // Verify that there is nothing after the filename, other than EOD.  Note that
1615   // we allow macros that expand to nothing after the filename, because this
1616   // falls into the category of "#include pp-tokens new-line" specified in
1617   // C99 6.10.2p4.
1618   CheckEndOfDirective(IncludeTok.getIdentifierInfo()->getNameStart(), true);
1619 
1620   // Check that we don't have infinite #include recursion.
1621   if (IncludeMacroStack.size() == MaxAllowedIncludeStackDepth-1) {
1622     Diag(FilenameTok, diag::err_pp_include_too_deep);
1623     return;
1624   }
1625 
1626   // Complain about attempts to #include files in an audit pragma.
1627   if (PragmaARCCFCodeAuditedLoc.isValid()) {
1628     Diag(HashLoc, diag::err_pp_include_in_arc_cf_code_audited);
1629     Diag(PragmaARCCFCodeAuditedLoc, diag::note_pragma_entered_here);
1630 
1631     // Immediately leave the pragma.
1632     PragmaARCCFCodeAuditedLoc = SourceLocation();
1633   }
1634 
1635   // Complain about attempts to #include files in an assume-nonnull pragma.
1636   if (PragmaAssumeNonNullLoc.isValid()) {
1637     Diag(HashLoc, diag::err_pp_include_in_assume_nonnull);
1638     Diag(PragmaAssumeNonNullLoc, diag::note_pragma_entered_here);
1639 
1640     // Immediately leave the pragma.
1641     PragmaAssumeNonNullLoc = SourceLocation();
1642   }
1643 
1644   if (HeaderInfo.HasIncludeAliasMap()) {
1645     // Map the filename with the brackets still attached.  If the name doesn't
1646     // map to anything, fall back on the filename we've already gotten the
1647     // spelling for.
1648     StringRef NewName = HeaderInfo.MapHeaderToIncludeAlias(OriginalFilename);
1649     if (!NewName.empty())
1650       Filename = NewName;
1651   }
1652 
1653   // Search include directories.
1654   const DirectoryLookup *CurDir;
1655   SmallString<1024> SearchPath;
1656   SmallString<1024> RelativePath;
1657   // We get the raw path only if we have 'Callbacks' to which we later pass
1658   // the path.
1659   ModuleMap::KnownHeader SuggestedModule;
1660   SourceLocation FilenameLoc = FilenameTok.getLocation();
1661   SmallString<128> NormalizedPath;
1662   if (LangOpts.MSVCCompat) {
1663     NormalizedPath = Filename.str();
1664 #ifndef LLVM_ON_WIN32
1665     llvm::sys::path::native(NormalizedPath);
1666 #endif
1667   }
1668   const FileEntry *File = LookupFile(
1669       FilenameLoc, LangOpts.MSVCCompat ? NormalizedPath.c_str() : Filename,
1670       isAngled, LookupFrom, LookupFromFile, CurDir,
1671       Callbacks ? &SearchPath : nullptr, Callbacks ? &RelativePath : nullptr,
1672       &SuggestedModule);
1673 
1674   if (!File) {
1675     if (Callbacks) {
1676       // Give the clients a chance to recover.
1677       SmallString<128> RecoveryPath;
1678       if (Callbacks->FileNotFound(Filename, RecoveryPath)) {
1679         if (const DirectoryEntry *DE = FileMgr.getDirectory(RecoveryPath)) {
1680           // Add the recovery path to the list of search paths.
1681           DirectoryLookup DL(DE, SrcMgr::C_User, false);
1682           HeaderInfo.AddSearchPath(DL, isAngled);
1683 
1684           // Try the lookup again, skipping the cache.
1685           File = LookupFile(
1686               FilenameLoc,
1687               LangOpts.MSVCCompat ? NormalizedPath.c_str() : Filename, isAngled,
1688               LookupFrom, LookupFromFile, CurDir, nullptr, nullptr,
1689               &SuggestedModule, /*SkipCache*/ true);
1690         }
1691       }
1692     }
1693 
1694     if (!SuppressIncludeNotFoundError) {
1695       // If the file could not be located and it was included via angle
1696       // brackets, we can attempt a lookup as though it were a quoted path to
1697       // provide the user with a possible fixit.
1698       if (isAngled) {
1699         File = LookupFile(
1700             FilenameLoc,
1701             LangOpts.MSVCCompat ? NormalizedPath.c_str() : Filename, false,
1702             LookupFrom, LookupFromFile, CurDir,
1703             Callbacks ? &SearchPath : nullptr,
1704             Callbacks ? &RelativePath : nullptr,
1705             &SuggestedModule);
1706         if (File) {
1707           SourceRange Range(FilenameTok.getLocation(), CharEnd);
1708           Diag(FilenameTok, diag::err_pp_file_not_found_not_fatal) <<
1709             Filename <<
1710             FixItHint::CreateReplacement(Range, "\"" + Filename.str() + "\"");
1711         }
1712       }
1713 
1714       // If the file is still not found, just go with the vanilla diagnostic
1715       if (!File)
1716         Diag(FilenameTok, diag::err_pp_file_not_found) << Filename;
1717     }
1718   }
1719 
1720   // Should we enter the source file? Set to false if either the source file is
1721   // known to have no effect beyond its effect on module visibility -- that is,
1722   // if it's got an include guard that is already defined or is a modular header
1723   // we've imported or already built.
1724   bool ShouldEnter = true;
1725 
1726   // Determine whether we should try to import the module for this #include, if
1727   // there is one. Don't do so if precompiled module support is disabled or we
1728   // are processing this module textually (because we're building the module).
1729   if (File && SuggestedModule && getLangOpts().Modules &&
1730       SuggestedModule.getModule()->getTopLevelModuleName() !=
1731           getLangOpts().CurrentModule) {
1732     // If this include corresponds to a module but that module is
1733     // unavailable, diagnose the situation and bail out.
1734     // FIXME: Remove this; loadModule does the same check (but produces
1735     // slightly worse diagnostics).
