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     StringRef Filename,
536     bool isAngled,
537     const DirectoryLookup *FromDir,
538     const DirectoryLookup *&CurDir,
539     SmallVectorImpl<char> *SearchPath,
540     SmallVectorImpl<char> *RelativePath,
541     Module **SuggestedModule,
542     bool SkipCache) {
543   // If the header lookup mechanism may be relative to the current file, pass in
544   // info about where the current file is.
545   const FileEntry *CurFileEnt = 0;
546   if (!FromDir) {
547     FileID FID = getCurrentFileLexer()->getFileID();
548     CurFileEnt = SourceMgr.getFileEntryForID(FID);
549 
550     // If there is no file entry associated with this file, it must be the
551     // predefines buffer.  Any other file is not lexed with a normal lexer, so
552     // it won't be scanned for preprocessor directives.   If we have the
553     // predefines buffer, resolve #include references (which come from the
554     // -include command line argument) as if they came from the main file, this
555     // affects file lookup etc.
556     if (CurFileEnt == 0) {
557       FID = SourceMgr.getMainFileID();
558       CurFileEnt = SourceMgr.getFileEntryForID(FID);
559     }
560   }
561 
562   // Do a standard file entry lookup.
563   CurDir = CurDirLookup;
564   const FileEntry *FE = HeaderInfo.LookupFile(
565       Filename, isAngled, FromDir, CurDir, CurFileEnt,
566       SearchPath, RelativePath, SuggestedModule, SkipCache);
567   if (FE) return FE;
568 
569   // Otherwise, see if this is a subframework header.  If so, this is relative
570   // to one of the headers on the #include stack.  Walk the list of the current
571   // headers on the #include stack and pass them to HeaderInfo.
572   if (IsFileLexer()) {
573     if ((CurFileEnt = SourceMgr.getFileEntryForID(CurPPLexer->getFileID())))
574       if ((FE = HeaderInfo.LookupSubframeworkHeader(Filename, CurFileEnt,
575                                                     SearchPath, RelativePath,
576                                                     SuggestedModule)))
577         return FE;
578   }
579 
580   for (unsigned i = 0, e = IncludeMacroStack.size(); i != e; ++i) {
581     IncludeStackInfo &ISEntry = IncludeMacroStack[e-i-1];
582     if (IsFileLexer(ISEntry)) {
583       if ((CurFileEnt =
584            SourceMgr.getFileEntryForID(ISEntry.ThePPLexer->getFileID())))
585         if ((FE = HeaderInfo.LookupSubframeworkHeader(
586                 Filename, CurFileEnt, SearchPath, RelativePath,
587                 SuggestedModule)))
588           return FE;
589     }
590   }
591 
592   // Otherwise, we really couldn't find the file.
593   return 0;
594 }
595 
596 
597 //===----------------------------------------------------------------------===//
598 // Preprocessor Directive Handling.
599 //===----------------------------------------------------------------------===//
600 
601 class Preprocessor::ResetMacroExpansionHelper {
602 public:
603   ResetMacroExpansionHelper(Preprocessor *pp)
604     : PP(pp), save(pp->DisableMacroExpansion) {
605     if (pp->MacroExpansionInDirectivesOverride)
606       pp->DisableMacroExpansion = false;
607   }
608   ~ResetMacroExpansionHelper() {
609     PP->DisableMacroExpansion = save;
610   }
611 private:
612   Preprocessor *PP;
613   bool save;
614 };
615 
616 /// HandleDirective - This callback is invoked when the lexer sees a # token
617 /// at the start of a line.  This consumes the directive, modifies the
618 /// lexer/preprocessor state, and advances the lexer(s) so that the next token
619 /// read is the correct one.
620 void Preprocessor::HandleDirective(Token &Result) {
621   // FIXME: Traditional: # with whitespace before it not recognized by K&R?
622 
623   // We just parsed a # character at the start of a line, so we're in directive
624   // mode.  Tell the lexer this so any newlines we see will be converted into an
625   // EOD token (which terminates the directive).
626   CurPPLexer->ParsingPreprocessorDirective = true;
627   if (CurLexer) CurLexer->SetKeepWhitespaceMode(false);
628 
629   bool ImmediatelyAfterTopLevelIfndef =
630       CurPPLexer->MIOpt.getImmediatelyAfterTopLevelIfndef();
631   CurPPLexer->MIOpt.resetImmediatelyAfterTopLevelIfndef();
632 
633   ++NumDirectives;
634 
635   // We are about to read a token.  For the multiple-include optimization FA to
636   // work, we have to remember if we had read any tokens *before* this
637   // pp-directive.
638   bool ReadAnyTokensBeforeDirective =CurPPLexer->MIOpt.getHasReadAnyTokensVal();
639 
640   // Save the '#' token in case we need to return it later.
641   Token SavedHash = Result;
642 
643   // Read the next token, the directive flavor.  This isn't expanded due to
644   // C99 6.10.3p8.
645   LexUnexpandedToken(Result);
646 
647   // C99 6.10.3p11: Is this preprocessor directive in macro invocation?  e.g.:
648   //   #define A(x) #x
649   //   A(abc
650   //     #warning blah
651   //   def)
652   // If so, the user is relying on undefined behavior, emit a diagnostic. Do
653   // not support this for #include-like directives, since that can result in
654   // terrible diagnostics, and does not work in GCC.
655   if (InMacroArgs) {
656     if (IdentifierInfo *II = Result.getIdentifierInfo()) {
657       switch (II->getPPKeywordID()) {
658       case tok::pp_include:
659       case tok::pp_import:
660       case tok::pp_include_next:
661       case tok::pp___include_macros:
662         Diag(Result, diag::err_embedded_include) << II->getName();
663         DiscardUntilEndOfDirective();
664         return;
665       default:
666         break;
667       }
668     }
669     Diag(Result, diag::ext_embedded_directive);
670   }
671 
672   // Temporarily enable macro expansion if set so
673   // and reset to previous state when returning from this function.
674   ResetMacroExpansionHelper helper(this);
675 
676   switch (Result.getKind()) {
677   case tok::eod:
678     return;   // null directive.
679   case tok::code_completion:
680     if (CodeComplete)
681       CodeComplete->CodeCompleteDirective(
682                                     CurPPLexer->getConditionalStackDepth() > 0);
683     setCodeCompletionReached();
684     return;
685   case tok::numeric_constant:  // # 7  GNU line marker directive.
686     if (getLangOpts().AsmPreprocessor)
687       break;  // # 4 is not a preprocessor directive in .S files.
688     return HandleDigitDirective(Result);
689   default:
690     IdentifierInfo *II = Result.getIdentifierInfo();
691     if (II == 0) break;  // Not an identifier.
692 
693     // Ask what the preprocessor keyword ID is.
694     switch (II->getPPKeywordID()) {
695     default: break;
696     // C99 6.10.1 - Conditional Inclusion.
697     case tok::pp_if:
698       return HandleIfDirective(Result, ReadAnyTokensBeforeDirective);
699     case tok::pp_ifdef:
700       return HandleIfdefDirective(Result, false, true/*not valid for miopt*/);
701     case tok::pp_ifndef:
702       return HandleIfdefDirective(Result, true, ReadAnyTokensBeforeDirective);
703     case tok::pp_elif:
704       return HandleElifDirective(Result);
705     case tok::pp_else:
706       return HandleElseDirective(Result);
707     case tok::pp_endif:
708       return HandleEndifDirective(Result);
709 
710     // C99 6.10.2 - Source File Inclusion.
711     case tok::pp_include:
712       // Handle #include.
713       return HandleIncludeDirective(SavedHash.getLocation(), Result);
714     case tok::pp___include_macros:
715       // Handle -imacros.
716       return HandleIncludeMacrosDirective(SavedHash.getLocation(), Result);
717 
718     // C99 6.10.3 - Macro Replacement.
719     case tok::pp_define:
720       return HandleDefineDirective(Result, ImmediatelyAfterTopLevelIfndef);
721     case tok::pp_undef:
722       return HandleUndefDirective(Result);
723 
724     // C99 6.10.4 - Line Control.
725     case tok::pp_line:
726       return HandleLineDirective(Result);
727 
728     // C99 6.10.5 - Error Directive.
729     case tok::pp_error:
730       return HandleUserDiagnosticDirective(Result, false);
731 
732     // C99 6.10.6 - Pragma Directive.
733     case tok::pp_pragma:
734       return HandlePragmaDirective(PIK_HashPragma);
735 
736     // GNU Extensions.
737     case tok::pp_import:
738       return HandleImportDirective(SavedHash.getLocation(), Result);
739     case tok::pp_include_next:
740       return HandleIncludeNextDirective(SavedHash.getLocation(), Result);
741 
742     case tok::pp_warning:
743       Diag(Result, diag::ext_pp_warning_directive);
744       return HandleUserDiagnosticDirective(Result, true);
745     case tok::pp_ident:
746       return HandleIdentSCCSDirective(Result);
747     case tok::pp_sccs:
748       return HandleIdentSCCSDirective(Result);
749     case tok::pp_assert:
750       //isExtension = true;  // FIXME: implement #assert
751       break;
752     case tok::pp_unassert:
753       //isExtension = true;  // FIXME: implement #unassert
754       break;
755 
756     case tok::pp___public_macro:
757       if (getLangOpts().Modules)
758         return HandleMacroPublicDirective(Result);
759       break;
760 
761     case tok::pp___private_macro:
762       if (getLangOpts().Modules)
763         return HandleMacroPrivateDirective(Result);
764       break;
765     }
766     break;
767   }
768 
769   // If this is a .S file, treat unknown # directives as non-preprocessor
770   // directives.  This is important because # may be a comment or introduce
771   // various pseudo-ops.  Just return the # token and push back the following
772   // token to be lexed next time.
