PPDirectives.cpp revision c56fff7fd231aebf4b152f60f8f11ef91835c48a
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"
28using namespace clang;
29
30//===----------------------------------------------------------------------===//
31// Utility Methods for Preprocessor Directive Handling.
32//===----------------------------------------------------------------------===//
33
34MacroInfo *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
54MacroInfo *Preprocessor::AllocateMacroInfo(SourceLocation L) {
55  MacroInfo *MI = AllocateMacroInfo();
56  new (MI) MacroInfo(L);
57  return MI;
58}
59
60MacroInfo *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  MacroInfo *MI =
65      (MacroInfo*)BP.Allocate(sizeof(MacroInfo) + sizeof(SubModuleID),
66                              llvm::AlignOf<MacroInfo>::Alignment);
67  new (MI) MacroInfo(L);
68  MI->FromASTFile = true;
69  MI->setOwningModuleID(SubModuleID);
70  return MI;
71}
72
73DefMacroDirective *
74Preprocessor::AllocateDefMacroDirective(MacroInfo *MI, SourceLocation Loc,
75                                        bool isImported) {
76  DefMacroDirective *MD = BP.Allocate<DefMacroDirective>();
77  new (MD) DefMacroDirective(MI, Loc, isImported);
78  return MD;
79}
80
81UndefMacroDirective *
82Preprocessor::AllocateUndefMacroDirective(SourceLocation UndefLoc) {
83  UndefMacroDirective *MD = BP.Allocate<UndefMacroDirective>();
84  new (MD) UndefMacroDirective(UndefLoc);
85  return MD;
86}
87
88VisibilityMacroDirective *
89Preprocessor::AllocateVisibilityMacroDirective(SourceLocation Loc,
90                                               bool isPublic) {
91  VisibilityMacroDirective *MD = BP.Allocate<VisibilityMacroDirective>();
92  new (MD) VisibilityMacroDirective(Loc, isPublic);
93  return MD;
94}
95
96/// \brief Release the specified MacroInfo to be reused for allocating
97/// new MacroInfo objects.
98void Preprocessor::ReleaseMacroInfo(MacroInfo *MI) {
99  MacroInfoChain *MIChain = (MacroInfoChain*) MI;
100  if (MacroInfoChain *Prev = MIChain->Prev) {
101    MacroInfoChain *Next = MIChain->Next;
102    Prev->Next = Next;
103    if (Next)
104      Next->Prev = Prev;
105  }
106  else {
107    assert(MIChainHead == MIChain);
108    MIChainHead = MIChain->Next;
109    MIChainHead->Prev = 0;
110  }
111  MIChain->Next = MICache;
112  MICache = MIChain;
113
114  MI->Destroy();
115}
116
117/// \brief Read and discard all tokens remaining on the current line until
118/// the tok::eod token is found.
119void Preprocessor::DiscardUntilEndOfDirective() {
120  Token Tmp;
121  do {
122    LexUnexpandedToken(Tmp);
123    assert(Tmp.isNot(tok::eof) && "EOF seen while discarding directive tokens");
124  } while (Tmp.isNot(tok::eod));
125}
126
127/// \brief Lex and validate a macro name, which occurs after a
128/// \#define or \#undef.
129///
130/// This sets the token kind to eod and discards the rest
131/// of the macro line if the macro name is invalid.  \p isDefineUndef is 1 if
132/// this is due to a a \#define, 2 if \#undef directive, 0 if it is something
133/// else (e.g. \#ifdef).
134void Preprocessor::ReadMacroName(Token &MacroNameTok, char isDefineUndef) {
135  // Read the token, don't allow macro expansion on it.
136  LexUnexpandedToken(MacroNameTok);
137
138  if (MacroNameTok.is(tok::code_completion)) {
139    if (CodeComplete)
140      CodeComplete->CodeCompleteMacroName(isDefineUndef == 1);
141    setCodeCompletionReached();
142    LexUnexpandedToken(MacroNameTok);
143  }
144
145  // Missing macro name?
146  if (MacroNameTok.is(tok::eod)) {
147    Diag(MacroNameTok, diag::err_pp_missing_macro_name);
148    return;
149  }
150
151  IdentifierInfo *II = MacroNameTok.getIdentifierInfo();
152  if (II == 0) {
153    bool Invalid = false;
154    std::string Spelling = getSpelling(MacroNameTok, &Invalid);
155    if (Invalid)
156      return;
157
158    const IdentifierInfo &Info = Identifiers.get(Spelling);
159
160    // Allow #defining |and| and friends in microsoft mode.
161    if (Info.isCPlusPlusOperatorKeyword() && getLangOpts().MicrosoftMode) {
162      MacroNameTok.setIdentifierInfo(getIdentifierInfo(Spelling));
163      return;
164    }
165
166    if (Info.isCPlusPlusOperatorKeyword())
167      // C++ 2.5p2: Alternative tokens behave the same as its primary token
168      // except for their spellings.
169      Diag(MacroNameTok, diag::err_pp_operator_used_as_macro_name) << Spelling;
170    else
171      Diag(MacroNameTok, diag::err_pp_macro_not_identifier);
172    // Fall through on error.
173  } else if (isDefineUndef && II->getPPKeywordID() == tok::pp_defined) {
174    // Error if defining "defined": C99 6.10.8/4, C++ [cpp.predefined]p4.
175    Diag(MacroNameTok, diag::err_defined_macro_name);
176  } else if (isDefineUndef == 2 && II->hasMacroDefinition() &&
177             getMacroInfo(II)->isBuiltinMacro()) {
178    // Warn if undefining "__LINE__" and other builtins, per C99 6.10.8/4
179    // and C++ [cpp.predefined]p4], but allow it as an extension.
180    Diag(MacroNameTok, diag::ext_pp_undef_builtin_macro);
181    return;
182  } else {
183    // Okay, we got a good identifier node.  Return it.
184    return;
185  }
186
187  // Invalid macro name, read and discard the rest of the line.  Then set the
188  // token kind to tok::eod.
189  MacroNameTok.setKind(tok::eod);
190  return DiscardUntilEndOfDirective();
191}
192
193/// \brief Ensure that the next token is a tok::eod token.
194///
195/// If not, emit a diagnostic and consume up until the eod.  If EnableMacros is
196/// true, then we consider macros that expand to zero tokens as being ok.
197void Preprocessor::CheckEndOfDirective(const char *DirType, bool EnableMacros) {
198  Token Tmp;
199  // Lex unexpanded tokens for most directives: macros might expand to zero
200  // tokens, causing us to miss diagnosing invalid lines.  Some directives (like
201  // #line) allow empty macros.
202  if (EnableMacros)
203    Lex(Tmp);
204  else
205    LexUnexpandedToken(Tmp);
206
207  // There should be no tokens after the directive, but we allow them as an
208  // extension.
209  while (Tmp.is(tok::comment))  // Skip comments in -C mode.
210    LexUnexpandedToken(Tmp);
211
212  if (Tmp.isNot(tok::eod)) {
213    // Add a fixit in GNU/C99/C++ mode.  Don't offer a fixit for strict-C89,
214    // or if this is a macro-style preprocessing directive, because it is more
215    // trouble than it is worth to insert /**/ and check that there is no /**/
216    // in the range also.
217    FixItHint Hint;
218    if ((LangOpts.GNUMode || LangOpts.C99 || LangOpts.CPlusPlus) &&
219        !CurTokenLexer)
220      Hint = FixItHint::CreateInsertion(Tmp.getLocation(),"//");
221    Diag(Tmp, diag::ext_pp_extra_tokens_at_eol) << DirType << Hint;
222    DiscardUntilEndOfDirective();
223  }
224}
225
226
227
228/// SkipExcludedConditionalBlock - We just read a \#if or related directive and
229/// decided that the subsequent tokens are in the \#if'd out portion of the
230/// file.  Lex the rest of the file, until we see an \#endif.  If
231/// FoundNonSkipPortion is true, then we have already emitted code for part of
232/// this \#if directive, so \#else/\#elif blocks should never be entered.