1736     if (!SuggestedModule.getModule()->isAvailable() &&
1737         !SuggestedModule.getModule()
1738              ->getTopLevelModule()
1739              ->HasIncompatibleModuleFile) {
1740       clang::Module::Requirement Requirement;
1741       clang::Module::UnresolvedHeaderDirective MissingHeader;
1742       Module *M = SuggestedModule.getModule();
1743       // Identify the cause.
1744       (void)M->isAvailable(getLangOpts(), getTargetInfo(), Requirement,
1745                            MissingHeader);
1746       if (MissingHeader.FileNameLoc.isValid()) {
1747         Diag(MissingHeader.FileNameLoc, diag::err_module_header_missing)
1748             << MissingHeader.IsUmbrella << MissingHeader.FileName;
1749       } else {
1750         Diag(M->DefinitionLoc, diag::err_module_unavailable)
1751             << M->getFullModuleName() << Requirement.second << Requirement.first;
1752       }
1753       Diag(FilenameTok.getLocation(),
1754            diag::note_implicit_top_level_module_import_here)
1755           << M->getTopLevelModuleName();
1756       return;
1757     }
1758 
1759     // Compute the module access path corresponding to this module.
1760     // FIXME: Should we have a second loadModule() overload to avoid this
1761     // extra lookup step?
1762     SmallVector<std::pair<IdentifierInfo *, SourceLocation>, 2> Path;
1763     for (Module *Mod = SuggestedModule.getModule(); Mod; Mod = Mod->Parent)
1764       Path.push_back(std::make_pair(getIdentifierInfo(Mod->Name),
1765                                     FilenameTok.getLocation()));
1766     std::reverse(Path.begin(), Path.end());
1767 
1768     // Warn that we're replacing the include/import with a module import.
1769     // We only do this in Objective-C, where we have a module-import syntax.
1770     if (getLangOpts().ObjC2)
1771       diagnoseAutoModuleImport(*this, HashLoc, IncludeTok, Path, CharEnd);
1772 
1773     // Load the module to import its macros. We'll make the declarations
1774     // visible when the parser gets here.
1775     // FIXME: Pass SuggestedModule in here rather than converting it to a path
1776     // and making the module loader convert it back again.
1777     ModuleLoadResult Imported = TheModuleLoader.loadModule(
1778         IncludeTok.getLocation(), Path, Module::Hidden,
1779         /*IsIncludeDirective=*/true);
1780     assert((Imported == nullptr || Imported == SuggestedModule.getModule()) &&
1781            "the imported module is different than the suggested one");
1782 
1783     if (Imported)
1784       ShouldEnter = false;
1785     else if (Imported.isMissingExpected()) {
1786       // We failed to find a submodule that we assumed would exist (because it
1787       // was in the directory of an umbrella header, for instance), but no
1788       // actual module exists for it (because the umbrella header is
1789       // incomplete).  Treat this as a textual inclusion.
1790       SuggestedModule = ModuleMap::KnownHeader();
1791     } else {
1792       // We hit an error processing the import. Bail out.
1793       if (hadModuleLoaderFatalFailure()) {
1794         // With a fatal failure in the module loader, we abort parsing.
1795         Token &Result = IncludeTok;
1796         if (CurLexer) {
1797           Result.startToken();
1798           CurLexer->FormTokenWithChars(Result, CurLexer->BufferEnd, tok::eof);
1799           CurLexer->cutOffLexing();
1800         } else {
1801           assert(CurPTHLexer && "#include but no current lexer set!");
1802           CurPTHLexer->getEOF(Result);
1803         }
1804       }
1805       return;
1806     }
1807   }
1808 
1809   if (Callbacks) {
1810     // Notify the callback object that we've seen an inclusion directive.
1811     Callbacks->InclusionDirective(
1812         HashLoc, IncludeTok,
1813         LangOpts.MSVCCompat ? NormalizedPath.c_str() : Filename, isAngled,
1814         FilenameRange, File, SearchPath, RelativePath,
1815         ShouldEnter ? nullptr : SuggestedModule.getModule());
1816   }
1817 
1818   if (!File)
1819     return;
1820 
1821   // The #included file will be considered to be a system header if either it is
1822   // in a system include directory, or if the #includer is a system include
1823   // header.
1824   SrcMgr::CharacteristicKind FileCharacter =
1825     std::max(HeaderInfo.getFileDirFlavor(File),
1826              SourceMgr.getFileCharacteristic(FilenameTok.getLocation()));
1827 
1828   // FIXME: If we have a suggested module, and we've already visited this file,
1829   // don't bother entering it again. We know it has no further effect.
1830 
1831   // Ask HeaderInfo if we should enter this #include file.  If not, #including
1832   // this file will have no effect.
1833   if (ShouldEnter &&
1834       !HeaderInfo.ShouldEnterIncludeFile(*this, File, isImport,
1835                                          SuggestedModule.getModule())) {
1836     ShouldEnter = false;
1837     if (Callbacks)
1838       Callbacks->FileSkipped(*File, FilenameTok, FileCharacter);
1839   }
1840 
1841   // If we don't need to enter the file, stop now.
1842   if (!ShouldEnter) {
1843     // If this is a module import, make it visible if needed.
1844     if (auto *M = SuggestedModule.getModule()) {
1845       makeModuleVisible(M, HashLoc);
1846 
1847       if (IncludeTok.getIdentifierInfo()->getPPKeywordID() !=
1848           tok::pp___include_macros)
1849         EnterAnnotationToken(*this, HashLoc, End, tok::annot_module_include, M);
1850     }
1851     return;
1852   }
1853 
1854   // Look up the file, create a File ID for it.
1855   SourceLocation IncludePos = End;
1856   // If the filename string was the result of macro expansions, set the include
1857   // position on the file where it will be included and after the expansions.