773   if (getLangOpts().AsmPreprocessor) {
774     Token *Toks = new Token[2];
775     // Return the # and the token after it.
776     Toks[0] = SavedHash;
777     Toks[1] = Result;
778 
779     // If the second token is a hashhash token, then we need to translate it to
780     // unknown so the token lexer doesn't try to perform token pasting.
781     if (Result.is(tok::hashhash))
782       Toks[1].setKind(tok::unknown);
783 
784     // Enter this token stream so that we re-lex the tokens.  Make sure to
785     // enable macro expansion, in case the token after the # is an identifier
786     // that is expanded.
787     EnterTokenStream(Toks, 2, false, true);
788     return;
789   }
790 
791   // If we reached here, the preprocessing token is not valid!
792   Diag(Result, diag::err_pp_invalid_directive);
793 
794   // Read the rest of the PP line.
795   DiscardUntilEndOfDirective();
796 
797   // Okay, we're done parsing the directive.
798 }
799 
800 /// GetLineValue - Convert a numeric token into an unsigned value, emitting
801 /// Diagnostic DiagID if it is invalid, and returning the value in Val.
802 static bool GetLineValue(Token &DigitTok, unsigned &Val,
803                          unsigned DiagID, Preprocessor &PP,
804                          bool IsGNULineDirective=false) {
805   if (DigitTok.isNot(tok::numeric_constant)) {
806     PP.Diag(DigitTok, DiagID);
807 
808     if (DigitTok.isNot(tok::eod))
809       PP.DiscardUntilEndOfDirective();
810     return true;
811   }
812 
813   SmallString<64> IntegerBuffer;
814   IntegerBuffer.resize(DigitTok.getLength());
815   const char *DigitTokBegin = &IntegerBuffer[0];
816   bool Invalid = false;
817   unsigned ActualLength = PP.getSpelling(DigitTok, DigitTokBegin, &Invalid);
818   if (Invalid)
819     return true;
820 
821   // Verify that we have a simple digit-sequence, and compute the value.  This
822   // is always a simple digit string computed in decimal, so we do this manually
823   // here.
824   Val = 0;
825   for (unsigned i = 0; i != ActualLength; ++i) {
826     if (!isDigit(DigitTokBegin[i])) {
827       PP.Diag(PP.AdvanceToTokenCharacter(DigitTok.getLocation(), i),
828               diag::err_pp_line_digit_sequence) << IsGNULineDirective;
829       PP.DiscardUntilEndOfDirective();
830       return true;
831     }
832 
833     unsigned NextVal = Val*10+(DigitTokBegin[i]-'0');
834     if (NextVal < Val) { // overflow.
835       PP.Diag(DigitTok, DiagID);
836       PP.DiscardUntilEndOfDirective();
837       return true;
838     }
839     Val = NextVal;
840   }
841 
842   if (DigitTokBegin[0] == '0' && Val)
843     PP.Diag(DigitTok.getLocation(), diag::warn_pp_line_decimal)
844       << IsGNULineDirective;
845 
846   return false;
847 }
848 
849 /// \brief Handle a \#line directive: C99 6.10.4.
850 ///
851 /// The two acceptable forms are:
852 /// \verbatim
853 ///   # line digit-sequence
854 ///   # line digit-sequence "s-char-sequence"
855 /// \endverbatim
856 void Preprocessor::HandleLineDirective(Token &Tok) {
857   // Read the line # and string argument.  Per C99 6.10.4p5, these tokens are
858   // expanded.
859   Token DigitTok;
860   Lex(DigitTok);
861 
862   // Validate the number and convert it to an unsigned.
863   unsigned LineNo;
864   if (GetLineValue(DigitTok, LineNo, diag::err_pp_line_requires_integer,*this))
865     return;
866 
867   if (LineNo == 0)
868     Diag(DigitTok, diag::ext_pp_line_zero);
869 
870   // Enforce C99 6.10.4p3: "The digit sequence shall not specify ... a
871   // number greater than 2147483647".  C90 requires that the line # be <= 32767.
872   unsigned LineLimit = 32768U;
873   if (LangOpts.C99 || LangOpts.CPlusPlus11)
874     LineLimit = 2147483648U;
875   if (LineNo >= LineLimit)
876     Diag(DigitTok, diag::ext_pp_line_too_big) << LineLimit;
877   else if (LangOpts.CPlusPlus11 && LineNo >= 32768U)
878     Diag(DigitTok, diag::warn_cxx98_compat_pp_line_too_big);
879 
880   int FilenameID = -1;
881   Token StrTok;
882   Lex(StrTok);
883 
884   // If the StrTok is "eod", then it wasn't present.  Otherwise, it must be a
885   // string followed by eod.
886   if (StrTok.is(tok::eod))
887     ; // ok
888   else if (StrTok.isNot(tok::string_literal)) {
889     Diag(StrTok, diag::err_pp_line_invalid_filename);
890     return DiscardUntilEndOfDirective();
891   } else if (StrTok.hasUDSuffix()) {
892     Diag(StrTok, diag::err_invalid_string_udl);
893     return DiscardUntilEndOfDirective();
894   } else {
895     // Parse and validate the string, converting it into a unique ID.
896     StringLiteralParser Literal(&StrTok, 1, *this);
897     assert(Literal.isAscii() && "Didn't allow wide strings in");
898     if (Literal.hadError)
899       return DiscardUntilEndOfDirective();
900     if (Literal.Pascal) {
901       Diag(StrTok, diag::err_pp_linemarker_invalid_filename);
902       return DiscardUntilEndOfDirective();
903     }
904     FilenameID = SourceMgr.getLineTableFilenameID(Literal.GetString());
905 
906     // Verify that there is nothing after the string, other than EOD.  Because
907     // of C99 6.10.4p5, macros that expand to empty tokens are ok.
908     CheckEndOfDirective("line", true);
909   }
910 
911   SourceMgr.AddLineNote(DigitTok.getLocation(), LineNo, FilenameID);
912 
913   if (Callbacks)
914     Callbacks->FileChanged(CurPPLexer->getSourceLocation(),
915                            PPCallbacks::RenameFile,
916                            SrcMgr::C_User);
917 }
918 
919 /// ReadLineMarkerFlags - Parse and validate any flags at the end of a GNU line
920 /// marker directive.
921 static bool ReadLineMarkerFlags(bool &IsFileEntry, bool &IsFileExit,
922                                 bool &IsSystemHeader, bool &IsExternCHeader,
923                                 Preprocessor &PP) {
924   unsigned FlagVal;
925   Token FlagTok;
926   PP.Lex(FlagTok);
927   if (FlagTok.is(tok::eod)) return false;
928   if (GetLineValue(FlagTok, FlagVal, diag::err_pp_linemarker_invalid_flag, PP))
929     return true;
930 
931   if (FlagVal == 1) {
932     IsFileEntry = true;
933 
934     PP.Lex(FlagTok);
935     if (FlagTok.is(tok::eod)) return false;
936     if (GetLineValue(FlagTok, FlagVal, diag::err_pp_linemarker_invalid_flag,PP))
937       return true;
938   } else if (FlagVal == 2) {
939     IsFileExit = true;
940 
941     SourceManager &SM = PP.getSourceManager();
942     // If we are leaving the current presumed file, check to make sure the
943     // presumed include stack isn't empty!
944     FileID CurFileID =
945       SM.getDecomposedExpansionLoc(FlagTok.getLocation()).first;
946     PresumedLoc PLoc = SM.getPresumedLoc(FlagTok.getLocation());
947     if (PLoc.isInvalid())
948       return true;
949 
950     // If there is no include loc (main file) or if the include loc is in a
951     // different physical file, then we aren't in a "1" line marker flag region.
952     SourceLocation IncLoc = PLoc.getIncludeLoc();
953     if (IncLoc.isInvalid() ||
954         SM.getDecomposedExpansionLoc(IncLoc).first != CurFileID) {
955       PP.Diag(FlagTok, diag::err_pp_linemarker_invalid_pop);
956       PP.DiscardUntilEndOfDirective();
957       return true;
958     }
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   }
965 
966   // We must have 3 if there are still flags.
967   if (FlagVal != 3) {
968     PP.Diag(FlagTok, diag::err_pp_linemarker_invalid_flag);
969     PP.DiscardUntilEndOfDirective();
970     return true;
971   }
972 
973   IsSystemHeader = true;
974 
975   PP.Lex(FlagTok);
976   if (FlagTok.is(tok::eod)) return false;
977   if (GetLineValue(FlagTok, FlagVal, diag::err_pp_linemarker_invalid_flag, PP))
978     return true;
979 
980   // We must have 4 if there is yet another flag.
981   if (FlagVal != 4) {
982     PP.Diag(FlagTok, diag::err_pp_linemarker_invalid_flag);
983     PP.DiscardUntilEndOfDirective();
984     return true;
985   }
986 
987   IsExternCHeader = true;
988 
989   PP.Lex(FlagTok);
990   if (FlagTok.is(tok::eod)) return false;
991 
992   // There are no more valid flags here.