233/// If ElseOk is true, then \#else directives are ok, if not, then we have
234/// already seen one so a \#else directive is a duplicate.  When this returns,
235/// the caller can lex the first valid token.
236void Preprocessor::SkipExcludedConditionalBlock(SourceLocation IfTokenLoc,
237                                                bool FoundNonSkipPortion,
238                                                bool FoundElse,
239                                                SourceLocation ElseLoc) {
240  ++NumSkipped;
241  assert(CurTokenLexer == 0 && CurPPLexer && "Lexing a macro, not a file?");
242
243  CurPPLexer->pushConditionalLevel(IfTokenLoc, /*isSkipping*/false,
244                                 FoundNonSkipPortion, FoundElse);
245
246  if (CurPTHLexer) {
247    PTHSkipExcludedConditionalBlock();
248    return;
249  }
250
251  // Enter raw mode to disable identifier lookup (and thus macro expansion),
252  // disabling warnings, etc.
253  CurPPLexer->LexingRawMode = true;
254  Token Tok;
255  while (1) {
256    CurLexer->Lex(Tok);
257
258    if (Tok.is(tok::code_completion)) {
259      if (CodeComplete)
260        CodeComplete->CodeCompleteInConditionalExclusion();
261      setCodeCompletionReached();
262      continue;
263    }
264
265    // If this is the end of the buffer, we have an error.
266    if (Tok.is(tok::eof)) {
267      // Emit errors for each unterminated conditional on the stack, including
268      // the current one.
269      while (!CurPPLexer->ConditionalStack.empty()) {
270        if (CurLexer->getFileLoc() != CodeCompletionFileLoc)
271          Diag(CurPPLexer->ConditionalStack.back().IfLoc,
272               diag::err_pp_unterminated_conditional);
273        CurPPLexer->ConditionalStack.pop_back();
274      }
275
276      // Just return and let the caller lex after this #include.
277      break;
278    }
279
280    // If this token is not a preprocessor directive, just skip it.
281    if (Tok.isNot(tok::hash) || !Tok.isAtStartOfLine())
282      continue;
283
284    // We just parsed a # character at the start of a line, so we're in
285    // directive mode.  Tell the lexer this so any newlines we see will be
286    // converted into an EOD token (this terminates the macro).
287    CurPPLexer->ParsingPreprocessorDirective = true;
288    if (CurLexer) CurLexer->SetKeepWhitespaceMode(false);
289
290
291    // Read the next token, the directive flavor.
292    LexUnexpandedToken(Tok);
293
294    // If this isn't an identifier directive (e.g. is "# 1\n" or "#\n", or
295    // something bogus), skip it.
296    if (Tok.isNot(tok::raw_identifier)) {
297      CurPPLexer->ParsingPreprocessorDirective = false;
298      // Restore comment saving mode.
299      if (CurLexer) CurLexer->resetExtendedTokenMode();
300      continue;
301    }
302
303    // If the first letter isn't i or e, it isn't intesting to us.  We know that
304    // this is safe in the face of spelling differences, because there is no way
305    // to spell an i/e in a strange way that is another letter.  Skipping this
306    // allows us to avoid looking up the identifier info for #define/#undef and
307    // other common directives.
308    const char *RawCharData = Tok.getRawIdentifierData();
309
310    char FirstChar = RawCharData[0];
311    if (FirstChar >= 'a' && FirstChar <= 'z' &&
312        FirstChar != 'i' && FirstChar != 'e') {
313      CurPPLexer->ParsingPreprocessorDirective = false;
314      // Restore comment saving mode.
315      if (CurLexer) CurLexer->resetExtendedTokenMode();
316      continue;
317    }
318
319    // Get the identifier name without trigraphs or embedded newlines.  Note
320    // that we can't use Tok.getIdentifierInfo() because its lookup is disabled
321    // when skipping.
322    char DirectiveBuf[20];
323    StringRef Directive;
324    if (!Tok.needsCleaning() && Tok.getLength() < 20) {
325      Directive = StringRef(RawCharData, Tok.getLength());
326    } else {
327      std::string DirectiveStr = getSpelling(Tok);
328      unsigned IdLen = DirectiveStr.size();
329      if (IdLen >= 20) {
330        CurPPLexer->ParsingPreprocessorDirective = false;
331        // Restore comment saving mode.
332        if (CurLexer) CurLexer->resetExtendedTokenMode();
333        continue;
334      }
335      memcpy(DirectiveBuf, &DirectiveStr[0], IdLen);
336      Directive = StringRef(DirectiveBuf, IdLen);
337    }
338
339    if (Directive.startswith("if")) {
340      StringRef Sub = Directive.substr(2);
341      if (Sub.empty() ||   // "if"
342          Sub == "def" ||   // "ifdef"
343          Sub == "ndef") {  // "ifndef"
344        // We know the entire #if/#ifdef/#ifndef block will be skipped, don't
345        // bother parsing the condition.
346        DiscardUntilEndOfDirective();
347        CurPPLexer->pushConditionalLevel(Tok.getLocation(), /*wasskipping*/true,
348                                       /*foundnonskip*/false,
349                                       /*foundelse*/false);
350      }
351    } else if (Directive[0] == 'e') {
352      StringRef Sub = Directive.substr(1);
353      if (Sub == "ndif") {  // "endif"
354        PPConditionalInfo CondInfo;
355        CondInfo.WasSkipping = true; // Silence bogus warning.
356        bool InCond = CurPPLexer->popConditionalLevel(CondInfo);
357        (void)InCond;  // Silence warning in no-asserts mode.
358        assert(!InCond && "Can't be skipping if not in a conditional!");
359
360        // If we popped the outermost skipping block, we're done skipping!
361        if (!CondInfo.WasSkipping) {
362          // Restore the value of LexingRawMode so that trailing comments
363          // are handled correctly, if we've reached the outermost block.
364          CurPPLexer->LexingRawMode = false;
365          CheckEndOfDirective("endif");
366          CurPPLexer->LexingRawMode = true;
367          if (Callbacks)
368            Callbacks->Endif(Tok.getLocation(), CondInfo.IfLoc);
369          break;
370        } else {
371          DiscardUntilEndOfDirective();
372        }
373      } else if (Sub == "lse") { // "else".
374        // #else directive in a skipping conditional.  If not in some other
375        // skipping conditional, and if #else hasn't already been seen, enter it
376        // as a non-skipping conditional.
377        PPConditionalInfo &CondInfo = CurPPLexer->peekConditionalLevel();
378
379        // If this is a #else with a #else before it, report the error.
380        if (CondInfo.FoundElse) Diag(Tok, diag::pp_err_else_after_else);
381
382        // Note that we've seen a #else in this conditional.
383        CondInfo.FoundElse = true;
384
385        // If the conditional is at the top level, and the #if block wasn't
386        // entered, enter the #else block now.
387        if (!CondInfo.WasSkipping && !CondInfo.FoundNonSkip) {
388          CondInfo.FoundNonSkip = true;
389          // Restore the value of LexingRawMode so that trailing comments
390          // are handled correctly.
391          CurPPLexer->LexingRawMode = false;
392          CheckEndOfDirective("else");
393          CurPPLexer->LexingRawMode = true;
394          if (Callbacks)
395            Callbacks->Else(Tok.getLocation(), CondInfo.IfLoc);
396          break;
397        } else {
398          DiscardUntilEndOfDirective();  // C99 6.10p4.
399        }
400      } else if (Sub == "lif") {  // "elif".
401        PPConditionalInfo &CondInfo = CurPPLexer->peekConditionalLevel();
402
403        bool ShouldEnter;
404        const SourceLocation ConditionalBegin = CurPPLexer->getSourceLocation();
405        // If this is in a skipping block or if we're already handled this #if
406        // block, don't bother parsing the condition.
407        if (CondInfo.WasSkipping || CondInfo.FoundNonSkip) {
408          DiscardUntilEndOfDirective();
409          ShouldEnter = false;
410        } else {
411          // Restore the value of LexingRawMode so that identifiers are
412          // looked up, etc, inside the #elif expression.