1858   if (IncludePos.isMacroID())
1859     IncludePos = SourceMgr.getExpansionRange(IncludePos).second;
1860   FileID FID = SourceMgr.createFileID(File, IncludePos, FileCharacter);
1861   assert(FID.isValid() && "Expected valid file ID");
1862 
1863   // If all is good, enter the new file!
1864   if (EnterSourceFile(FID, CurDir, FilenameTok.getLocation()))
1865     return;
1866 
1867   // Determine if we're switching to building a new submodule, and which one.
1868   if (auto *M = SuggestedModule.getModule()) {
1869     assert(!CurSubmodule && "should not have marked this as a module yet");
1870     CurSubmodule = M;
1871 
1872     // Let the macro handling code know that any future macros are within
1873     // the new submodule.
1874     EnterSubmodule(M, HashLoc);
1875 
1876     // Let the parser know that any future declarations are within the new
1877     // submodule.
1878     // FIXME: There's no point doing this if we're handling a #__include_macros
1879     // directive.
1880     EnterAnnotationToken(*this, HashLoc, End, tok::annot_module_begin, M);
1881   }
1882 }
1883 
1884 /// HandleIncludeNextDirective - Implements \#include_next.
1885 ///
1886 void Preprocessor::HandleIncludeNextDirective(SourceLocation HashLoc,
1887                                               Token &IncludeNextTok) {
1888   Diag(IncludeNextTok, diag::ext_pp_include_next_directive);
1889 
1890   // #include_next is like #include, except that we start searching after
1891   // the current found directory.  If we can't do this, issue a
1892   // diagnostic.
1893   const DirectoryLookup *Lookup = CurDirLookup;
1894   const FileEntry *LookupFromFile = nullptr;
1895   if (isInPrimaryFile()) {
1896     Lookup = nullptr;
1897     Diag(IncludeNextTok, diag::pp_include_next_in_primary);
1898   } else if (CurSubmodule) {
1899     // Start looking up in the directory *after* the one in which the current
1900     // file would be found, if any.
1901     assert(CurPPLexer && "#include_next directive in macro?");
1902     LookupFromFile = CurPPLexer->getFileEntry();
1903     Lookup = nullptr;
1904   } else if (!Lookup) {
1905     Diag(IncludeNextTok, diag::pp_include_next_absolute_path);
1906   } else {
1907     // Start looking up in the next directory.
1908     ++Lookup;
1909   }
1910 
1911   return HandleIncludeDirective(HashLoc, IncludeNextTok, Lookup,
1912                                 LookupFromFile);
1913 }
1914 
1915 /// HandleMicrosoftImportDirective - Implements \#import for Microsoft Mode
1916 void Preprocessor::HandleMicrosoftImportDirective(Token &Tok) {
1917   // The Microsoft #import directive takes a type library and generates header
1918   // files from it, and includes those.  This is beyond the scope of what clang
1919   // does, so we ignore it and error out.  However, #import can optionally have
1920   // trailing attributes that span multiple lines.  We're going to eat those
1921   // so we can continue processing from there.
1922   Diag(Tok, diag::err_pp_import_directive_ms );
1923 
1924   // Read tokens until we get to the end of the directive.  Note that the
1925   // directive can be split over multiple lines using the backslash character.
1926   DiscardUntilEndOfDirective();
1927 }
1928 
1929 /// HandleImportDirective - Implements \#import.
1930 ///
1931 void Preprocessor::HandleImportDirective(SourceLocation HashLoc,
1932                                          Token &ImportTok) {
1933   if (!LangOpts.ObjC1) {  // #import is standard for ObjC.
1934     if (LangOpts.MSVCCompat)
1935       return HandleMicrosoftImportDirective(ImportTok);
1936     Diag(ImportTok, diag::ext_pp_import_directive);
1937   }
1938   return HandleIncludeDirective(HashLoc, ImportTok, nullptr, nullptr, true);
1939 }
1940 
1941 /// HandleIncludeMacrosDirective - The -imacros command line option turns into a
1942 /// pseudo directive in the predefines buffer.  This handles it by sucking all
1943 /// tokens through the preprocessor and discarding them (only keeping the side
1944 /// effects on the preprocessor).
1945 void Preprocessor::HandleIncludeMacrosDirective(SourceLocation HashLoc,
1946                                                 Token &IncludeMacrosTok) {
1947   // This directive should only occur in the predefines buffer.  If not, emit an
1948   // error and reject it.
1949   SourceLocation Loc = IncludeMacrosTok.getLocation();
1950   if (strcmp(SourceMgr.getBufferName(Loc), "<built-in>") != 0) {
1951     Diag(IncludeMacrosTok.getLocation(),
1952          diag::pp_include_macros_out_of_predefines);
1953     DiscardUntilEndOfDirective();
1954     return;
1955   }
1956 
1957   // Treat this as a normal #include for checking purposes.  If this is
1958   // successful, it will push a new lexer onto the include stack.
1959   HandleIncludeDirective(HashLoc, IncludeMacrosTok);
1960 
1961   Token TmpTok;
1962   do {
1963     Lex(TmpTok);
1964     assert(TmpTok.isNot(tok::eof) && "Didn't find end of -imacros!");
1965   } while (TmpTok.isNot(tok::hashhash));
1966 }
1967 
1968 //===----------------------------------------------------------------------===//
1969 // Preprocessor Macro Directive Handling.
1970 //===----------------------------------------------------------------------===//
1971 
1972 /// ReadMacroDefinitionArgList - The ( starting an argument list of a macro
1973 /// definition has just been read.  Lex the rest of the arguments and the
1974 /// closing ), updating MI with what we learn.  Return true if an error occurs
1975 /// parsing the arg list.
1976 bool Preprocessor::ReadMacroDefinitionArgList(MacroInfo *MI, Token &Tok) {
1977   SmallVector<IdentifierInfo*, 32> Arguments;
1978 
1979   while (1) {
1980     LexUnexpandedToken(Tok);
1981     switch (Tok.getKind()) {
1982     case tok::r_paren:
1983       // Found the end of the argument list.