993   PP.Diag(FlagTok, diag::err_pp_linemarker_invalid_flag);
994   PP.DiscardUntilEndOfDirective();
995   return true;
996 }
997 
998 /// HandleDigitDirective - Handle a GNU line marker directive, whose syntax is
999 /// one of the following forms:
1000 ///
1001 ///     # 42
1002 ///     # 42 "file" ('1' | '2')?
1003 ///     # 42 "file" ('1' | '2')? '3' '4'?
1004 ///
1005 void Preprocessor::HandleDigitDirective(Token &DigitTok) {
1006   // Validate the number and convert it to an unsigned.  GNU does not have a
1007   // line # limit other than it fit in 32-bits.
1008   unsigned LineNo;
1009   if (GetLineValue(DigitTok, LineNo, diag::err_pp_linemarker_requires_integer,
1010                    *this, true))
1011     return;
1012 
1013   Token StrTok;
1014   Lex(StrTok);
1015 
1016   bool IsFileEntry = false, IsFileExit = false;
1017   bool IsSystemHeader = false, IsExternCHeader = false;
1018   int FilenameID = -1;
1019 
1020   // If the StrTok is "eod", then it wasn't present.  Otherwise, it must be a
1021   // string followed by eod.
1022   if (StrTok.is(tok::eod))
1023     ; // ok
1024   else if (StrTok.isNot(tok::string_literal)) {
1025     Diag(StrTok, diag::err_pp_linemarker_invalid_filename);
1026     return DiscardUntilEndOfDirective();
1027   } else if (StrTok.hasUDSuffix()) {
1028     Diag(StrTok, diag::err_invalid_string_udl);
1029     return DiscardUntilEndOfDirective();
1030   } else {
1031     // Parse and validate the string, converting it into a unique ID.
1032     StringLiteralParser Literal(&StrTok, 1, *this);
1033     assert(Literal.isAscii() && "Didn't allow wide strings in");
1034     if (Literal.hadError)
1035       return DiscardUntilEndOfDirective();
1036     if (Literal.Pascal) {
1037       Diag(StrTok, diag::err_pp_linemarker_invalid_filename);
1038       return DiscardUntilEndOfDirective();
1039     }
1040     FilenameID = SourceMgr.getLineTableFilenameID(Literal.GetString());
1041 
1042     // If a filename was present, read any flags that are present.
1043     if (ReadLineMarkerFlags(IsFileEntry, IsFileExit,
1044                             IsSystemHeader, IsExternCHeader, *this))
1045       return;
1046   }
1047 
1048   // Create a line note with this information.
1049   SourceMgr.AddLineNote(DigitTok.getLocation(), LineNo, FilenameID,
1050                         IsFileEntry, IsFileExit,
1051                         IsSystemHeader, IsExternCHeader);
1052 
1053   // If the preprocessor has callbacks installed, notify them of the #line
1054   // change.  This is used so that the line marker comes out in -E mode for
1055   // example.
1056   if (Callbacks) {
1057     PPCallbacks::FileChangeReason Reason = PPCallbacks::RenameFile;
1058     if (IsFileEntry)
1059       Reason = PPCallbacks::EnterFile;
1060     else if (IsFileExit)
1061       Reason = PPCallbacks::ExitFile;
1062     SrcMgr::CharacteristicKind FileKind = SrcMgr::C_User;
1063     if (IsExternCHeader)
1064       FileKind = SrcMgr::C_ExternCSystem;
1065     else if (IsSystemHeader)
1066       FileKind = SrcMgr::C_System;
1067 
1068     Callbacks->FileChanged(CurPPLexer->getSourceLocation(), Reason, FileKind);
1069   }
1070 }
1071 
1072 
1073 /// HandleUserDiagnosticDirective - Handle a #warning or #error directive.
1074 ///
1075 void Preprocessor::HandleUserDiagnosticDirective(Token &Tok,
1076                                                  bool isWarning) {
1077   // PTH doesn't emit #warning or #error directives.
1078   if (CurPTHLexer)
1079     return CurPTHLexer->DiscardToEndOfLine();
1080 
1081   // Read the rest of the line raw.  We do this because we don't want macros
1082   // to be expanded and we don't require that the tokens be valid preprocessing
1083   // tokens.  For example, this is allowed: "#warning `   'foo".  GCC does
1084   // collapse multiple consequtive white space between tokens, but this isn't
1085   // specified by the standard.
1086   SmallString<128> Message;
1087   CurLexer->ReadToEndOfLine(&Message);
1088 
1089   // Find the first non-whitespace character, so that we can make the
1090   // diagnostic more succinct.
1091   StringRef Msg = Message.str().ltrim(" ");
1092 
1093   if (isWarning)
1094     Diag(Tok, diag::pp_hash_warning) << Msg;
1095   else
1096     Diag(Tok, diag::err_pp_hash_error) << Msg;
1097 }
1098 
1099 /// HandleIdentSCCSDirective - Handle a #ident/#sccs directive.
1100 ///
1101 void Preprocessor::HandleIdentSCCSDirective(Token &Tok) {
1102   // Yes, this directive is an extension.
1103   Diag(Tok, diag::ext_pp_ident_directive);
1104 
1105   // Read the string argument.
1106   Token StrTok;
1107   Lex(StrTok);
1108 
1109   // If the token kind isn't a string, it's a malformed directive.
1110   if (StrTok.isNot(tok::string_literal) &&
1111       StrTok.isNot(tok::wide_string_literal)) {
1112     Diag(StrTok, diag::err_pp_malformed_ident);
1113     if (StrTok.isNot(tok::eod))
1114       DiscardUntilEndOfDirective();
1115     return;
1116   }
1117 
1118   if (StrTok.hasUDSuffix()) {
1119     Diag(StrTok, diag::err_invalid_string_udl);
1120     return DiscardUntilEndOfDirective();
1121   }
1122 
1123   // Verify that there is nothing after the string, other than EOD.
1124   CheckEndOfDirective("ident");
1125 
1126   if (Callbacks) {
1127     bool Invalid = false;
1128     std::string Str = getSpelling(StrTok, &Invalid);
1129     if (!Invalid)
1130       Callbacks->Ident(Tok.getLocation(), Str);
1131   }
1132 }
1133 
1134 /// \brief Handle a #public directive.
1135 void Preprocessor::HandleMacroPublicDirective(Token &Tok) {
1136   Token MacroNameTok;
1137   ReadMacroName(MacroNameTok, 2);
1138 
1139   // Error reading macro name?  If so, diagnostic already issued.
1140   if (MacroNameTok.is(tok::eod))
1141     return;
1142 
1143   // Check to see if this is the last token on the #__public_macro line.
1144   CheckEndOfDirective("__public_macro");
1145 
1146   IdentifierInfo *II = MacroNameTok.getIdentifierInfo();
1147   // Okay, we finally have a valid identifier to undef.
1148   MacroDirective *MD = getMacroDirective(II);
1149 
1150   // If the macro is not defined, this is an error.
1151   if (MD == 0) {
1152     Diag(MacroNameTok, diag::err_pp_visibility_non_macro) << II;
1153     return;
1154   }
1155 
1156   // Note that this macro has now been exported.
1157   appendMacroDirective(II, AllocateVisibilityMacroDirective(
1158                                 MacroNameTok.getLocation(), /*IsPublic=*/true));
1159 }
1160 
1161 /// \brief Handle a #private directive.
1162 void Preprocessor::HandleMacroPrivateDirective(Token &Tok) {
1163   Token MacroNameTok;
1164   ReadMacroName(MacroNameTok, 2);
1165 
1166   // Error reading macro name?  If so, diagnostic already issued.
1167   if (MacroNameTok.is(tok::eod))
1168     return;
1169 
1170   // Check to see if this is the last token on the #__private_macro line.
1171   CheckEndOfDirective("__private_macro");
1172 
1173   IdentifierInfo *II = MacroNameTok.getIdentifierInfo();
1174   // Okay, we finally have a valid identifier to undef.
1175   MacroDirective *MD = getMacroDirective(II);
1176 
1177   // If the macro is not defined, this is an error.
1178   if (MD == 0) {
1179     Diag(MacroNameTok, diag::err_pp_visibility_non_macro) << II;
1180     return;
1181   }
1182 
1183   // Note that this macro has now been marked private.
1184   appendMacroDirective(II, AllocateVisibilityMacroDirective(
1185                                MacroNameTok.getLocation(), /*IsPublic=*/false));
1186 }
1187 
1188 //===----------------------------------------------------------------------===//
1189 // Preprocessor Include Directive Handling.
1190 //===----------------------------------------------------------------------===//
1191 
1192 /// GetIncludeFilenameSpelling - Turn the specified lexer token into a fully
1193 /// checked and spelled filename, e.g. as an operand of \#include. This returns
1194 /// true if the input filename was in <>'s or false if it were in ""'s.  The
1195 /// caller is expected to provide a buffer that is large enough to hold the
1196 /// spelling of the filename, but is also expected to handle the case when
1197 /// this method decides to use a different buffer.
1198 bool Preprocessor::GetIncludeFilenameSpelling(SourceLocation Loc,
1199                                               StringRef &Buffer) {
1200   // Get the text form of the filename.
1201   assert(!Buffer.empty() && "Can't have tokens with empty spellings!");
1202 
1203   // Make sure the filename is <x> or "x".