413          assert(CurPPLexer->LexingRawMode && "We have to be skipping here!");
414          CurPPLexer->LexingRawMode = false;
415          IdentifierInfo *IfNDefMacro = 0;
416          ShouldEnter = EvaluateDirectiveExpression(IfNDefMacro);
417          CurPPLexer->LexingRawMode = true;
418        }
419        const SourceLocation ConditionalEnd = CurPPLexer->getSourceLocation();
420
421        // If this is a #elif with a #else before it, report the error.
422        if (CondInfo.FoundElse) Diag(Tok, diag::pp_err_elif_after_else);
423
424        // If this condition is true, enter it!
425        if (ShouldEnter) {
426          CondInfo.FoundNonSkip = true;
427          if (Callbacks)
428            Callbacks->Elif(Tok.getLocation(),
429                            SourceRange(ConditionalBegin, ConditionalEnd),
430                            CondInfo.IfLoc);
431          break;
432        }
433      }
434    }
435
436    CurPPLexer->ParsingPreprocessorDirective = false;
437    // Restore comment saving mode.
438    if (CurLexer) CurLexer->resetExtendedTokenMode();
439  }
440
441  // Finally, if we are out of the conditional (saw an #endif or ran off the end
442  // of the file, just stop skipping and return to lexing whatever came after
443  // the #if block.
444  CurPPLexer->LexingRawMode = false;
445
446  if (Callbacks) {
447    SourceLocation BeginLoc = ElseLoc.isValid() ? ElseLoc : IfTokenLoc;
448    Callbacks->SourceRangeSkipped(SourceRange(BeginLoc, Tok.getLocation()));
449  }
450}
451
452void Preprocessor::PTHSkipExcludedConditionalBlock() {
453
454  while (1) {
455    assert(CurPTHLexer);
456    assert(CurPTHLexer->LexingRawMode == false);
457
458    // Skip to the next '#else', '#elif', or #endif.
459    if (CurPTHLexer->SkipBlock()) {
460      // We have reached an #endif.  Both the '#' and 'endif' tokens
461      // have been consumed by the PTHLexer.  Just pop off the condition level.
462      PPConditionalInfo CondInfo;
463      bool InCond = CurPTHLexer->popConditionalLevel(CondInfo);
464      (void)InCond;  // Silence warning in no-asserts mode.
465      assert(!InCond && "Can't be skipping if not in a conditional!");
466      break;
467    }
468
469    // We have reached a '#else' or '#elif'.  Lex the next token to get
470    // the directive flavor.
471    Token Tok;
472    LexUnexpandedToken(Tok);
473
474    // We can actually look up the IdentifierInfo here since we aren't in
475    // raw mode.
476    tok::PPKeywordKind K = Tok.getIdentifierInfo()->getPPKeywordID();
477
478    if (K == tok::pp_else) {
479      // #else: Enter the else condition.  We aren't in a nested condition
480      //  since we skip those. We're always in the one matching the last
481      //  blocked we skipped.
482      PPConditionalInfo &CondInfo = CurPTHLexer->peekConditionalLevel();
483      // Note that we've seen a #else in this conditional.
484      CondInfo.FoundElse = true;
485
486      // If the #if block wasn't entered then enter the #else block now.
487      if (!CondInfo.FoundNonSkip) {
488        CondInfo.FoundNonSkip = true;
489
490        // Scan until the eod token.
491        CurPTHLexer->ParsingPreprocessorDirective = true;
492        DiscardUntilEndOfDirective();
493        CurPTHLexer->ParsingPreprocessorDirective = false;
494
495        break;
496      }
497
498      // Otherwise skip this block.
499      continue;
500    }
501
502    assert(K == tok::pp_elif);
503    PPConditionalInfo &CondInfo = CurPTHLexer->peekConditionalLevel();
504
505    // If this is a #elif with a #else before it, report the error.
506    if (CondInfo.FoundElse)
507      Diag(Tok, diag::pp_err_elif_after_else);
508
509    // If this is in a skipping block or if we're already handled this #if
510    // block, don't bother parsing the condition.  We just skip this block.
511    if (CondInfo.FoundNonSkip)
512      continue;
513
514    // Evaluate the condition of the #elif.
515    IdentifierInfo *IfNDefMacro = 0;
516    CurPTHLexer->ParsingPreprocessorDirective = true;
517    bool ShouldEnter = EvaluateDirectiveExpression(IfNDefMacro);
518    CurPTHLexer->ParsingPreprocessorDirective = false;
519
520    // If this condition is true, enter it!
521    if (ShouldEnter) {
522      CondInfo.FoundNonSkip = true;
523      break;
524    }
525
526    // Otherwise, skip this block and go to the next one.
527    continue;
528  }
529}
530
531const FileEntry *Preprocessor::LookupFile(
532    StringRef Filename,
533    bool isAngled,
534    const DirectoryLookup *FromDir,
535    const DirectoryLookup *&CurDir,
536    SmallVectorImpl<char> *SearchPath,
537    SmallVectorImpl<char> *RelativePath,
538    Module **SuggestedModule,
539    bool SkipCache) {
540  // If the header lookup mechanism may be relative to the current file, pass in
541  // info about where the current file is.
542  const FileEntry *CurFileEnt = 0;
543  if (!FromDir) {
544    FileID FID = getCurrentFileLexer()->getFileID();
545    CurFileEnt = SourceMgr.getFileEntryForID(FID);
546
547    // If there is no file entry associated with this file, it must be the
548    // predefines buffer.  Any other file is not lexed with a normal lexer, so
549    // it won't be scanned for preprocessor directives.   If we have the
550    // predefines buffer, resolve #include references (which come from the
551    // -include command line argument) as if they came from the main file, this
552    // affects file lookup etc.
553    if (CurFileEnt == 0) {
554      FID = SourceMgr.getMainFileID();
555      CurFileEnt = SourceMgr.getFileEntryForID(FID);
556    }
557  }
558
559  // Do a standard file entry lookup.
560  CurDir = CurDirLookup;
561  const FileEntry *FE = HeaderInfo.LookupFile(
562      Filename, isAngled, FromDir, CurDir, CurFileEnt,
563      SearchPath, RelativePath, SuggestedModule, SkipCache);
564  if (FE) return FE;
565
566  // Otherwise, see if this is a subframework header.  If so, this is relative
567  // to one of the headers on the #include stack.  Walk the list of the current
568  // headers on the #include stack and pass them to HeaderInfo.
569  if (IsFileLexer()) {
570    if ((CurFileEnt = SourceMgr.getFileEntryForID(CurPPLexer->getFileID())))
571      if ((FE = HeaderInfo.LookupSubframeworkHeader(Filename, CurFileEnt,
572                                                    SearchPath, RelativePath,
573                                                    SuggestedModule)))
574        return FE;
575  }
576
577  for (unsigned i = 0, e = IncludeMacroStack.size(); i != e; ++i) {
578    IncludeStackInfo &ISEntry = IncludeMacroStack[e-i-1];
579    if (IsFileLexer(ISEntry)) {
580      if ((CurFileEnt =
581           SourceMgr.getFileEntryForID(ISEntry.ThePPLexer->getFileID())))
582        if ((FE = HeaderInfo.LookupSubframeworkHeader(
583                Filename, CurFileEnt, SearchPath, RelativePath,
584                SuggestedModule)))
585          return FE;
586    }
587  }
588
589  // Otherwise, we really couldn't find the file.
590  return 0;
591}
592
593
594//===----------------------------------------------------------------------===//
595// Preprocessor Directive Handling.