1984       if (Arguments.empty())  // #define FOO()
1985         return false;
1986       // Otherwise we have #define FOO(A,)
1987       Diag(Tok, diag::err_pp_expected_ident_in_arg_list);
1988       return true;
1989     case tok::ellipsis:  // #define X(... -> C99 varargs
1990       if (!LangOpts.C99)
1991         Diag(Tok, LangOpts.CPlusPlus11 ?
1992              diag::warn_cxx98_compat_variadic_macro :
1993              diag::ext_variadic_macro);
1994 
1995       // OpenCL v1.2 s6.9.e: variadic macros are not supported.
1996       if (LangOpts.OpenCL) {
1997         Diag(Tok, diag::err_pp_opencl_variadic_macros);
1998         return true;
1999       }
2000 
2001       // Lex the token after the identifier.
2002       LexUnexpandedToken(Tok);
2003       if (Tok.isNot(tok::r_paren)) {
2004         Diag(Tok, diag::err_pp_missing_rparen_in_macro_def);
2005         return true;
2006       }
2007       // Add the __VA_ARGS__ identifier as an argument.
2008       Arguments.push_back(Ident__VA_ARGS__);
2009       MI->setIsC99Varargs();
2010       MI->setArgumentList(Arguments, BP);
2011       return false;
2012     case tok::eod:  // #define X(
2013       Diag(Tok, diag::err_pp_missing_rparen_in_macro_def);
2014       return true;
2015     default:
2016       // Handle keywords and identifiers here to accept things like
2017       // #define Foo(for) for.
2018       IdentifierInfo *II = Tok.getIdentifierInfo();
2019       if (!II) {
2020         // #define X(1
2021         Diag(Tok, diag::err_pp_invalid_tok_in_arg_list);
2022         return true;
2023       }
2024 
2025       // If this is already used as an argument, it is used multiple times (e.g.
2026       // #define X(A,A.
2027       if (std::find(Arguments.begin(), Arguments.end(), II) !=
2028           Arguments.end()) {  // C99 6.10.3p6
2029         Diag(Tok, diag::err_pp_duplicate_name_in_arg_list) << II;
2030         return true;
2031       }
2032 
2033       // Add the argument to the macro info.
2034       Arguments.push_back(II);
2035 
2036       // Lex the token after the identifier.
2037       LexUnexpandedToken(Tok);
2038 
2039       switch (Tok.getKind()) {
2040       default:          // #define X(A B
2041         Diag(Tok, diag::err_pp_expected_comma_in_arg_list);
2042         return true;
2043       case tok::r_paren: // #define X(A)
2044         MI->setArgumentList(Arguments, BP);
2045         return false;
2046       case tok::comma:  // #define X(A,
2047         break;
2048       case tok::ellipsis:  // #define X(A... -> GCC extension
2049         // Diagnose extension.
2050         Diag(Tok, diag::ext_named_variadic_macro);
2051 
2052         // Lex the token after the identifier.
2053         LexUnexpandedToken(Tok);
2054         if (Tok.isNot(tok::r_paren)) {
2055           Diag(Tok, diag::err_pp_missing_rparen_in_macro_def);
2056           return true;
2057         }
2058 
2059         MI->setIsGNUVarargs();
2060         MI->setArgumentList(Arguments, BP);
2061         return false;
2062       }
2063     }
2064   }
2065 }
2066 
2067 static bool isConfigurationPattern(Token &MacroName, MacroInfo *MI,
2068                                    const LangOptions &LOptions) {
2069   if (MI->getNumTokens() == 1) {
2070     const Token &Value = MI->getReplacementToken(0);
2071 
2072     // Macro that is identity, like '#define inline inline' is a valid pattern.
2073     if (MacroName.getKind() == Value.getKind())
2074       return true;
2075 
2076     // Macro that maps a keyword to the same keyword decorated with leading/
2077     // trailing underscores is a valid pattern:
2078     //    #define inline __inline
2079     //    #define inline __inline__
2080     //    #define inline _inline (in MS compatibility mode)
2081     StringRef MacroText = MacroName.getIdentifierInfo()->getName();
2082     if (IdentifierInfo *II = Value.getIdentifierInfo()) {
2083       if (!II->isKeyword(LOptions))
2084         return false;
2085       StringRef ValueText = II->getName();
2086       StringRef TrimmedValue = ValueText;
2087       if (!ValueText.startswith("__")) {
2088         if (ValueText.startswith("_"))
2089           TrimmedValue = TrimmedValue.drop_front(1);
2090         else
2091           return false;
2092       } else {
2093         TrimmedValue = TrimmedValue.drop_front(2);
2094         if (TrimmedValue.endswith("__"))
2095           TrimmedValue = TrimmedValue.drop_back(2);
2096       }
2097       return TrimmedValue.equals(MacroText);
2098     } else {
2099       return false;
2100     }
2101   }
2102 
2103   // #define inline
2104   return MacroName.isOneOf(tok::kw_extern, tok::kw_inline, tok::kw_static,
2105                            tok::kw_const) &&
2106          MI->getNumTokens() == 0;
2107 }
2108 
2109 /// HandleDefineDirective - Implements \#define.  This consumes the entire macro
2110 /// line then lets the caller lex the next real token.
2111 void Preprocessor::HandleDefineDirective(Token &DefineTok,
2112                                          bool ImmediatelyAfterHeaderGuard) {
2113   ++NumDefined;
2114 
2115   Token MacroNameTok;
2116   bool MacroShadowsKeyword;
2117   ReadMacroName(MacroNameTok, MU_Define, &MacroShadowsKeyword);
2118 
2119   // Error reading macro name?  If so, diagnostic already issued.
2120   if (MacroNameTok.is(tok::eod))
2121     return;
2122 
2123   Token LastTok = MacroNameTok;
2124 
2125   // If we are supposed to keep comments in #defines, reenable comment saving
2126   // mode.