1204   bool isAngled;
1205   if (Buffer[0] == '<') {
1206     if (Buffer.back() != '>') {
1207       Diag(Loc, diag::err_pp_expects_filename);
1208       Buffer = StringRef();
1209       return true;
1210     }
1211     isAngled = true;
1212   } else if (Buffer[0] == '"') {
1213     if (Buffer.back() != '"') {
1214       Diag(Loc, diag::err_pp_expects_filename);
1215       Buffer = StringRef();
1216       return true;
1217     }
1218     isAngled = false;
1219   } else {
1220     Diag(Loc, diag::err_pp_expects_filename);
1221     Buffer = StringRef();
1222     return true;
1223   }
1224 
1225   // Diagnose #include "" as invalid.
1226   if (Buffer.size() <= 2) {
1227     Diag(Loc, diag::err_pp_empty_filename);
1228     Buffer = StringRef();
1229     return true;
1230   }
1231 
1232   // Skip the brackets.
1233   Buffer = Buffer.substr(1, Buffer.size()-2);
1234   return isAngled;
1235 }
1236 
1237 /// \brief Handle cases where the \#include name is expanded from a macro
1238 /// as multiple tokens, which need to be glued together.
1239 ///
1240 /// This occurs for code like:
1241 /// \code
1242 ///    \#define FOO <a/b.h>
1243 ///    \#include FOO
1244 /// \endcode
1245 /// because in this case, "<a/b.h>" is returned as 7 tokens, not one.
1246 ///
1247 /// This code concatenates and consumes tokens up to the '>' token.  It returns
1248 /// false if the > was found, otherwise it returns true if it finds and consumes
1249 /// the EOD marker.
1250 bool Preprocessor::ConcatenateIncludeName(
1251                                         SmallString<128> &FilenameBuffer,
1252                                           SourceLocation &End) {
1253   Token CurTok;
1254 
1255   Lex(CurTok);
1256   while (CurTok.isNot(tok::eod)) {
1257     End = CurTok.getLocation();
1258 
1259     // FIXME: Provide code completion for #includes.
1260     if (CurTok.is(tok::code_completion)) {
1261       setCodeCompletionReached();
1262       Lex(CurTok);
1263       continue;
1264     }
1265 
1266     // Append the spelling of this token to the buffer. If there was a space
1267     // before it, add it now.
1268     if (CurTok.hasLeadingSpace())
1269       FilenameBuffer.push_back(' ');
1270 
1271     // Get the spelling of the token, directly into FilenameBuffer if possible.
1272     unsigned PreAppendSize = FilenameBuffer.size();
1273     FilenameBuffer.resize(PreAppendSize+CurTok.getLength());
1274 
1275     const char *BufPtr = &FilenameBuffer[PreAppendSize];
1276     unsigned ActualLen = getSpelling(CurTok, BufPtr);
1277 
1278     // If the token was spelled somewhere else, copy it into FilenameBuffer.
1279     if (BufPtr != &FilenameBuffer[PreAppendSize])
1280       memcpy(&FilenameBuffer[PreAppendSize], BufPtr, ActualLen);
1281 
1282     // Resize FilenameBuffer to the correct size.
1283     if (CurTok.getLength() != ActualLen)
1284       FilenameBuffer.resize(PreAppendSize+ActualLen);
1285 
1286     // If we found the '>' marker, return success.
1287     if (CurTok.is(tok::greater))
1288       return false;
1289 
1290     Lex(CurTok);
1291   }
1292 
1293   // If we hit the eod marker, emit an error and return true so that the caller
1294   // knows the EOD has been read.
1295   Diag(CurTok.getLocation(), diag::err_pp_expects_filename);
1296   return true;
1297 }
1298 
1299 /// HandleIncludeDirective - The "\#include" tokens have just been read, read
1300 /// the file to be included from the lexer, then include it!  This is a common
1301 /// routine with functionality shared between \#include, \#include_next and
1302 /// \#import.  LookupFrom is set when this is a \#include_next directive, it
1303 /// specifies the file to start searching from.
1304 void Preprocessor::HandleIncludeDirective(SourceLocation HashLoc,
1305                                           Token &IncludeTok,
1306                                           const DirectoryLookup *LookupFrom,
1307                                           bool isImport) {
1308 
1309   Token FilenameTok;
1310   CurPPLexer->LexIncludeFilename(FilenameTok);
1311 
1312   // Reserve a buffer to get the spelling.
1313   SmallString<128> FilenameBuffer;
1314   StringRef Filename;
1315   SourceLocation End;
1316   SourceLocation CharEnd; // the end of this directive, in characters
1317 
1318   switch (FilenameTok.getKind()) {
1319   case tok::eod:
1320     // If the token kind is EOD, the error has already been diagnosed.
1321     return;
1322 
1323   case tok::angle_string_literal:
1324   case tok::string_literal:
1325     Filename = getSpelling(FilenameTok, FilenameBuffer);
1326     End = FilenameTok.getLocation();
1327     CharEnd = End.getLocWithOffset(FilenameTok.getLength());
1328     break;
1329 
1330   case tok::less:
1331     // This could be a <foo/bar.h> file coming from a macro expansion.  In this
1332     // case, glue the tokens together into FilenameBuffer and interpret those.
1333     FilenameBuffer.push_back('<');
1334     if (ConcatenateIncludeName(FilenameBuffer, End))
1335       return;   // Found <eod> but no ">"?  Diagnostic already emitted.
1336     Filename = FilenameBuffer.str();
1337     CharEnd = End.getLocWithOffset(1);
1338     break;
1339   default:
1340     Diag(FilenameTok.getLocation(), diag::err_pp_expects_filename);
1341     DiscardUntilEndOfDirective();
1342     return;
1343   }
1344 
1345   CharSourceRange FilenameRange
1346     = CharSourceRange::getCharRange(FilenameTok.getLocation(), CharEnd);
1347   StringRef OriginalFilename = Filename;
1348   bool isAngled =
1349     GetIncludeFilenameSpelling(FilenameTok.getLocation(), Filename);
1350   // If GetIncludeFilenameSpelling set the start ptr to null, there was an
1351   // error.
1352   if (Filename.empty()) {
1353     DiscardUntilEndOfDirective();
1354     return;
1355   }
1356 
1357   // Verify that there is nothing after the filename, other than EOD.  Note that
1358   // we allow macros that expand to nothing after the filename, because this
1359   // falls into the category of "#include pp-tokens new-line" specified in
1360   // C99 6.10.2p4.
1361   CheckEndOfDirective(IncludeTok.getIdentifierInfo()->getNameStart(), true);
1362 
1363   // Check that we don't have infinite #include recursion.
1364   if (IncludeMacroStack.size() == MaxAllowedIncludeStackDepth-1) {
1365     Diag(FilenameTok, diag::err_pp_include_too_deep);
1366     return;
1367   }
1368 
1369   // Complain about attempts to #include files in an audit pragma.
1370   if (PragmaARCCFCodeAuditedLoc.isValid()) {
1371     Diag(HashLoc, diag::err_pp_include_in_arc_cf_code_audited);
1372     Diag(PragmaARCCFCodeAuditedLoc, diag::note_pragma_entered_here);
1373 
1374     // Immediately leave the pragma.
1375     PragmaARCCFCodeAuditedLoc = SourceLocation();
1376   }
1377 
1378   if (HeaderInfo.HasIncludeAliasMap()) {
1379     // Map the filename with the brackets still attached.  If the name doesn't
1380     // map to anything, fall back on the filename we've already gotten the
1381     // spelling for.
1382     StringRef NewName = HeaderInfo.MapHeaderToIncludeAlias(OriginalFilename);
1383     if (!NewName.empty())
1384       Filename = NewName;
1385   }
1386 
1387   // Search include directories.
1388   const DirectoryLookup *CurDir;
1389   SmallString<1024> SearchPath;
1390   SmallString<1024> RelativePath;
1391   // We get the raw path only if we have 'Callbacks' to which we later pass
1392   // the path.
1393   Module *SuggestedModule = 0;
1394   const FileEntry *File = LookupFile(
1395       Filename, isAngled, LookupFrom, CurDir,
1396       Callbacks ? &SearchPath : NULL, Callbacks ? &RelativePath : NULL,
1397       getLangOpts().Modules? &SuggestedModule : 0);
1398 
1399   if (Callbacks) {
1400     if (!File) {
1401       // Give the clients a chance to recover.
1402       SmallString<128> RecoveryPath;
1403       if (Callbacks->FileNotFound(Filename, RecoveryPath)) {
1404         if (const DirectoryEntry *DE = FileMgr.getDirectory(RecoveryPath)) {
1405           // Add the recovery path to the list of search paths.
1406           DirectoryLookup DL(DE, SrcMgr::C_User, false);
1407           HeaderInfo.AddSearchPath(DL, isAngled);
1408 
1409           // Try the lookup again, skipping the cache.
1410           File = LookupFile(Filename, isAngled, LookupFrom, CurDir, 0, 0,
1411                             getLangOpts().Modules? &SuggestedModule : 0,
1412                             /*SkipCache*/true);
1413         }
1414       }
1415     }
1416 
1417     if (!SuggestedModule) {
1418       // Notify the callback object that we've seen an inclusion directive.