596//===----------------------------------------------------------------------===//
597
598class Preprocessor::ResetMacroExpansionHelper {
599public:
600  ResetMacroExpansionHelper(Preprocessor *pp)
601    : PP(pp), save(pp->DisableMacroExpansion) {
602    if (pp->MacroExpansionInDirectivesOverride)
603      pp->DisableMacroExpansion = false;
604  }
605  ~ResetMacroExpansionHelper() {
606    PP->DisableMacroExpansion = save;
607  }
608private:
609  Preprocessor *PP;
610  bool save;
611};
612
613/// HandleDirective - This callback is invoked when the lexer sees a # token
614/// at the start of a line.  This consumes the directive, modifies the
615/// lexer/preprocessor state, and advances the lexer(s) so that the next token
616/// read is the correct one.
617void Preprocessor::HandleDirective(Token &Result) {
618  // FIXME: Traditional: # with whitespace before it not recognized by K&R?
619
620  // We just parsed a # character at the start of a line, so we're in directive
621  // mode.  Tell the lexer this so any newlines we see will be converted into an
622  // EOD token (which terminates the directive).
623  CurPPLexer->ParsingPreprocessorDirective = true;
624  if (CurLexer) CurLexer->SetKeepWhitespaceMode(false);
625
626  ++NumDirectives;
627
628  // We are about to read a token.  For the multiple-include optimization FA to
629  // work, we have to remember if we had read any tokens *before* this
630  // pp-directive.
631  bool ReadAnyTokensBeforeDirective =CurPPLexer->MIOpt.getHasReadAnyTokensVal();
632
633  // Save the '#' token in case we need to return it later.
634  Token SavedHash = Result;
635
636  // Read the next token, the directive flavor.  This isn't expanded due to
637  // C99 6.10.3p8.
638  LexUnexpandedToken(Result);
639
640  // C99 6.10.3p11: Is this preprocessor directive in macro invocation?  e.g.:
641  //   #define A(x) #x
642  //   A(abc
643  //     #warning blah
644  //   def)
645  // If so, the user is relying on undefined behavior, emit a diagnostic. Do
646  // not support this for #include-like directives, since that can result in
647  // terrible diagnostics, and does not work in GCC.
648  if (InMacroArgs) {
649    if (IdentifierInfo *II = Result.getIdentifierInfo()) {
650      switch (II->getPPKeywordID()) {
651      case tok::pp_include:
652      case tok::pp_import:
653      case tok::pp_include_next:
654      case tok::pp___include_macros:
655        Diag(Result, diag::err_embedded_include) << II->getName();
656        DiscardUntilEndOfDirective();
657        return;
658      default:
659        break;
660      }
661    }
662    Diag(Result, diag::ext_embedded_directive);
663  }
664
665  // Temporarily enable macro expansion if set so
666  // and reset to previous state when returning from this function.
667  ResetMacroExpansionHelper helper(this);
668
669  switch (Result.getKind()) {
670  case tok::eod:
671    return;   // null directive.
672  case tok::code_completion:
673    if (CodeComplete)
674      CodeComplete->CodeCompleteDirective(
675                                    CurPPLexer->getConditionalStackDepth() > 0);
676    setCodeCompletionReached();
677    return;
678  case tok::numeric_constant:  // # 7  GNU line marker directive.
679    if (getLangOpts().AsmPreprocessor)
680      break;  // # 4 is not a preprocessor directive in .S files.
681    return HandleDigitDirective(Result);
682  default:
683    IdentifierInfo *II = Result.getIdentifierInfo();
684    if (II == 0) break;  // Not an identifier.
685
686    // Ask what the preprocessor keyword ID is.
687    switch (II->getPPKeywordID()) {
688    default: break;
689    // C99 6.10.1 - Conditional Inclusion.
690    case tok::pp_if:
691      return HandleIfDirective(Result, ReadAnyTokensBeforeDirective);
692    case tok::pp_ifdef:
693      return HandleIfdefDirective(Result, false, true/*not valid for miopt*/);
694    case tok::pp_ifndef:
695      return HandleIfdefDirective(Result, true, ReadAnyTokensBeforeDirective);
696    case tok::pp_elif:
697      return HandleElifDirective(Result);
698    case tok::pp_else:
699      return HandleElseDirective(Result);
700    case tok::pp_endif:
701      return HandleEndifDirective(Result);
702
703    // C99 6.10.2 - Source File Inclusion.
704    case tok::pp_include:
705      // Handle #include.
706      return HandleIncludeDirective(SavedHash.getLocation(), Result);
707    case tok::pp___include_macros:
708      // Handle -imacros.
709      return HandleIncludeMacrosDirective(SavedHash.getLocation(), Result);
710
711    // C99 6.10.3 - Macro Replacement.
712    case tok::pp_define:
713      return HandleDefineDirective(Result);
714    case tok::pp_undef:
715      return HandleUndefDirective(Result);
716
717    // C99 6.10.4 - Line Control.
718    case tok::pp_line:
719      return HandleLineDirective(Result);
720
721    // C99 6.10.5 - Error Directive.
722    case tok::pp_error:
723      return HandleUserDiagnosticDirective(Result, false);
724
725    // C99 6.10.6 - Pragma Directive.
726    case tok::pp_pragma:
727      return HandlePragmaDirective(PIK_HashPragma);
728
729    // GNU Extensions.
730    case tok::pp_import:
731      return HandleImportDirective(SavedHash.getLocation(), Result);
732    case tok::pp_include_next:
733      return HandleIncludeNextDirective(SavedHash.getLocation(), Result);
734
735    case tok::pp_warning:
736      Diag(Result, diag::ext_pp_warning_directive);
737      return HandleUserDiagnosticDirective(Result, true);
738    case tok::pp_ident:
739      return HandleIdentSCCSDirective(Result);
740    case tok::pp_sccs:
741      return HandleIdentSCCSDirective(Result);
742    case tok::pp_assert:
743      //isExtension = true;  // FIXME: implement #assert
744      break;
745    case tok::pp_unassert:
746      //isExtension = true;  // FIXME: implement #unassert
747      break;
748
749    case tok::pp___public_macro:
750      if (getLangOpts().Modules)
751        return HandleMacroPublicDirective(Result);
752      break;
753
754    case tok::pp___private_macro:
755      if (getLangOpts().Modules)
756        return HandleMacroPrivateDirective(Result);
757      break;
758    }
759    break;
760  }
761
762  // If this is a .S file, treat unknown # directives as non-preprocessor
763  // directives.  This is important because # may be a comment or introduce
764  // various pseudo-ops.  Just return the # token and push back the following
765  // token to be lexed next time.
766  if (getLangOpts().AsmPreprocessor) {
767    Token *Toks = new Token[2];
768    // Return the # and the token after it.
769    Toks[0] = SavedHash;
770    Toks[1] = Result;
771
772    // If the second token is a hashhash token, then we need to translate it to
773    // unknown so the token lexer doesn't try to perform token pasting.
774    if (Result.is(tok::hashhash))
775      Toks[1].setKind(tok::unknown);
776
777    // Enter this token stream so that we re-lex the tokens.  Make sure to
778    // enable macro expansion, in case the token after the # is an identifier
779    // that is expanded.
780    EnterTokenStream(Toks, 2, false, true);
781    return;
782  }
783
784  // If we reached here, the preprocessing token is not valid!
785  Diag(Result, diag::err_pp_invalid_directive);
786
787  // Read the rest of the PP line.
788  DiscardUntilEndOfDirective();
789
790  // Okay, we're done parsing the directive.
791}
792
793/// GetLineValue - Convert a numeric token into an unsigned value, emitting
794/// Diagnostic DiagID if it is invalid, and returning the value in Val.
795static bool GetLineValue(Token &DigitTok, unsigned &Val,
796                         unsigned DiagID, Preprocessor &PP) {
797  if (DigitTok.isNot(tok::numeric_constant)) {
798    PP.Diag(DigitTok, DiagID);
799
800    if (DigitTok.isNot(tok::eod))
801      PP.DiscardUntilEndOfDirective();
802    return true;
803  }
804
805  SmallString<64> IntegerBuffer;
806  IntegerBuffer.resize(DigitTok.getLength());
807  const char *DigitTokBegin = &IntegerBuffer[0];
808  bool Invalid = false;
809  unsigned ActualLength = PP.getSpelling(DigitTok, DigitTokBegin, &Invalid);
810  if (Invalid)
811    return true;
812
813  // Verify that we have a simple digit-sequence, and compute the value.  This
814  // is always a simple digit string computed in decimal, so we do this manually
815  // here.