2127   if (CurLexer) CurLexer->SetCommentRetentionState(KeepMacroComments);
2128 
2129   // Create the new macro.
2130   MacroInfo *MI = AllocateMacroInfo(MacroNameTok.getLocation());
2131 
2132   Token Tok;
2133   LexUnexpandedToken(Tok);
2134 
2135   // If this is a function-like macro definition, parse the argument list,
2136   // marking each of the identifiers as being used as macro arguments.  Also,
2137   // check other constraints on the first token of the macro body.
2138   if (Tok.is(tok::eod)) {
2139     if (ImmediatelyAfterHeaderGuard) {
2140       // Save this macro information since it may part of a header guard.
2141       CurPPLexer->MIOpt.SetDefinedMacro(MacroNameTok.getIdentifierInfo(),
2142                                         MacroNameTok.getLocation());
2143     }
2144     // If there is no body to this macro, we have no special handling here.
2145   } else if (Tok.hasLeadingSpace()) {
2146     // This is a normal token with leading space.  Clear the leading space
2147     // marker on the first token to get proper expansion.
2148     Tok.clearFlag(Token::LeadingSpace);
2149   } else if (Tok.is(tok::l_paren)) {
2150     // This is a function-like macro definition.  Read the argument list.
2151     MI->setIsFunctionLike();
2152     if (ReadMacroDefinitionArgList(MI, LastTok)) {
2153       // Throw away the rest of the line.
2154       if (CurPPLexer->ParsingPreprocessorDirective)
2155         DiscardUntilEndOfDirective();
2156       return;
2157     }
2158 
2159     // If this is a definition of a variadic C99 function-like macro, not using
2160     // the GNU named varargs extension, enabled __VA_ARGS__.
2161 
2162     // "Poison" __VA_ARGS__, which can only appear in the expansion of a macro.
2163     // This gets unpoisoned where it is allowed.
2164     assert(Ident__VA_ARGS__->isPoisoned() && "__VA_ARGS__ should be poisoned!");
2165     if (MI->isC99Varargs())
2166       Ident__VA_ARGS__->setIsPoisoned(false);
2167 
2168     // Read the first token after the arg list for down below.
2169     LexUnexpandedToken(Tok);
2170   } else if (LangOpts.C99 || LangOpts.CPlusPlus11) {
2171     // C99 requires whitespace between the macro definition and the body.  Emit
2172     // a diagnostic for something like "#define X+".
2173     Diag(Tok, diag::ext_c99_whitespace_required_after_macro_name);
2174   } else {
2175     // C90 6.8 TC1 says: "In the definition of an object-like macro, if the
2176     // first character of a replacement list is not a character required by
2177     // subclause 5.2.1, then there shall be white-space separation between the
2178     // identifier and the replacement list.".  5.2.1 lists this set:
2179     //   "A-Za-z0-9!"#%&'()*+,_./:;<=>?[\]^_{|}~" as well as whitespace, which
2180     // is irrelevant here.
2181     bool isInvalid = false;
2182     if (Tok.is(tok::at)) // @ is not in the list above.
2183       isInvalid = true;
2184     else if (Tok.is(tok::unknown)) {
2185       // If we have an unknown token, it is something strange like "`".  Since
2186       // all of valid characters would have lexed into a single character
2187       // token of some sort, we know this is not a valid case.
2188       isInvalid = true;
2189     }
2190     if (isInvalid)
2191       Diag(Tok, diag::ext_missing_whitespace_after_macro_name);
2192     else
2193       Diag(Tok, diag::warn_missing_whitespace_after_macro_name);
2194   }
2195 
2196   if (!Tok.is(tok::eod))
2197     LastTok = Tok;
2198 
2199   // Read the rest of the macro body.
2200   if (MI->isObjectLike()) {
2201     // Object-like macros are very simple, just read their body.
2202     while (Tok.isNot(tok::eod)) {
2203       LastTok = Tok;
2204       MI->AddTokenToBody(Tok);
2205       // Get the next token of the macro.
2206       LexUnexpandedToken(Tok);
2207     }
2208   } else {
2209     // Otherwise, read the body of a function-like macro.  While we are at it,
2210     // check C99 6.10.3.2p1: ensure that # operators are followed by macro
2211     // parameters in function-like macro expansions.
2212     while (Tok.isNot(tok::eod)) {
2213       LastTok = Tok;
2214 
2215       if (!Tok.isOneOf(tok::hash, tok::hashat, tok::hashhash)) {
2216         MI->AddTokenToBody(Tok);
2217 
2218         // Get the next token of the macro.
2219         LexUnexpandedToken(Tok);
2220         continue;
2221       }
2222 
2223       // If we're in -traditional mode, then we should ignore stringification
2224       // and token pasting. Mark the tokens as unknown so as not to confuse
2225       // things.
2226       if (getLangOpts().TraditionalCPP) {
2227         Tok.setKind(tok::unknown);
2228         MI->AddTokenToBody(Tok);
2229 
2230         // Get the next token of the macro.
2231         LexUnexpandedToken(Tok);
2232         continue;
2233       }
2234 
2235       if (Tok.is(tok::hashhash)) {
2236         // If we see token pasting, check if it looks like the gcc comma
2237         // pasting extension.  We'll use this information to suppress
2238         // diagnostics later on.
2239 
2240         // Get the next token of the macro.
2241         LexUnexpandedToken(Tok);
2242 
2243         if (Tok.is(tok::eod)) {
2244           MI->AddTokenToBody(LastTok);
2245           break;
2246         }
2247 
2248         unsigned NumTokens = MI->getNumTokens();
2249         if (NumTokens && Tok.getIdentifierInfo() == Ident__VA_ARGS__ &&
2250             MI->getReplacementToken(NumTokens-1).is(tok::comma))
2251           MI->setHasCommaPasting();
2252 
2253         // Things look ok, add the '##' token to the macro.
2254         MI->AddTokenToBody(LastTok);
2255         continue;
2256       }
2257 
2258       // Get the next token of the macro.