1419       Callbacks->InclusionDirective(HashLoc, IncludeTok, Filename, isAngled,
1420                                     FilenameRange, File,
1421                                     SearchPath, RelativePath,
1422                                     /*ImportedModule=*/0);
1423     }
1424   }
1425 
1426   if (File == 0) {
1427     if (!SuppressIncludeNotFoundError) {
1428       // If the file could not be located and it was included via angle
1429       // brackets, we can attempt a lookup as though it were a quoted path to
1430       // provide the user with a possible fixit.
1431       if (isAngled) {
1432         File = LookupFile(Filename, false, LookupFrom, CurDir,
1433                           Callbacks ? &SearchPath : 0,
1434                           Callbacks ? &RelativePath : 0,
1435                           getLangOpts().Modules ? &SuggestedModule : 0);
1436         if (File) {
1437           SourceRange Range(FilenameTok.getLocation(), CharEnd);
1438           Diag(FilenameTok, diag::err_pp_file_not_found_not_fatal) <<
1439             Filename <<
1440             FixItHint::CreateReplacement(Range, "\"" + Filename.str() + "\"");
1441         }
1442       }
1443       // If the file is still not found, just go with the vanilla diagnostic
1444       if (!File)
1445         Diag(FilenameTok, diag::err_pp_file_not_found) << Filename;
1446     }
1447     if (!File)
1448       return;
1449   }
1450 
1451   // If we are supposed to import a module rather than including the header,
1452   // do so now.
1453   if (SuggestedModule) {
1454     // Compute the module access path corresponding to this module.
1455     // FIXME: Should we have a second loadModule() overload to avoid this
1456     // extra lookup step?
1457     SmallVector<std::pair<IdentifierInfo *, SourceLocation>, 2> Path;
1458     for (Module *Mod = SuggestedModule; Mod; Mod = Mod->Parent)
1459       Path.push_back(std::make_pair(getIdentifierInfo(Mod->Name),
1460                                     FilenameTok.getLocation()));
1461     std::reverse(Path.begin(), Path.end());
1462 
1463     // Warn that we're replacing the include/import with a module import.
1464     SmallString<128> PathString;
1465     for (unsigned I = 0, N = Path.size(); I != N; ++I) {
1466       if (I)
1467         PathString += '.';
1468       PathString += Path[I].first->getName();
1469     }
1470     int IncludeKind = 0;
1471 
1472     switch (IncludeTok.getIdentifierInfo()->getPPKeywordID()) {
1473     case tok::pp_include:
1474       IncludeKind = 0;
1475       break;
1476 
1477     case tok::pp_import:
1478       IncludeKind = 1;
1479       break;
1480 
1481     case tok::pp_include_next:
1482       IncludeKind = 2;
1483       break;
1484 
1485     case tok::pp___include_macros:
1486       IncludeKind = 3;
1487       break;
1488 
1489     default:
1490       llvm_unreachable("unknown include directive kind");
1491     }
1492 
1493     // Determine whether we are actually building the module that this
1494     // include directive maps to.
1495     bool BuildingImportedModule
1496       = Path[0].first->getName() == getLangOpts().CurrentModule;
1497 
1498     if (!BuildingImportedModule && getLangOpts().ObjC2) {
1499       // If we're not building the imported module, warn that we're going
1500       // to automatically turn this inclusion directive into a module import.
1501       // We only do this in Objective-C, where we have a module-import syntax.
1502       CharSourceRange ReplaceRange(SourceRange(HashLoc, CharEnd),
1503                                    /*IsTokenRange=*/false);
1504       Diag(HashLoc, diag::warn_auto_module_import)
1505         << IncludeKind << PathString
1506         << FixItHint::CreateReplacement(ReplaceRange,
1507              "@import " + PathString.str().str() + ";");
1508     }
1509 
1510     // Load the module.
1511     // If this was an #__include_macros directive, only make macros visible.
1512     Module::NameVisibilityKind Visibility
1513       = (IncludeKind == 3)? Module::MacrosVisible : Module::AllVisible;
1514     ModuleLoadResult Imported
1515       = TheModuleLoader.loadModule(IncludeTok.getLocation(), Path, Visibility,
1516                                    /*IsIncludeDirective=*/true);
1517     assert((Imported == 0 || Imported == SuggestedModule) &&
1518            "the imported module is different than the suggested one");
1519 
1520     if (!Imported && hadModuleLoaderFatalFailure()) {
1521       // With a fatal failure in the module loader, we abort parsing.
1522       Token &Result = IncludeTok;
1523       if (CurLexer) {
1524         Result.startToken();
1525         CurLexer->FormTokenWithChars(Result, CurLexer->BufferEnd, tok::eof);
1526         CurLexer->cutOffLexing();
1527       } else {
1528         assert(CurPTHLexer && "#include but no current lexer set!");
1529         CurPTHLexer->getEOF(Result);
1530       }
1531       return;
1532     }
1533 
1534     // If this header isn't part of the module we're building, we're done.
1535     if (!BuildingImportedModule && Imported) {
1536       if (Callbacks) {
1537         Callbacks->InclusionDirective(HashLoc, IncludeTok, Filename, isAngled,
1538                                       FilenameRange, File,
1539                                       SearchPath, RelativePath, Imported);
1540       }
1541       return;
1542     }
1543 
1544     // If we failed to find a submodule that we expected to find, we can
1545     // continue. Otherwise, there's an error in the included file, so we
1546     // don't want to include it.
1547     if (!BuildingImportedModule && !Imported.isMissingExpected()) {
1548       return;
1549     }
1550   }
1551 
1552   if (Callbacks && SuggestedModule) {
1553     // We didn't notify the callback object that we've seen an inclusion
1554     // directive before. Now that we are parsing the include normally and not
1555     // turning it to a module import, notify the callback object.
1556     Callbacks->InclusionDirective(HashLoc, IncludeTok, Filename, isAngled,
1557                                   FilenameRange, File,
1558                                   SearchPath, RelativePath,
1559                                   /*ImportedModule=*/0);
1560   }
1561 
1562   // The #included file will be considered to be a system header if either it is
1563   // in a system include directory, or if the #includer is a system include
1564   // header.
1565   SrcMgr::CharacteristicKind FileCharacter =
1566     std::max(HeaderInfo.getFileDirFlavor(File),
1567              SourceMgr.getFileCharacteristic(FilenameTok.getLocation()));
1568 
1569   // Ask HeaderInfo if we should enter this #include file.  If not, #including
1570   // this file will have no effect.
1571   if (!HeaderInfo.ShouldEnterIncludeFile(File, isImport)) {
1572     if (Callbacks)
1573       Callbacks->FileSkipped(*File, FilenameTok, FileCharacter);
1574     return;
1575   }
1576 
1577   // Look up the file, create a File ID for it.
1578   SourceLocation IncludePos = End;
1579   // If the filename string was the result of macro expansions, set the include
1580   // position on the file where it will be included and after the expansions.
1581   if (IncludePos.isMacroID())
1582     IncludePos = SourceMgr.getExpansionRange(IncludePos).second;
1583   FileID FID = SourceMgr.createFileID(File, IncludePos, FileCharacter);
1584   assert(!FID.isInvalid() && "Expected valid file ID");
1585 
1586   // Finally, if all is good, enter the new file!
1587   EnterSourceFile(FID, CurDir, FilenameTok.getLocation());
1588 }
1589 
1590 /// HandleIncludeNextDirective - Implements \#include_next.
1591 ///
1592 void Preprocessor::HandleIncludeNextDirective(SourceLocation HashLoc,
1593                                               Token &IncludeNextTok) {
1594   Diag(IncludeNextTok, diag::ext_pp_include_next_directive);
1595 
1596   // #include_next is like #include, except that we start searching after
1597   // the current found directory.  If we can't do this, issue a
1598   // diagnostic.
1599   const DirectoryLookup *Lookup = CurDirLookup;
1600   if (isInPrimaryFile()) {
1601     Lookup = 0;
1602     Diag(IncludeNextTok, diag::pp_include_next_in_primary);
1603   } else if (Lookup == 0) {
1604     Diag(IncludeNextTok, diag::pp_include_next_absolute_path);
1605   } else {
1606     // Start looking up in the next directory.
1607     ++Lookup;
1608   }
1609 
1610   return HandleIncludeDirective(HashLoc, IncludeNextTok, Lookup);
1611 }
1612 
1613 /// HandleMicrosoftImportDirective - Implements \#import for Microsoft Mode
1614 void Preprocessor::HandleMicrosoftImportDirective(Token &Tok) {
1615   // The Microsoft #import directive takes a type library and generates header
1616   // files from it, and includes those.  This is beyond the scope of what clang
1617   // does, so we ignore it and error out.  However, #import can optionally have
1618   // trailing attributes that span multiple lines.  We're going to eat those
1619   // so we can continue processing from there.
1620   Diag(Tok, diag::err_pp_import_directive_ms );
1621 
1622   // Read tokens until we get to the end of the directive.  Note that the
1623   // directive can be split over multiple lines using the backslash character.
1624   DiscardUntilEndOfDirective();
1625 }
1626 
1627 /// HandleImportDirective - Implements \#import.
1628 ///
1629 void Preprocessor::HandleImportDirective(SourceLocation HashLoc,
1630                                          Token &ImportTok) {
1631   if (!LangOpts.ObjC1) {  // #import is standard for ObjC.