816  Val = 0;
817  for (unsigned i = 0; i != ActualLength; ++i) {
818    if (!isDigit(DigitTokBegin[i])) {
819      PP.Diag(PP.AdvanceToTokenCharacter(DigitTok.getLocation(), i),
820              diag::err_pp_line_digit_sequence);
821      PP.DiscardUntilEndOfDirective();
822      return true;
823    }
824
825    unsigned NextVal = Val*10+(DigitTokBegin[i]-'0');
826    if (NextVal < Val) { // overflow.
827      PP.Diag(DigitTok, DiagID);
828      PP.DiscardUntilEndOfDirective();
829      return true;
830    }
831    Val = NextVal;
832  }
833
834  if (DigitTokBegin[0] == '0' && Val)
835    PP.Diag(DigitTok.getLocation(), diag::warn_pp_line_decimal);
836
837  return false;
838}
839
840/// \brief Handle a \#line directive: C99 6.10.4.
841///
842/// The two acceptable forms are:
843/// \verbatim
844///   # line digit-sequence
845///   # line digit-sequence "s-char-sequence"
846/// \endverbatim
847void Preprocessor::HandleLineDirective(Token &Tok) {
848  // Read the line # and string argument.  Per C99 6.10.4p5, these tokens are
849  // expanded.
850  Token DigitTok;
851  Lex(DigitTok);
852
853  // Validate the number and convert it to an unsigned.
854  unsigned LineNo;
855  if (GetLineValue(DigitTok, LineNo, diag::err_pp_line_requires_integer,*this))
856    return;
857
858  if (LineNo == 0)
859    Diag(DigitTok, diag::ext_pp_line_zero);
860
861  // Enforce C99 6.10.4p3: "The digit sequence shall not specify ... a
862  // number greater than 2147483647".  C90 requires that the line # be <= 32767.
863  unsigned LineLimit = 32768U;
864  if (LangOpts.C99 || LangOpts.CPlusPlus11)
865    LineLimit = 2147483648U;
866  if (LineNo >= LineLimit)
867    Diag(DigitTok, diag::ext_pp_line_too_big) << LineLimit;
868  else if (LangOpts.CPlusPlus11 && LineNo >= 32768U)
869    Diag(DigitTok, diag::warn_cxx98_compat_pp_line_too_big);
870
871  int FilenameID = -1;
872  Token StrTok;
873  Lex(StrTok);
874
875  // If the StrTok is "eod", then it wasn't present.  Otherwise, it must be a
876  // string followed by eod.
877  if (StrTok.is(tok::eod))
878    ; // ok
879  else if (StrTok.isNot(tok::string_literal)) {
880    Diag(StrTok, diag::err_pp_line_invalid_filename);
881    return DiscardUntilEndOfDirective();
882  } else if (StrTok.hasUDSuffix()) {
883    Diag(StrTok, diag::err_invalid_string_udl);
884    return DiscardUntilEndOfDirective();
885  } else {
886    // Parse and validate the string, converting it into a unique ID.
887    StringLiteralParser Literal(&StrTok, 1, *this);
888    assert(Literal.isAscii() && "Didn't allow wide strings in");
889    if (Literal.hadError)
890      return DiscardUntilEndOfDirective();
891    if (Literal.Pascal) {
892      Diag(StrTok, diag::err_pp_linemarker_invalid_filename);
893      return DiscardUntilEndOfDirective();
894    }
895    FilenameID = SourceMgr.getLineTableFilenameID(Literal.GetString());
896
897    // Verify that there is nothing after the string, other than EOD.  Because
898    // of C99 6.10.4p5, macros that expand to empty tokens are ok.
899    CheckEndOfDirective("line", true);
900  }
901
902  SourceMgr.AddLineNote(DigitTok.getLocation(), LineNo, FilenameID);
903
904  if (Callbacks)
905    Callbacks->FileChanged(CurPPLexer->getSourceLocation(),
906                           PPCallbacks::RenameFile,
907                           SrcMgr::C_User);
908}
909
910/// ReadLineMarkerFlags - Parse and validate any flags at the end of a GNU line
911/// marker directive.
912static bool ReadLineMarkerFlags(bool &IsFileEntry, bool &IsFileExit,
913                                bool &IsSystemHeader, bool &IsExternCHeader,
914                                Preprocessor &PP) {
915  unsigned FlagVal;
916  Token FlagTok;
917  PP.Lex(FlagTok);
918  if (FlagTok.is(tok::eod)) return false;
919  if (GetLineValue(FlagTok, FlagVal, diag::err_pp_linemarker_invalid_flag, PP))
920    return true;
921
922  if (FlagVal == 1) {
923    IsFileEntry = true;
924
925    PP.Lex(FlagTok);
926    if (FlagTok.is(tok::eod)) return false;
927    if (GetLineValue(FlagTok, FlagVal, diag::err_pp_linemarker_invalid_flag,PP))
928      return true;
929  } else if (FlagVal == 2) {
930    IsFileExit = true;
931
932    SourceManager &SM = PP.getSourceManager();
933    // If we are leaving the current presumed file, check to make sure the
934    // presumed include stack isn't empty!
935    FileID CurFileID =
936      SM.getDecomposedExpansionLoc(FlagTok.getLocation()).first;
937    PresumedLoc PLoc = SM.getPresumedLoc(FlagTok.getLocation());
938    if (PLoc.isInvalid())
939      return true;
940
941    // If there is no include loc (main file) or if the include loc is in a
942    // different physical file, then we aren't in a "1" line marker flag region.
943    SourceLocation IncLoc = PLoc.getIncludeLoc();
944    if (IncLoc.isInvalid() ||
945        SM.getDecomposedExpansionLoc(IncLoc).first != CurFileID) {
946      PP.Diag(FlagTok, diag::err_pp_linemarker_invalid_pop);
947      PP.DiscardUntilEndOfDirective();
948      return true;
949    }
950
951    PP.Lex(FlagTok);
952    if (FlagTok.is(tok::eod)) return false;
953    if (GetLineValue(FlagTok, FlagVal, diag::err_pp_linemarker_invalid_flag,PP))
954      return true;
955  }
956
957  // We must have 3 if there are still flags.
958  if (FlagVal != 3) {
959    PP.Diag(FlagTok, diag::err_pp_linemarker_invalid_flag);
960    PP.DiscardUntilEndOfDirective();
961    return true;
962  }
963
964  IsSystemHeader = true;
965
966  PP.Lex(FlagTok);
967  if (FlagTok.is(tok::eod)) return false;
968  if (GetLineValue(FlagTok, FlagVal, diag::err_pp_linemarker_invalid_flag, PP))
969    return true;
970
971  // We must have 4 if there is yet another flag.
972  if (FlagVal != 4) {
973    PP.Diag(FlagTok, diag::err_pp_linemarker_invalid_flag);
974    PP.DiscardUntilEndOfDirective();
975    return true;
976  }
977
978  IsExternCHeader = true;
979
980  PP.Lex(FlagTok);
981  if (FlagTok.is(tok::eod)) return false;
982
983  // There are no more valid flags here.
984  PP.Diag(FlagTok, diag::err_pp_linemarker_invalid_flag);
985  PP.DiscardUntilEndOfDirective();
986  return true;
987}
988
989/// HandleDigitDirective - Handle a GNU line marker directive, whose syntax is
990/// one of the following forms:
991///
992///     # 42
993///     # 42 "file" ('1' | '2')?
994///     # 42 "file" ('1' | '2')? '3' '4'?
995///
996void Preprocessor::HandleDigitDirective(Token &DigitTok) {
997  // Validate the number and convert it to an unsigned.  GNU does not have a
998  // line # limit other than it fit in 32-bits.