2259       LexUnexpandedToken(Tok);
2260 
2261       // Check for a valid macro arg identifier.
2262       if (Tok.getIdentifierInfo() == nullptr ||
2263           MI->getArgumentNum(Tok.getIdentifierInfo()) == -1) {
2264 
2265         // If this is assembler-with-cpp mode, we accept random gibberish after
2266         // the '#' because '#' is often a comment character.  However, change
2267         // the kind of the token to tok::unknown so that the preprocessor isn't
2268         // confused.
2269         if (getLangOpts().AsmPreprocessor && Tok.isNot(tok::eod)) {
2270           LastTok.setKind(tok::unknown);
2271           MI->AddTokenToBody(LastTok);
2272           continue;
2273         } else {
2274           Diag(Tok, diag::err_pp_stringize_not_parameter)
2275             << LastTok.is(tok::hashat);
2276 
2277           // Disable __VA_ARGS__ again.
2278           Ident__VA_ARGS__->setIsPoisoned(true);
2279           return;
2280         }
2281       }
2282 
2283       // Things look ok, add the '#' and param name tokens to the macro.
2284       MI->AddTokenToBody(LastTok);
2285       MI->AddTokenToBody(Tok);
2286       LastTok = Tok;
2287 
2288       // Get the next token of the macro.
2289       LexUnexpandedToken(Tok);
2290     }
2291   }
2292 
2293   if (MacroShadowsKeyword &&
2294       !isConfigurationPattern(MacroNameTok, MI, getLangOpts())) {
2295     Diag(MacroNameTok, diag::warn_pp_macro_hides_keyword);
2296   }
2297 
2298   // Disable __VA_ARGS__ again.
2299   Ident__VA_ARGS__->setIsPoisoned(true);
2300 
2301   // Check that there is no paste (##) operator at the beginning or end of the
2302   // replacement list.
2303   unsigned NumTokens = MI->getNumTokens();
2304   if (NumTokens != 0) {
2305     if (MI->getReplacementToken(0).is(tok::hashhash)) {
2306       Diag(MI->getReplacementToken(0), diag::err_paste_at_start);
2307       return;
2308     }
2309     if (MI->getReplacementToken(NumTokens-1).is(tok::hashhash)) {
2310       Diag(MI->getReplacementToken(NumTokens-1), diag::err_paste_at_end);
2311       return;
2312     }
2313   }
2314 
2315   MI->setDefinitionEndLoc(LastTok.getLocation());
2316 
2317   // Finally, if this identifier already had a macro defined for it, verify that
2318   // the macro bodies are identical, and issue diagnostics if they are not.
2319   if (const MacroInfo *OtherMI=getMacroInfo(MacroNameTok.getIdentifierInfo())) {
2320     // In Objective-C, ignore attempts to directly redefine the builtin
2321     // definitions of the ownership qualifiers.  It's still possible to
2322     // #undef them.
2323     auto isObjCProtectedMacro = [](const IdentifierInfo *II) -> bool {
2324       return II->isStr("__strong") ||
2325              II->isStr("__weak") ||
2326              II->isStr("__unsafe_unretained") ||
2327              II->isStr("__autoreleasing");
2328     };
2329    if (getLangOpts().ObjC1 &&
2330         SourceMgr.getFileID(OtherMI->getDefinitionLoc())
2331           == getPredefinesFileID() &&
2332         isObjCProtectedMacro(MacroNameTok.getIdentifierInfo())) {
2333       // Warn if it changes the tokens.
2334       if ((!getDiagnostics().getSuppressSystemWarnings() ||
2335            !SourceMgr.isInSystemHeader(DefineTok.getLocation())) &&
2336           !MI->isIdenticalTo(*OtherMI, *this,
2337                              /*Syntactic=*/LangOpts.MicrosoftExt)) {
2338         Diag(MI->getDefinitionLoc(), diag::warn_pp_objc_macro_redef_ignored);
2339       }
2340       assert(!OtherMI->isWarnIfUnused());
2341       return;
2342     }
2343 
2344     // It is very common for system headers to have tons of macro redefinitions
2345     // and for warnings to be disabled in system headers.  If this is the case,
2346     // then don't bother calling MacroInfo::isIdenticalTo.
2347     if (!getDiagnostics().getSuppressSystemWarnings() ||
2348         !SourceMgr.isInSystemHeader(DefineTok.getLocation())) {
2349       if (!OtherMI->isUsed() && OtherMI->isWarnIfUnused())
2350         Diag(OtherMI->getDefinitionLoc(), diag::pp_macro_not_used);
2351 
2352       // Warn if defining "__LINE__" and other builtins, per C99 6.10.8/4 and
2353       // C++ [cpp.predefined]p4, but allow it as an extension.
2354       if (OtherMI->isBuiltinMacro())
2355         Diag(MacroNameTok, diag::ext_pp_redef_builtin_macro);
2356       // Macros must be identical.  This means all tokens and whitespace
2357       // separation must be the same.  C99 6.10.3p2.
2358       else if (!OtherMI->isAllowRedefinitionsWithoutWarning() &&
2359                !MI->isIdenticalTo(*OtherMI, *this, /*Syntactic=*/LangOpts.MicrosoftExt)) {
2360         Diag(MI->getDefinitionLoc(), diag::ext_pp_macro_redef)
2361           << MacroNameTok.getIdentifierInfo();
2362         Diag(OtherMI->getDefinitionLoc(), diag::note_previous_definition);
2363       }
2364     }
2365     if (OtherMI->isWarnIfUnused())
2366       WarnUnusedMacroLocs.erase(OtherMI->getDefinitionLoc());
2367   }
2368 
2369   DefMacroDirective *MD =
2370       appendDefMacroDirective(MacroNameTok.getIdentifierInfo(), MI);
2371 
2372   assert(!MI->isUsed());
2373   // If we need warning for not using the macro, add its location in the
2374   // warn-because-unused-macro set. If it gets used it will be removed from set.