1632     if (LangOpts.MicrosoftMode)
1633       return HandleMicrosoftImportDirective(ImportTok);
1634     Diag(ImportTok, diag::ext_pp_import_directive);
1635   }
1636   return HandleIncludeDirective(HashLoc, ImportTok, 0, true);
1637 }
1638 
1639 /// HandleIncludeMacrosDirective - The -imacros command line option turns into a
1640 /// pseudo directive in the predefines buffer.  This handles it by sucking all
1641 /// tokens through the preprocessor and discarding them (only keeping the side
1642 /// effects on the preprocessor).
1643 void Preprocessor::HandleIncludeMacrosDirective(SourceLocation HashLoc,
1644                                                 Token &IncludeMacrosTok) {
1645   // This directive should only occur in the predefines buffer.  If not, emit an
1646   // error and reject it.
1647   SourceLocation Loc = IncludeMacrosTok.getLocation();
1648   if (strcmp(SourceMgr.getBufferName(Loc), "<built-in>") != 0) {
1649     Diag(IncludeMacrosTok.getLocation(),
1650          diag::pp_include_macros_out_of_predefines);
1651     DiscardUntilEndOfDirective();
1652     return;
1653   }
1654 
1655   // Treat this as a normal #include for checking purposes.  If this is
1656   // successful, it will push a new lexer onto the include stack.
1657   HandleIncludeDirective(HashLoc, IncludeMacrosTok, 0, false);
1658 
1659   Token TmpTok;
1660   do {
1661     Lex(TmpTok);
1662     assert(TmpTok.isNot(tok::eof) && "Didn't find end of -imacros!");
1663   } while (TmpTok.isNot(tok::hashhash));
1664 }
1665 
1666 //===----------------------------------------------------------------------===//
1667 // Preprocessor Macro Directive Handling.
1668 //===----------------------------------------------------------------------===//
1669 
1670 /// ReadMacroDefinitionArgList - The ( starting an argument list of a macro
1671 /// definition has just been read.  Lex the rest of the arguments and the
1672 /// closing ), updating MI with what we learn.  Return true if an error occurs
1673 /// parsing the arg list.
1674 bool Preprocessor::ReadMacroDefinitionArgList(MacroInfo *MI, Token &Tok) {
1675   SmallVector<IdentifierInfo*, 32> Arguments;
1676 
1677   while (1) {
1678     LexUnexpandedToken(Tok);
1679     switch (Tok.getKind()) {
1680     case tok::r_paren:
1681       // Found the end of the argument list.
1682       if (Arguments.empty())  // #define FOO()
1683         return false;
1684       // Otherwise we have #define FOO(A,)
1685       Diag(Tok, diag::err_pp_expected_ident_in_arg_list);
1686       return true;
1687     case tok::ellipsis:  // #define X(... -> C99 varargs
1688       if (!LangOpts.C99)
1689         Diag(Tok, LangOpts.CPlusPlus11 ?
1690              diag::warn_cxx98_compat_variadic_macro :
1691              diag::ext_variadic_macro);
1692 
1693       // OpenCL v1.2 s6.9.e: variadic macros are not supported.
1694       if (LangOpts.OpenCL) {
1695         Diag(Tok, diag::err_pp_opencl_variadic_macros);
1696         return true;
1697       }
1698 
1699       // Lex the token after the identifier.
1700       LexUnexpandedToken(Tok);
1701       if (Tok.isNot(tok::r_paren)) {
1702         Diag(Tok, diag::err_pp_missing_rparen_in_macro_def);
1703         return true;
1704       }
1705       // Add the __VA_ARGS__ identifier as an argument.
1706       Arguments.push_back(Ident__VA_ARGS__);
1707       MI->setIsC99Varargs();
1708       MI->setArgumentList(&Arguments[0], Arguments.size(), BP);
1709       return false;
1710     case tok::eod:  // #define X(
1711       Diag(Tok, diag::err_pp_missing_rparen_in_macro_def);
1712       return true;
1713     default:
1714       // Handle keywords and identifiers here to accept things like
1715       // #define Foo(for) for.
1716       IdentifierInfo *II = Tok.getIdentifierInfo();
1717       if (II == 0) {
1718         // #define X(1
1719         Diag(Tok, diag::err_pp_invalid_tok_in_arg_list);
1720         return true;
1721       }
1722 
1723       // If this is already used as an argument, it is used multiple times (e.g.
1724       // #define X(A,A.
1725       if (std::find(Arguments.begin(), Arguments.end(), II) !=
1726           Arguments.end()) {  // C99 6.10.3p6
1727         Diag(Tok, diag::err_pp_duplicate_name_in_arg_list) << II;
1728         return true;
1729       }
1730 
1731       // Add the argument to the macro info.
1732       Arguments.push_back(II);
1733 
1734       // Lex the token after the identifier.
1735       LexUnexpandedToken(Tok);
1736 
1737       switch (Tok.getKind()) {
1738       default:          // #define X(A B
1739         Diag(Tok, diag::err_pp_expected_comma_in_arg_list);
1740         return true;
1741       case tok::r_paren: // #define X(A)
1742         MI->setArgumentList(&Arguments[0], Arguments.size(), BP);
1743         return false;
1744       case tok::comma:  // #define X(A,
1745         break;
1746       case tok::ellipsis:  // #define X(A... -> GCC extension
1747         // Diagnose extension.
1748         Diag(Tok, diag::ext_named_variadic_macro);
1749 
1750         // Lex the token after the identifier.
1751         LexUnexpandedToken(Tok);
1752         if (Tok.isNot(tok::r_paren)) {
1753           Diag(Tok, diag::err_pp_missing_rparen_in_macro_def);
1754           return true;
1755         }
1756 
1757         MI->setIsGNUVarargs();
1758         MI->setArgumentList(&Arguments[0], Arguments.size(), BP);
1759         return false;
1760       }
1761     }
1762   }
1763 }
1764 
1765 /// HandleDefineDirective - Implements \#define.  This consumes the entire macro
1766 /// line then lets the caller lex the next real token.
1767 void Preprocessor::HandleDefineDirective(Token &DefineTok,
1768                                          bool ImmediatelyAfterHeaderGuard) {
1769   ++NumDefined;
1770 
1771   Token MacroNameTok;
1772   ReadMacroName(MacroNameTok, 1);
1773 
1774   // Error reading macro name?  If so, diagnostic already issued.
1775   if (MacroNameTok.is(tok::eod))
1776     return;
1777 
1778   Token LastTok = MacroNameTok;
1779 
1780   // If we are supposed to keep comments in #defines, reenable comment saving
1781   // mode.
1782   if (CurLexer) CurLexer->SetCommentRetentionState(KeepMacroComments);
1783 
1784   // Create the new macro.
1785   MacroInfo *MI = AllocateMacroInfo(MacroNameTok.getLocation());
1786 
1787   Token Tok;
1788   LexUnexpandedToken(Tok);
1789 
1790   // If this is a function-like macro definition, parse the argument list,
1791   // marking each of the identifiers as being used as macro arguments.  Also,
1792   // check other constraints on the first token of the macro body.
1793   if (Tok.is(tok::eod)) {
1794     if (ImmediatelyAfterHeaderGuard) {
1795       // Save this macro information since it may part of a header guard.
1796       CurPPLexer->MIOpt.SetDefinedMacro(MacroNameTok.getIdentifierInfo(),
1797                                         MacroNameTok.getLocation());
1798     }
1799     // If there is no body to this macro, we have no special handling here.
1800   } else if (Tok.hasLeadingSpace()) {
1801     // This is a normal token with leading space.  Clear the leading space
1802     // marker on the first token to get proper expansion.
1803     Tok.clearFlag(Token::LeadingSpace);
1804   } else if (Tok.is(tok::l_paren)) {
1805     // This is a function-like macro definition.  Read the argument list.
1806     MI->setIsFunctionLike();
1807     if (ReadMacroDefinitionArgList(MI, LastTok)) {
1808       // Forget about MI.
1809       ReleaseMacroInfo(MI);
1810       // Throw away the rest of the line.
1811       if (CurPPLexer->ParsingPreprocessorDirective)
1812         DiscardUntilEndOfDirective();
1813       return;
1814     }
1815 
1816     // If this is a definition of a variadic C99 function-like macro, not using
1817     // the GNU named varargs extension, enabled __VA_ARGS__.
1818 
1819     // "Poison" __VA_ARGS__, which can only appear in the expansion of a macro.
1820     // This gets unpoisoned where it is allowed.
1821     assert(Ident__VA_ARGS__->isPoisoned() && "__VA_ARGS__ should be poisoned!");
1822     if (MI->isC99Varargs())
1823       Ident__VA_ARGS__->setIsPoisoned(false);
1824 
1825     // Read the first token after the arg list for down below.
1826     LexUnexpandedToken(Tok);
1827   } else if (LangOpts.C99 || LangOpts.CPlusPlus11) {
1828     // C99 requires whitespace between the macro definition and the body.  Emit
1829     // a diagnostic for something like "#define X+".
1830     Diag(Tok, diag::ext_c99_whitespace_required_after_macro_name);
1831   } else {
1832     // C90 6.8 TC1 says: "In the definition of an object-like macro, if the
1833     // first character of a replacement list is not a character required by
1834     // subclause 5.2.1, then there shall be white-space separation between the
1835     // identifier and the replacement list.".  5.2.1 lists this set:
1836     //   "A-Za-z0-9!"#%&'()*+,_./:;<=>?[\]^_{|}~" as well as whitespace, which
1837     // is irrelevant here.