999  unsigned LineNo;
1000  if (GetLineValue(DigitTok, LineNo, diag::err_pp_linemarker_requires_integer,
1001                   *this))
1002    return;
1003
1004  Token StrTok;
1005  Lex(StrTok);
1006
1007  bool IsFileEntry = false, IsFileExit = false;
1008  bool IsSystemHeader = false, IsExternCHeader = false;
1009  int FilenameID = -1;
1010
1011  // If the StrTok is "eod", then it wasn't present.  Otherwise, it must be a
1012  // string followed by eod.
1013  if (StrTok.is(tok::eod))
1014    ; // ok
1015  else if (StrTok.isNot(tok::string_literal)) {
1016    Diag(StrTok, diag::err_pp_linemarker_invalid_filename);
1017    return DiscardUntilEndOfDirective();
1018  } else if (StrTok.hasUDSuffix()) {
1019    Diag(StrTok, diag::err_invalid_string_udl);
1020    return DiscardUntilEndOfDirective();
1021  } else {
1022    // Parse and validate the string, converting it into a unique ID.
1023    StringLiteralParser Literal(&StrTok, 1, *this);
1024    assert(Literal.isAscii() && "Didn't allow wide strings in");
1025    if (Literal.hadError)
1026      return DiscardUntilEndOfDirective();
1027    if (Literal.Pascal) {
1028      Diag(StrTok, diag::err_pp_linemarker_invalid_filename);
1029      return DiscardUntilEndOfDirective();
1030    }
1031    FilenameID = SourceMgr.getLineTableFilenameID(Literal.GetString());
1032
1033    // If a filename was present, read any flags that are present.
1034    if (ReadLineMarkerFlags(IsFileEntry, IsFileExit,
1035                            IsSystemHeader, IsExternCHeader, *this))
1036      return;
1037  }
1038
1039  // Create a line note with this information.
1040  SourceMgr.AddLineNote(DigitTok.getLocation(), LineNo, FilenameID,
1041                        IsFileEntry, IsFileExit,
1042                        IsSystemHeader, IsExternCHeader);
1043
1044  // If the preprocessor has callbacks installed, notify them of the #line
1045  // change.  This is used so that the line marker comes out in -E mode for
1046  // example.
1047  if (Callbacks) {
1048    PPCallbacks::FileChangeReason Reason = PPCallbacks::RenameFile;
1049    if (IsFileEntry)
1050      Reason = PPCallbacks::EnterFile;
1051    else if (IsFileExit)
1052      Reason = PPCallbacks::ExitFile;
1053    SrcMgr::CharacteristicKind FileKind = SrcMgr::C_User;
1054    if (IsExternCHeader)
1055      FileKind = SrcMgr::C_ExternCSystem;
1056    else if (IsSystemHeader)
1057      FileKind = SrcMgr::C_System;
1058
1059    Callbacks->FileChanged(CurPPLexer->getSourceLocation(), Reason, FileKind);
1060  }
1061}
1062
1063
1064/// HandleUserDiagnosticDirective - Handle a #warning or #error directive.
1065///
1066void Preprocessor::HandleUserDiagnosticDirective(Token &Tok,
1067                                                 bool isWarning) {
1068  // PTH doesn't emit #warning or #error directives.
1069  if (CurPTHLexer)
1070    return CurPTHLexer->DiscardToEndOfLine();
1071
1072  // Read the rest of the line raw.  We do this because we don't want macros
1073  // to be expanded and we don't require that the tokens be valid preprocessing
1074  // tokens.  For example, this is allowed: "#warning `   'foo".  GCC does
1075  // collapse multiple consequtive white space between tokens, but this isn't
1076  // specified by the standard.
1077  SmallString<128> Message;
1078  CurLexer->ReadToEndOfLine(&Message);
1079
1080  // Find the first non-whitespace character, so that we can make the
1081  // diagnostic more succinct.
1082  StringRef Msg = Message.str().ltrim(" ");
1083
1084  if (isWarning)
1085    Diag(Tok, diag::pp_hash_warning) << Msg;
1086  else
1087    Diag(Tok, diag::err_pp_hash_error) << Msg;
1088}
1089
1090/// HandleIdentSCCSDirective - Handle a #ident/#sccs directive.
1091///
1092void Preprocessor::HandleIdentSCCSDirective(Token &Tok) {
1093  // Yes, this directive is an extension.
1094  Diag(Tok, diag::ext_pp_ident_directive);
1095
1096  // Read the string argument.
1097  Token StrTok;
1098  Lex(StrTok);
1099
1100  // If the token kind isn't a string, it's a malformed directive.
1101  if (StrTok.isNot(tok::string_literal) &&
1102      StrTok.isNot(tok::wide_string_literal)) {
1103    Diag(StrTok, diag::err_pp_malformed_ident);
1104    if (StrTok.isNot(tok::eod))
1105      DiscardUntilEndOfDirective();
1106    return;
1107  }
1108
1109  if (StrTok.hasUDSuffix()) {
1110    Diag(StrTok, diag::err_invalid_string_udl);
1111    return DiscardUntilEndOfDirective();
1112  }
1113
1114  // Verify that there is nothing after the string, other than EOD.
1115  CheckEndOfDirective("ident");
1116
1117  if (Callbacks) {
1118    bool Invalid = false;
1119    std::string Str = getSpelling(StrTok, &Invalid);
1120    if (!Invalid)
1121      Callbacks->Ident(Tok.getLocation(), Str);
1122  }
1123}
1124
1125/// \brief Handle a #public directive.
1126void Preprocessor::HandleMacroPublicDirective(Token &Tok) {
1127  Token MacroNameTok;
1128  ReadMacroName(MacroNameTok, 2);
1129
1130  // Error reading macro name?  If so, diagnostic already issued.
1131  if (MacroNameTok.is(tok::eod))
1132    return;
1133
1134  // Check to see if this is the last token on the #__public_macro line.
1135  CheckEndOfDirective("__public_macro");
1136
1137  IdentifierInfo *II = MacroNameTok.getIdentifierInfo();
1138  // Okay, we finally have a valid identifier to undef.
1139  MacroDirective *MD = getMacroDirective(II);
1140
1141  // If the macro is not defined, this is an error.
1142  if (MD == 0) {
1143    Diag(MacroNameTok, diag::err_pp_visibility_non_macro) << II;
1144    return;
1145  }
1146
1147  // Note that this macro has now been exported.
1148  appendMacroDirective(II, AllocateVisibilityMacroDirective(
1149                                MacroNameTok.getLocation(), /*IsPublic=*/true));
1150}
1151
1152/// \brief Handle a #private directive.
1153void Preprocessor::HandleMacroPrivateDirective(Token &Tok) {
1154  Token MacroNameTok;
1155  ReadMacroName(MacroNameTok, 2);
1156
1157  // Error reading macro name?  If so, diagnostic already issued.
1158  if (MacroNameTok.is(tok::eod))
1159    return;
1160
1161  // Check to see if this is the last token on the #__private_macro line.
1162  CheckEndOfDirective("__private_macro");
1163
1164  IdentifierInfo *II = MacroNameTok.getIdentifierInfo();
1165  // Okay, we finally have a valid identifier to undef.
1166  MacroDirective *MD = getMacroDirective(II);
1167
1168  // If the macro is not defined, this is an error.
1169  if (MD == 0) {
1170    Diag(MacroNameTok, diag::err_pp_visibility_non_macro) << II;
1171    return;
1172  }
1173
1174  // Note that this macro has now been marked private.
1175  appendMacroDirective(II, AllocateVisibilityMacroDirective(
1176                               MacroNameTok.getLocation(), /*IsPublic=*/false));
1177}
1178
1179//===----------------------------------------------------------------------===//
1180// Preprocessor Include Directive Handling.
1181//===----------------------------------------------------------------------===//
1182
1183/// GetIncludeFilenameSpelling - Turn the specified lexer token into a fully
1184/// checked and spelled filename, e.g. as an operand of \#include. This returns
1185/// true if the input filename was in <>'s or false if it were in ""'s.  The
1186/// caller is expected to provide a buffer that is large enough to hold the
1187/// spelling of the filename, but is also expected to handle the case when
1188/// this method decides to use a different buffer.