2375   if (getSourceManager().isInMainFile(MI->getDefinitionLoc()) &&
2376       !Diags->isIgnored(diag::pp_macro_not_used, MI->getDefinitionLoc())) {
2377     MI->setIsWarnIfUnused(true);
2378     WarnUnusedMacroLocs.insert(MI->getDefinitionLoc());
2379   }
2380 
2381   // If the callbacks want to know, tell them about the macro definition.
2382   if (Callbacks)
2383     Callbacks->MacroDefined(MacroNameTok, MD);
2384 }
2385 
2386 /// HandleUndefDirective - Implements \#undef.
2387 ///
2388 void Preprocessor::HandleUndefDirective(Token &UndefTok) {
2389   ++NumUndefined;
2390 
2391   Token MacroNameTok;
2392   ReadMacroName(MacroNameTok, MU_Undef);
2393 
2394   // Error reading macro name?  If so, diagnostic already issued.
2395   if (MacroNameTok.is(tok::eod))
2396     return;
2397 
2398   // Check to see if this is the last token on the #undef line.
2399   CheckEndOfDirective("undef");
2400 
2401   // Okay, we have a valid identifier to undef.
2402   auto *II = MacroNameTok.getIdentifierInfo();
2403   auto MD = getMacroDefinition(II);
2404 
2405   // If the callbacks want to know, tell them about the macro #undef.
2406   // Note: no matter if the macro was defined or not.
2407   if (Callbacks)
2408     Callbacks->MacroUndefined(MacroNameTok, MD);
2409 
2410   // If the macro is not defined, this is a noop undef, just return.
2411   const MacroInfo *MI = MD.getMacroInfo();
2412   if (!MI)
2413     return;
2414 
2415   if (!MI->isUsed() && MI->isWarnIfUnused())
2416     Diag(MI->getDefinitionLoc(), diag::pp_macro_not_used);
2417 
2418   if (MI->isWarnIfUnused())
2419     WarnUnusedMacroLocs.erase(MI->getDefinitionLoc());
2420 
2421   appendMacroDirective(MacroNameTok.getIdentifierInfo(),
2422                        AllocateUndefMacroDirective(MacroNameTok.getLocation()));
2423 }
2424 
2425 //===----------------------------------------------------------------------===//
2426 // Preprocessor Conditional Directive Handling.
2427 //===----------------------------------------------------------------------===//
2428 
2429 /// HandleIfdefDirective - Implements the \#ifdef/\#ifndef directive.  isIfndef
2430 /// is true when this is a \#ifndef directive.  ReadAnyTokensBeforeDirective is
2431 /// true if any tokens have been returned or pp-directives activated before this
2432 /// \#ifndef has been lexed.
2433 ///
2434 void Preprocessor::HandleIfdefDirective(Token &Result, bool isIfndef,
2435                                         bool ReadAnyTokensBeforeDirective) {
2436   ++NumIf;
2437   Token DirectiveTok = Result;
2438 
2439   Token MacroNameTok;
2440   ReadMacroName(MacroNameTok);
2441 
2442   // Error reading macro name?  If so, diagnostic already issued.
2443   if (MacroNameTok.is(tok::eod)) {
2444     // Skip code until we get to #endif.  This helps with recovery by not
2445     // emitting an error when the #endif is reached.
2446     SkipExcludedConditionalBlock(DirectiveTok.getLocation(),
2447                                  /*Foundnonskip*/false, /*FoundElse*/false);
2448     return;
2449   }
2450 
2451   // Check to see if this is the last token on the #if[n]def line.
2452   CheckEndOfDirective(isIfndef ? "ifndef" : "ifdef");
2453 
2454   IdentifierInfo *MII = MacroNameTok.getIdentifierInfo();
2455   auto MD = getMacroDefinition(MII);
2456   MacroInfo *MI = MD.getMacroInfo();
2457 
2458   if (CurPPLexer->getConditionalStackDepth() == 0) {
2459     // If the start of a top-level #ifdef and if the macro is not defined,
2460     // inform MIOpt that this might be the start of a proper include guard.
2461     // Otherwise it is some other form of unknown conditional which we can't
2462     // handle.
2463     if (!ReadAnyTokensBeforeDirective && !MI) {
2464       assert(isIfndef && "#ifdef shouldn't reach here");
2465       CurPPLexer->MIOpt.EnterTopLevelIfndef(MII, MacroNameTok.getLocation());
2466     } else
2467       CurPPLexer->MIOpt.EnterTopLevelConditional();
2468   }
2469 
2470   // If there is a macro, process it.
2471   if (MI)  // Mark it used.
2472     markMacroAsUsed(MI);
2473 
2474   if (Callbacks) {
2475     if (isIfndef)
2476       Callbacks->Ifndef(DirectiveTok.getLocation(), MacroNameTok, MD);
2477     else
2478       Callbacks->Ifdef(DirectiveTok.getLocation(), MacroNameTok, MD);
2479   }
2480 
2481   // Should we include the stuff contained by this directive?
2482   if (!MI == isIfndef) {
2483     // Yes, remember that we are inside a conditional, then lex the next token.
2484     CurPPLexer->pushConditionalLevel(DirectiveTok.getLocation(),
2485                                      /*wasskip*/false, /*foundnonskip*/true,
2486                                      /*foundelse*/false);
2487   } else {
2488     // No, skip the contents of this block.
2489     SkipExcludedConditionalBlock(DirectiveTok.getLocation(),
2490                                  /*Foundnonskip*/false,
2491                                  /*FoundElse*/false);
2492   }
2493 }
2494 
2495 /// HandleIfDirective - Implements the \#if directive.
2496 ///
2497 void Preprocessor::HandleIfDirective(Token &IfToken,
2498                                      bool ReadAnyTokensBeforeDirective) {
2499   ++NumIf;
2500 
2501   // Parse and evaluate the conditional expression.