1838     bool isInvalid = false;
1839     if (Tok.is(tok::at)) // @ is not in the list above.
1840       isInvalid = true;
1841     else if (Tok.is(tok::unknown)) {
1842       // If we have an unknown token, it is something strange like "`".  Since
1843       // all of valid characters would have lexed into a single character
1844       // token of some sort, we know this is not a valid case.
1845       isInvalid = true;
1846     }
1847     if (isInvalid)
1848       Diag(Tok, diag::ext_missing_whitespace_after_macro_name);
1849     else
1850       Diag(Tok, diag::warn_missing_whitespace_after_macro_name);
1851   }
1852 
1853   if (!Tok.is(tok::eod))
1854     LastTok = Tok;
1855 
1856   // Read the rest of the macro body.
1857   if (MI->isObjectLike()) {
1858     // Object-like macros are very simple, just read their body.
1859     while (Tok.isNot(tok::eod)) {
1860       LastTok = Tok;
1861       MI->AddTokenToBody(Tok);
1862       // Get the next token of the macro.
1863       LexUnexpandedToken(Tok);
1864     }
1865 
1866   } else {
1867     // Otherwise, read the body of a function-like macro.  While we are at it,
1868     // check C99 6.10.3.2p1: ensure that # operators are followed by macro
1869     // parameters in function-like macro expansions.
1870     while (Tok.isNot(tok::eod)) {
1871       LastTok = Tok;
1872 
1873       if (Tok.isNot(tok::hash) && Tok.isNot(tok::hashhash)) {
1874         MI->AddTokenToBody(Tok);
1875 
1876         // Get the next token of the macro.
1877         LexUnexpandedToken(Tok);
1878         continue;
1879       }
1880 
1881       if (Tok.is(tok::hashhash)) {
1882 
1883         // If we see token pasting, check if it looks like the gcc comma
1884         // pasting extension.  We'll use this information to suppress
1885         // diagnostics later on.
1886 
1887         // Get the next token of the macro.
1888         LexUnexpandedToken(Tok);
1889 
1890         if (Tok.is(tok::eod)) {
1891           MI->AddTokenToBody(LastTok);
1892           break;
1893         }
1894 
1895         unsigned NumTokens = MI->getNumTokens();
1896         if (NumTokens && Tok.getIdentifierInfo() == Ident__VA_ARGS__ &&
1897             MI->getReplacementToken(NumTokens-1).is(tok::comma))
1898           MI->setHasCommaPasting();
1899 
1900         // Things look ok, add the '##' and param name tokens to the macro.
1901         MI->AddTokenToBody(LastTok);
1902         MI->AddTokenToBody(Tok);
1903         LastTok = Tok;
1904 
1905         // Get the next token of the macro.
1906         LexUnexpandedToken(Tok);
1907         continue;
1908       }
1909 
1910       // Get the next token of the macro.
1911       LexUnexpandedToken(Tok);
1912 
1913       // Check for a valid macro arg identifier.
1914       if (Tok.getIdentifierInfo() == 0 ||
1915           MI->getArgumentNum(Tok.getIdentifierInfo()) == -1) {
1916 
1917         // If this is assembler-with-cpp mode, we accept random gibberish after
1918         // the '#' because '#' is often a comment character.  However, change
1919         // the kind of the token to tok::unknown so that the preprocessor isn't
1920         // confused.
1921         if (getLangOpts().AsmPreprocessor && Tok.isNot(tok::eod)) {
1922           LastTok.setKind(tok::unknown);
1923         } else {
1924           Diag(Tok, diag::err_pp_stringize_not_parameter);
1925           ReleaseMacroInfo(MI);
1926 
1927           // Disable __VA_ARGS__ again.
1928           Ident__VA_ARGS__->setIsPoisoned(true);
1929           return;
1930         }
1931       }
1932 
1933       // Things look ok, add the '#' and param name tokens to the macro.
1934       MI->AddTokenToBody(LastTok);
1935       MI->AddTokenToBody(Tok);
1936       LastTok = Tok;
1937 
1938       // Get the next token of the macro.
1939       LexUnexpandedToken(Tok);
1940     }
1941   }
1942 
1943 
1944   // Disable __VA_ARGS__ again.
1945   Ident__VA_ARGS__->setIsPoisoned(true);
1946 
1947   // Check that there is no paste (##) operator at the beginning or end of the
1948   // replacement list.
1949   unsigned NumTokens = MI->getNumTokens();
1950   if (NumTokens != 0) {
1951     if (MI->getReplacementToken(0).is(tok::hashhash)) {
1952       Diag(MI->getReplacementToken(0), diag::err_paste_at_start);
1953       ReleaseMacroInfo(MI);
1954       return;
1955     }
1956     if (MI->getReplacementToken(NumTokens-1).is(tok::hashhash)) {
1957       Diag(MI->getReplacementToken(NumTokens-1), diag::err_paste_at_end);
1958       ReleaseMacroInfo(MI);
1959       return;
1960     }
1961   }
1962 
1963   MI->setDefinitionEndLoc(LastTok.getLocation());
1964 
1965   // Finally, if this identifier already had a macro defined for it, verify that
1966   // the macro bodies are identical, and issue diagnostics if they are not.
1967   if (const MacroInfo *OtherMI=getMacroInfo(MacroNameTok.getIdentifierInfo())) {
1968     // It is very common for system headers to have tons of macro redefinitions
1969     // and for warnings to be disabled in system headers.  If this is the case,
1970     // then don't bother calling MacroInfo::isIdenticalTo.
1971     if (!getDiagnostics().getSuppressSystemWarnings() ||
1972         !SourceMgr.isInSystemHeader(DefineTok.getLocation())) {
1973       if (!OtherMI->isUsed() && OtherMI->isWarnIfUnused())
1974         Diag(OtherMI->getDefinitionLoc(), diag::pp_macro_not_used);
1975 
1976       // Warn if defining "__LINE__" and other builtins, per C99 6.10.8/4 and
1977       // C++ [cpp.predefined]p4, but allow it as an extension.
1978       if (OtherMI->isBuiltinMacro())
1979         Diag(MacroNameTok, diag::ext_pp_redef_builtin_macro);
1980       // Macros must be identical.  This means all tokens and whitespace
1981       // separation must be the same.  C99 6.10.3p2.
1982       else if (!OtherMI->isAllowRedefinitionsWithoutWarning() &&
1983                !MI->isIdenticalTo(*OtherMI, *this, /*Syntactic=*/LangOpts.MicrosoftExt)) {
1984         Diag(MI->getDefinitionLoc(), diag::ext_pp_macro_redef)
1985           << MacroNameTok.getIdentifierInfo();
1986         Diag(OtherMI->getDefinitionLoc(), diag::note_previous_definition);
1987       }
1988     }
1989     if (OtherMI->isWarnIfUnused())
1990       WarnUnusedMacroLocs.erase(OtherMI->getDefinitionLoc());
1991   }
1992 
1993   DefMacroDirective *MD =
1994       appendDefMacroDirective(MacroNameTok.getIdentifierInfo(), MI);
1995 
1996   assert(!MI->isUsed());
1997   // If we need warning for not using the macro, add its location in the
1998   // warn-because-unused-macro set. If it gets used it will be removed from set.
1999   if (isInPrimaryFile() && // don't warn for include'd macros.
2000       Diags->getDiagnosticLevel(diag::pp_macro_not_used,
2001           MI->getDefinitionLoc()) != DiagnosticsEngine::Ignored) {
2002     MI->setIsWarnIfUnused(true);
2003     WarnUnusedMacroLocs.insert(MI->getDefinitionLoc());
2004   }
2005 
2006   // If the callbacks want to know, tell them about the macro definition.
2007   if (Callbacks)
2008     Callbacks->MacroDefined(MacroNameTok, MD);
2009 }
2010 
2011 /// HandleUndefDirective - Implements \#undef.
2012 ///
2013 void Preprocessor::HandleUndefDirective(Token &UndefTok) {
2014   ++NumUndefined;
2015 
2016   Token MacroNameTok;
2017   ReadMacroName(MacroNameTok, 2);
2018 
2019   // Error reading macro name?  If so, diagnostic already issued.
2020   if (MacroNameTok.is(tok::eod))
2021     return;
2022 
2023   // Check to see if this is the last token on the #undef line.
2024   CheckEndOfDirective("undef");
2025 
2026   // Okay, we finally have a valid identifier to undef.
2027   MacroDirective *MD = getMacroDirective(MacroNameTok.getIdentifierInfo());
2028   const MacroInfo *MI = MD ? MD->getMacroInfo() : 0;
2029 
2030   // If the callbacks want to know, tell them about the macro #undef.
2031   // Note: no matter if the macro was defined or not.
2032   if (Callbacks)
2033     Callbacks->MacroUndefined(MacroNameTok, MD);
2034 
2035   // If the macro is not defined, this is a noop undef, just return.
2036   if (MI == 0) return;
2037 
2038   if (!MI->isUsed() && MI->isWarnIfUnused())
2039     Diag(MI->getDefinitionLoc(), diag::pp_macro_not_used);
2040 
2041   if (MI->isWarnIfUnused())
2042     WarnUnusedMacroLocs.erase(MI->getDefinitionLoc());
2043 
2044   appendMacroDirective(MacroNameTok.getIdentifierInfo(),
2045                        AllocateUndefMacroDirective(MacroNameTok.getLocation()));
2046 }
2047 
2048 
2049 //===----------------------------------------------------------------------===//
2050 // Preprocessor Conditional Directive Handling.