1189bool Preprocessor::GetIncludeFilenameSpelling(SourceLocation Loc,
1190                                              StringRef &Buffer) {
1191  // Get the text form of the filename.
1192  assert(!Buffer.empty() && "Can't have tokens with empty spellings!");
1193
1194  // Make sure the filename is <x> or "x".
1195  bool isAngled;
1196  if (Buffer[0] == '<') {
1197    if (Buffer.back() != '>') {
1198      Diag(Loc, diag::err_pp_expects_filename);
1199      Buffer = StringRef();
1200      return true;
1201    }
1202    isAngled = true;
1203  } else if (Buffer[0] == '"') {
1204    if (Buffer.back() != '"') {
1205      Diag(Loc, diag::err_pp_expects_filename);
1206      Buffer = StringRef();
1207      return true;
1208    }
1209    isAngled = false;
1210  } else {
1211    Diag(Loc, diag::err_pp_expects_filename);
1212    Buffer = StringRef();
1213    return true;
1214  }
1215
1216  // Diagnose #include "" as invalid.
1217  if (Buffer.size() <= 2) {
1218    Diag(Loc, diag::err_pp_empty_filename);
1219    Buffer = StringRef();
1220    return true;
1221  }
1222
1223  // Skip the brackets.
1224  Buffer = Buffer.substr(1, Buffer.size()-2);
1225  return isAngled;
1226}
1227
1228/// \brief Handle cases where the \#include name is expanded from a macro
1229/// as multiple tokens, which need to be glued together.
1230///
1231/// This occurs for code like:
1232/// \code
1233///    \#define FOO <a/b.h>
1234///    \#include FOO
1235/// \endcode
1236/// because in this case, "<a/b.h>" is returned as 7 tokens, not one.
1237///
1238/// This code concatenates and consumes tokens up to the '>' token.  It returns
1239/// false if the > was found, otherwise it returns true if it finds and consumes
1240/// the EOD marker.
1241bool Preprocessor::ConcatenateIncludeName(
1242                                        SmallString<128> &FilenameBuffer,
1243                                          SourceLocation &End) {
1244  Token CurTok;
1245
1246  Lex(CurTok);
1247  while (CurTok.isNot(tok::eod)) {
1248    End = CurTok.getLocation();
1249
1250    // FIXME: Provide code completion for #includes.
1251    if (CurTok.is(tok::code_completion)) {
1252      setCodeCompletionReached();
1253      Lex(CurTok);
1254      continue;
1255    }
1256
1257    // Append the spelling of this token to the buffer. If there was a space
1258    // before it, add it now.
1259    if (CurTok.hasLeadingSpace())
1260      FilenameBuffer.push_back(' ');
1261
1262    // Get the spelling of the token, directly into FilenameBuffer if possible.
1263    unsigned PreAppendSize = FilenameBuffer.size();
1264    FilenameBuffer.resize(PreAppendSize+CurTok.getLength());
1265
1266    const char *BufPtr = &FilenameBuffer[PreAppendSize];
1267    unsigned ActualLen = getSpelling(CurTok, BufPtr);
1268
1269    // If the token was spelled somewhere else, copy it into FilenameBuffer.
1270    if (BufPtr != &FilenameBuffer[PreAppendSize])
1271      memcpy(&FilenameBuffer[PreAppendSize], BufPtr, ActualLen);
1272
1273    // Resize FilenameBuffer to the correct size.
1274    if (CurTok.getLength() != ActualLen)
1275      FilenameBuffer.resize(PreAppendSize+ActualLen);
1276
1277    // If we found the '>' marker, return success.
1278    if (CurTok.is(tok::greater))
1279      return false;
1280
1281    Lex(CurTok);
1282  }
1283
1284  // If we hit the eod marker, emit an error and return true so that the caller
1285  // knows the EOD has been read.
1286  Diag(CurTok.getLocation(), diag::err_pp_expects_filename);
1287  return true;
1288}
1289
1290/// HandleIncludeDirective - The "\#include" tokens have just been read, read
1291/// the file to be included from the lexer, then include it!  This is a common
1292/// routine with functionality shared between \#include, \#include_next and
1293/// \#import.  LookupFrom is set when this is a \#include_next directive, it
1294/// specifies the file to start searching from.
1295void Preprocessor::HandleIncludeDirective(SourceLocation HashLoc,
1296                                          Token &IncludeTok,
1297                                          const DirectoryLookup *LookupFrom,
1298                                          bool isImport) {
1299
1300  Token FilenameTok;
1301  CurPPLexer->LexIncludeFilename(FilenameTok);
1302
1303  // Reserve a buffer to get the spelling.
1304  SmallString<128> FilenameBuffer;
1305  StringRef Filename;
1306  SourceLocation End;
1307  SourceLocation CharEnd; // the end of this directive, in characters
1308
1309  switch (FilenameTok.getKind()) {
1310  case tok::eod:
1311    // If the token kind is EOD, the error has already been diagnosed.
1312    return;
1313
1314  case tok::angle_string_literal:
1315  case tok::string_literal:
1316    Filename = getSpelling(FilenameTok, FilenameBuffer);
1317    End = FilenameTok.getLocation();
1318    CharEnd = End.getLocWithOffset(FilenameTok.getLength());
1319    break;
1320
1321  case tok::less:
1322    // This could be a <foo/bar.h> file coming from a macro expansion.  In this
1323    // case, glue the tokens together into FilenameBuffer and interpret those.
1324    FilenameBuffer.push_back('<');
1325    if (ConcatenateIncludeName(FilenameBuffer, End))
1326      return;   // Found <eod> but no ">"?  Diagnostic already emitted.
1327    Filename = FilenameBuffer.str();
1328    CharEnd = End.getLocWithOffset(1);
1329    break;
1330  default:
1331    Diag(FilenameTok.getLocation(), diag::err_pp_expects_filename);
1332    DiscardUntilEndOfDirective();
1333    return;
1334  }
1335
1336  CharSourceRange FilenameRange
1337    = CharSourceRange::getCharRange(FilenameTok.getLocation(), CharEnd);
1338  StringRef OriginalFilename = Filename;
1339  bool isAngled =
1340    GetIncludeFilenameSpelling(FilenameTok.getLocation(), Filename);
1341  // If GetIncludeFilenameSpelling set the start ptr to null, there was an
1342  // error.
1343  if (Filename.empty()) {
1344    DiscardUntilEndOfDirective();
1345    return;
1346  }
1347
1348  // Verify that there is nothing after the filename, other than EOD.  Note that
1349  // we allow macros that expand to nothing after the filename, because this
1350  // falls into the category of "#include pp-tokens new-line" specified in
1351  // C99 6.10.2p4.
1352  CheckEndOfDirective(IncludeTok.getIdentifierInfo()->getNameStart(), true);
1353
1354  // Check that we don't have infinite #include recursion.
1355  if (IncludeMacroStack.size() == MaxAllowedIncludeStackDepth-1) {
1356    Diag(FilenameTok, diag::err_pp_include_too_deep);
1357    return;
1358  }
1359
1360  // Complain about attempts to #include files in an audit pragma.
1361  if (PragmaARCCFCodeAuditedLoc.isValid()) {
1362    Diag(HashLoc, diag::err_pp_include_in_arc_cf_code_audited);
1363    Diag(PragmaARCCFCodeAuditedLoc, diag::note_pragma_entered_here);
1364
1365    // Immediately leave the pragma.
1366    PragmaARCCFCodeAuditedLoc = SourceLocation();
1367  }
1368
1369  if (HeaderInfo.HasIncludeAliasMap()) {
1370    // Map the filename with the brackets still attached.  If the name doesn't
1371    // map to anything, fall back on the filename we've already gotten the
1372    // spelling for.