2502   IdentifierInfo *IfNDefMacro = nullptr;
2503   const SourceLocation ConditionalBegin = CurPPLexer->getSourceLocation();
2504   const bool ConditionalTrue = EvaluateDirectiveExpression(IfNDefMacro);
2505   const SourceLocation ConditionalEnd = CurPPLexer->getSourceLocation();
2506 
2507   // If this condition is equivalent to #ifndef X, and if this is the first
2508   // directive seen, handle it for the multiple-include optimization.
2509   if (CurPPLexer->getConditionalStackDepth() == 0) {
2510     if (!ReadAnyTokensBeforeDirective && IfNDefMacro && ConditionalTrue)
2511       // FIXME: Pass in the location of the macro name, not the 'if' token.
2512       CurPPLexer->MIOpt.EnterTopLevelIfndef(IfNDefMacro, IfToken.getLocation());
2513     else
2514       CurPPLexer->MIOpt.EnterTopLevelConditional();
2515   }
2516 
2517   if (Callbacks)
2518     Callbacks->If(IfToken.getLocation(),
2519                   SourceRange(ConditionalBegin, ConditionalEnd),
2520                   (ConditionalTrue ? PPCallbacks::CVK_True : PPCallbacks::CVK_False));
2521 
2522   // Should we include the stuff contained by this directive?
2523   if (ConditionalTrue) {
2524     // Yes, remember that we are inside a conditional, then lex the next token.
2525     CurPPLexer->pushConditionalLevel(IfToken.getLocation(), /*wasskip*/false,
2526                                    /*foundnonskip*/true, /*foundelse*/false);
2527   } else {
2528     // No, skip the contents of this block.
2529     SkipExcludedConditionalBlock(IfToken.getLocation(), /*Foundnonskip*/false,
2530                                  /*FoundElse*/false);
2531   }
2532 }
2533 
2534 /// HandleEndifDirective - Implements the \#endif directive.
2535 ///
2536 void Preprocessor::HandleEndifDirective(Token &EndifToken) {
2537   ++NumEndif;
2538 
2539   // Check that this is the whole directive.
2540   CheckEndOfDirective("endif");
2541 
2542   PPConditionalInfo CondInfo;
2543   if (CurPPLexer->popConditionalLevel(CondInfo)) {
2544     // No conditionals on the stack: this is an #endif without an #if.
2545     Diag(EndifToken, diag::err_pp_endif_without_if);
2546     return;
2547   }
2548 
2549   // If this the end of a top-level #endif, inform MIOpt.
2550   if (CurPPLexer->getConditionalStackDepth() == 0)
2551     CurPPLexer->MIOpt.ExitTopLevelConditional();
2552 
2553   assert(!CondInfo.WasSkipping && !CurPPLexer->LexingRawMode &&
2554          "This code should only be reachable in the non-skipping case!");
2555 
2556   if (Callbacks)
2557     Callbacks->Endif(EndifToken.getLocation(), CondInfo.IfLoc);
2558 }
2559 
2560 /// HandleElseDirective - Implements the \#else directive.
2561 ///
2562 void Preprocessor::HandleElseDirective(Token &Result) {
2563   ++NumElse;
2564 
2565   // #else directive in a non-skipping conditional... start skipping.
2566   CheckEndOfDirective("else");
2567 
2568   PPConditionalInfo CI;
2569   if (CurPPLexer->popConditionalLevel(CI)) {
2570     Diag(Result, diag::pp_err_else_without_if);
2571     return;
2572   }
2573 
2574   // If this is a top-level #else, inform the MIOpt.
2575   if (CurPPLexer->getConditionalStackDepth() == 0)
2576     CurPPLexer->MIOpt.EnterTopLevelConditional();
2577 
2578   // If this is a #else with a #else before it, report the error.
2579   if (CI.FoundElse) Diag(Result, diag::pp_err_else_after_else);
2580 
2581   if (Callbacks)
2582     Callbacks->Else(Result.getLocation(), CI.IfLoc);
2583 
2584   // Finally, skip the rest of the contents of this block.
2585   SkipExcludedConditionalBlock(CI.IfLoc, /*Foundnonskip*/true,
2586                                /*FoundElse*/true, Result.getLocation());
2587 }
2588 
2589 /// HandleElifDirective - Implements the \#elif directive.
2590 ///
2591 void Preprocessor::HandleElifDirective(Token &ElifToken) {
2592   ++NumElse;
2593 
2594   // #elif directive in a non-skipping conditional... start skipping.
2595   // We don't care what the condition is, because we will always skip it (since
2596   // the block immediately before it was included).
2597   const SourceLocation ConditionalBegin = CurPPLexer->getSourceLocation();
2598   DiscardUntilEndOfDirective();
2599   const SourceLocation ConditionalEnd = CurPPLexer->getSourceLocation();
2600 
2601   PPConditionalInfo CI;
2602   if (CurPPLexer->popConditionalLevel(CI)) {
2603     Diag(ElifToken, diag::pp_err_elif_without_if);
2604     return;
2605   }
2606 
2607   // If this is a top-level #elif, inform the MIOpt.
2608   if (CurPPLexer->getConditionalStackDepth() == 0)
2609     CurPPLexer->MIOpt.EnterTopLevelConditional();
2610 
2611   // If this is a #elif with a #else before it, report the error.
2612   if (CI.FoundElse) Diag(ElifToken, diag::pp_err_elif_after_else);
2613 
2614   if (Callbacks)
2615     Callbacks->Elif(ElifToken.getLocation(),
2616                     SourceRange(ConditionalBegin, ConditionalEnd),
2617                     PPCallbacks::CVK_NotEvaluated, CI.IfLoc);
2618 
2619   // Finally, skip the rest of the contents of this block.
2620   SkipExcludedConditionalBlock(CI.IfLoc, /*Foundnonskip*/true,
2621                                /*FoundElse*/CI.FoundElse,
2622                                ElifToken.getLocation());
2623 }
2624