2051 //===----------------------------------------------------------------------===//
2052 
2053 /// HandleIfdefDirective - Implements the \#ifdef/\#ifndef directive.  isIfndef
2054 /// is true when this is a \#ifndef directive.  ReadAnyTokensBeforeDirective is
2055 /// true if any tokens have been returned or pp-directives activated before this
2056 /// \#ifndef has been lexed.
2057 ///
2058 void Preprocessor::HandleIfdefDirective(Token &Result, bool isIfndef,
2059                                         bool ReadAnyTokensBeforeDirective) {
2060   ++NumIf;
2061   Token DirectiveTok = Result;
2062 
2063   Token MacroNameTok;
2064   ReadMacroName(MacroNameTok);
2065 
2066   // Error reading macro name?  If so, diagnostic already issued.
2067   if (MacroNameTok.is(tok::eod)) {
2068     // Skip code until we get to #endif.  This helps with recovery by not
2069     // emitting an error when the #endif is reached.
2070     SkipExcludedConditionalBlock(DirectiveTok.getLocation(),
2071                                  /*Foundnonskip*/false, /*FoundElse*/false);
2072     return;
2073   }
2074 
2075   // Check to see if this is the last token on the #if[n]def line.
2076   CheckEndOfDirective(isIfndef ? "ifndef" : "ifdef");
2077 
2078   IdentifierInfo *MII = MacroNameTok.getIdentifierInfo();
2079   MacroDirective *MD = getMacroDirective(MII);
2080   MacroInfo *MI = MD ? MD->getMacroInfo() : 0;
2081 
2082   if (CurPPLexer->getConditionalStackDepth() == 0) {
2083     // If the start of a top-level #ifdef and if the macro is not defined,
2084     // inform MIOpt that this might be the start of a proper include guard.
2085     // Otherwise it is some other form of unknown conditional which we can't
2086     // handle.
2087     if (!ReadAnyTokensBeforeDirective && MI == 0) {
2088       assert(isIfndef && "#ifdef shouldn't reach here");
2089       CurPPLexer->MIOpt.EnterTopLevelIfndef(MII, MacroNameTok.getLocation());
2090     } else
2091       CurPPLexer->MIOpt.EnterTopLevelConditional();
2092   }
2093 
2094   // If there is a macro, process it.
2095   if (MI)  // Mark it used.
2096     markMacroAsUsed(MI);
2097 
2098   if (Callbacks) {
2099     if (isIfndef)
2100       Callbacks->Ifndef(DirectiveTok.getLocation(), MacroNameTok, MD);
2101     else
2102       Callbacks->Ifdef(DirectiveTok.getLocation(), MacroNameTok, MD);
2103   }
2104 
2105   // Should we include the stuff contained by this directive?
2106   if (!MI == isIfndef) {
2107     // Yes, remember that we are inside a conditional, then lex the next token.
2108     CurPPLexer->pushConditionalLevel(DirectiveTok.getLocation(),
2109                                      /*wasskip*/false, /*foundnonskip*/true,
2110                                      /*foundelse*/false);
2111   } else {
2112     // No, skip the contents of this block.
2113     SkipExcludedConditionalBlock(DirectiveTok.getLocation(),
2114                                  /*Foundnonskip*/false,
2115                                  /*FoundElse*/false);
2116   }
2117 }
2118 
2119 /// HandleIfDirective - Implements the \#if directive.
2120 ///
2121 void Preprocessor::HandleIfDirective(Token &IfToken,
2122                                      bool ReadAnyTokensBeforeDirective) {
2123   ++NumIf;
2124 
2125   // Parse and evaluate the conditional expression.
2126   IdentifierInfo *IfNDefMacro = 0;
2127   const SourceLocation ConditionalBegin = CurPPLexer->getSourceLocation();
2128   const bool ConditionalTrue = EvaluateDirectiveExpression(IfNDefMacro);
2129   const SourceLocation ConditionalEnd = CurPPLexer->getSourceLocation();
2130 
2131   // If this condition is equivalent to #ifndef X, and if this is the first
2132   // directive seen, handle it for the multiple-include optimization.
2133   if (CurPPLexer->getConditionalStackDepth() == 0) {
2134     if (!ReadAnyTokensBeforeDirective && IfNDefMacro && ConditionalTrue)
2135       // FIXME: Pass in the location of the macro name, not the 'if' token.
2136       CurPPLexer->MIOpt.EnterTopLevelIfndef(IfNDefMacro, IfToken.getLocation());
2137     else
2138       CurPPLexer->MIOpt.EnterTopLevelConditional();
2139   }
2140 
2141   if (Callbacks)
2142     Callbacks->If(IfToken.getLocation(),
2143                   SourceRange(ConditionalBegin, ConditionalEnd));
2144 
2145   // Should we include the stuff contained by this directive?
2146   if (ConditionalTrue) {
2147     // Yes, remember that we are inside a conditional, then lex the next token.
2148     CurPPLexer->pushConditionalLevel(IfToken.getLocation(), /*wasskip*/false,
2149                                    /*foundnonskip*/true, /*foundelse*/false);
2150   } else {
2151     // No, skip the contents of this block.
2152     SkipExcludedConditionalBlock(IfToken.getLocation(), /*Foundnonskip*/false,
2153                                  /*FoundElse*/false);
2154   }
2155 }
2156 
2157 /// HandleEndifDirective - Implements the \#endif directive.
2158 ///
2159 void Preprocessor::HandleEndifDirective(Token &EndifToken) {
2160   ++NumEndif;
2161 
2162   // Check that this is the whole directive.
2163   CheckEndOfDirective("endif");
2164 
2165   PPConditionalInfo CondInfo;
2166   if (CurPPLexer->popConditionalLevel(CondInfo)) {
2167     // No conditionals on the stack: this is an #endif without an #if.
2168     Diag(EndifToken, diag::err_pp_endif_without_if);
2169     return;
2170   }
2171 
2172   // If this the end of a top-level #endif, inform MIOpt.
2173   if (CurPPLexer->getConditionalStackDepth() == 0)
2174     CurPPLexer->MIOpt.ExitTopLevelConditional();
2175 
2176   assert(!CondInfo.WasSkipping && !CurPPLexer->LexingRawMode &&
2177          "This code should only be reachable in the non-skipping case!");
2178 
2179   if (Callbacks)
2180     Callbacks->Endif(EndifToken.getLocation(), CondInfo.IfLoc);
2181 }
2182 
2183 /// HandleElseDirective - Implements the \#else directive.
2184 ///
2185 void Preprocessor::HandleElseDirective(Token &Result) {
2186   ++NumElse;
2187 
2188   // #else directive in a non-skipping conditional... start skipping.
2189   CheckEndOfDirective("else");
2190 
2191   PPConditionalInfo CI;
2192   if (CurPPLexer->popConditionalLevel(CI)) {
2193     Diag(Result, diag::pp_err_else_without_if);
2194     return;
2195   }
2196 
2197   // If this is a top-level #else, inform the MIOpt.
2198   if (CurPPLexer->getConditionalStackDepth() == 0)
2199     CurPPLexer->MIOpt.EnterTopLevelConditional();
2200 
2201   // If this is a #else with a #else before it, report the error.
2202   if (CI.FoundElse) Diag(Result, diag::pp_err_else_after_else);
2203 
2204   if (Callbacks)
2205     Callbacks->Else(Result.getLocation(), CI.IfLoc);
2206 
2207   // Finally, skip the rest of the contents of this block.
2208   SkipExcludedConditionalBlock(CI.IfLoc, /*Foundnonskip*/true,
2209                                /*FoundElse*/true, Result.getLocation());
2210 }
2211 
2212 /// HandleElifDirective - Implements the \#elif directive.
2213 ///
2214 void Preprocessor::HandleElifDirective(Token &ElifToken) {
2215   ++NumElse;
2216 
2217   // #elif directive in a non-skipping conditional... start skipping.
2218   // We don't care what the condition is, because we will always skip it (since
2219   // the block immediately before it was included).
2220   const SourceLocation ConditionalBegin = CurPPLexer->getSourceLocation();
2221   DiscardUntilEndOfDirective();
2222   const SourceLocation ConditionalEnd = CurPPLexer->getSourceLocation();
2223 
2224   PPConditionalInfo CI;
2225   if (CurPPLexer->popConditionalLevel(CI)) {
2226     Diag(ElifToken, diag::pp_err_elif_without_if);
2227     return;
2228   }
2229 
2230   // If this is a top-level #elif, inform the MIOpt.
2231   if (CurPPLexer->getConditionalStackDepth() == 0)
2232     CurPPLexer->MIOpt.EnterTopLevelConditional();
2233 
2234   // If this is a #elif with a #else before it, report the error.
2235   if (CI.FoundElse) Diag(ElifToken, diag::pp_err_elif_after_else);
2236 
2237   if (Callbacks)
2238     Callbacks->Elif(ElifToken.getLocation(),
2239                     SourceRange(ConditionalBegin, ConditionalEnd), CI.IfLoc);
2240 
2241   // Finally, skip the rest of the contents of this block.
2242   SkipExcludedConditionalBlock(CI.IfLoc, /*Foundnonskip*/true,
2243                                /*FoundElse*/CI.FoundElse,
2244                                ElifToken.getLocation());
2245 }
2246