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