1373    StringRef NewName = HeaderInfo.MapHeaderToIncludeAlias(OriginalFilename);
1374    if (!NewName.empty())
1375      Filename = NewName;
1376  }
1377
1378  // Search include directories.
1379  const DirectoryLookup *CurDir;
1380  SmallString<1024> SearchPath;
1381  SmallString<1024> RelativePath;
1382  // We get the raw path only if we have 'Callbacks' to which we later pass
1383  // the path.
1384  Module *SuggestedModule = 0;
1385  const FileEntry *File = LookupFile(
1386      Filename, isAngled, LookupFrom, CurDir,
1387      Callbacks ? &SearchPath : NULL, Callbacks ? &RelativePath : NULL,
1388      getLangOpts().Modules? &SuggestedModule : 0);
1389
1390  if (Callbacks) {
1391    if (!File) {
1392      // Give the clients a chance to recover.
1393      SmallString<128> RecoveryPath;
1394      if (Callbacks->FileNotFound(Filename, RecoveryPath)) {
1395        if (const DirectoryEntry *DE = FileMgr.getDirectory(RecoveryPath)) {
1396          // Add the recovery path to the list of search paths.
1397          DirectoryLookup DL(DE, SrcMgr::C_User, false);
1398          HeaderInfo.AddSearchPath(DL, isAngled);
1399
1400          // Try the lookup again, skipping the cache.
1401          File = LookupFile(Filename, isAngled, LookupFrom, CurDir, 0, 0,
1402                            getLangOpts().Modules? &SuggestedModule : 0,
1403                            /*SkipCache*/true);
1404        }
1405      }
1406    }
1407
1408    if (!SuggestedModule) {
1409      // Notify the callback object that we've seen an inclusion directive.
1410      Callbacks->InclusionDirective(HashLoc, IncludeTok, Filename, isAngled,
1411                                    FilenameRange, File,
1412                                    SearchPath, RelativePath,
1413                                    /*ImportedModule=*/0);
1414    }
1415  }
1416
1417  if (File == 0) {
1418    if (!SuppressIncludeNotFoundError) {
1419      // If the file could not be located and it was included via angle
1420      // brackets, we can attempt a lookup as though it were a quoted path to
1421      // provide the user with a possible fixit.
1422      if (isAngled) {
1423        File = LookupFile(Filename, false, LookupFrom, CurDir,
1424                          Callbacks ? &SearchPath : 0,
1425                          Callbacks ? &RelativePath : 0,
1426                          getLangOpts().Modules ? &SuggestedModule : 0);
1427        if (File) {
1428          SourceRange Range(FilenameTok.getLocation(), CharEnd);
1429          Diag(FilenameTok, diag::err_pp_file_not_found_not_fatal) <<
1430            Filename <<
1431            FixItHint::CreateReplacement(Range, "\"" + Filename.str() + "\"");
1432        }
1433      }
1434      // If the file is still not found, just go with the vanilla diagnostic
1435      if (!File)
1436        Diag(FilenameTok, diag::err_pp_file_not_found) << Filename;
1437    }
1438    if (!File)
1439      return;
1440  }
1441
1442  // If we are supposed to import a module rather than including the header,
1443  // do so now.
1444  if (SuggestedModule) {
1445    // Compute the module access path corresponding to this module.
1446    // FIXME: Should we have a second loadModule() overload to avoid this
1447    // extra lookup step?
1448    SmallVector<std::pair<IdentifierInfo *, SourceLocation>, 2> Path;
1449    for (Module *Mod = SuggestedModule; Mod; Mod = Mod->Parent)
1450      Path.push_back(std::make_pair(getIdentifierInfo(Mod->Name),
1451                                    FilenameTok.getLocation()));
1452    std::reverse(Path.begin(), Path.end());
1453
1454    // Warn that we're replacing the include/import with a module import.
1455    SmallString<128> PathString;
1456    for (unsigned I = 0, N = Path.size(); I != N; ++I) {
1457      if (I)
1458        PathString += '.';
1459      PathString += Path[I].first->getName();
1460    }
1461    int IncludeKind = 0;
1462
1463    switch (IncludeTok.getIdentifierInfo()->getPPKeywordID()) {
1464    case tok::pp_include:
1465      IncludeKind = 0;
1466      break;
1467
1468    case tok::pp_import:
1469      IncludeKind = 1;
1470      break;
1471
1472    case tok::pp_include_next:
1473      IncludeKind = 2;
1474      break;
1475
1476    case tok::pp___include_macros:
1477      IncludeKind = 3;
1478      break;
1479
1480    default:
1481      llvm_unreachable("unknown include directive kind");
1482    }
1483
1484    // Determine whether we are actually building the module that this
1485    // include directive maps to.
1486    bool BuildingImportedModule
1487      = Path[0].first->getName() == getLangOpts().CurrentModule;
1488
1489    if (!BuildingImportedModule && getLangOpts().ObjC2) {
1490      // If we're not building the imported module, warn that we're going
1491      // to automatically turn this inclusion directive into a module import.
1492      // We only do this in Objective-C, where we have a module-import syntax.
1493      CharSourceRange ReplaceRange(SourceRange(HashLoc, CharEnd),
1494                                   /*IsTokenRange=*/false);
1495      Diag(HashLoc, diag::warn_auto_module_import)
1496        << IncludeKind << PathString
1497        << FixItHint::CreateReplacement(ReplaceRange,
1498             "@import " + PathString.str().str() + ";");
1499    }
1500
1501    // Load the module.
1502    // If this was an #__include_macros directive, only make macros visible.
1503    Module::NameVisibilityKind Visibility
1504      = (IncludeKind == 3)? Module::MacrosVisible : Module::AllVisible;
1505    ModuleLoadResult Imported
1506      = TheModuleLoader.loadModule(IncludeTok.getLocation(), Path, Visibility,
1507                                   /*IsIncludeDirective=*/true);
1508    assert((Imported == 0 || Imported == SuggestedModule) &&
1509           "the imported module is different than the suggested one");
1510
1511    // If this header isn't part of the module we're building, we're done.
1512    if (!BuildingImportedModule && Imported) {
1513      if (Callbacks) {
1514        Callbacks->InclusionDirective(HashLoc, IncludeTok, Filename, isAngled,
1515                                      FilenameRange, File,
1516                                      SearchPath, RelativePath, Imported);
1517      }
1518      return;
1519    }
1520
1521    // If we failed to find a submodule that we expected to find, we can
1522    // continue. Otherwise, there's an error in the included file, so we
1523    // don't want to include it.
1524    if (!BuildingImportedModule && !Imported.isMissingExpected()) {
1525      return;
1526    }
1527  }
1528
1529  if (Callbacks && SuggestedModule) {
1530    // We didn't notify the callback object that we've seen an inclusion
1531    // directive before. Now that we are parsing the include normally and not
1532    // turning it to a module import, notify the callback object.
1533    Callbacks->InclusionDirective(HashLoc, IncludeTok, Filename, isAngled,
1534                                  FilenameRange, File,
1535                                  SearchPath, RelativePath,
1536                                  /*ImportedModule=*/0);
1537  }
1538
1539  // The #included file will be considered to be a system header if either it is
1540  // in a system include directory, or if the #includer is a system include
1541  // header.
1542  SrcMgr::CharacteristicKind FileCharacter =
1543    std::max(HeaderInfo.getFileDirFlavor(File),
1544             SourceMgr.getFileCharacteristic(FilenameTok.getLocation()));
1545
1546  // Ask HeaderInfo if we should enter this #include file.  If not, #including
1547  // this file will have no effect.
1548  if (!HeaderInfo.ShouldEnterIncludeFile(File, isImport)) {
1549    if (Callbacks)
1550      Callbacks->FileSkipped(*File, FilenameTok, FileCharacter);
1551    return;
1552  }
1553
1554  // Look up the file, create a File ID for it.
1555  SourceLocation IncludePos = End;
1556  // If the filename string was the result of macro expansions, set the include
1557  // position on the file where it will be included and after the expansions.
1558  if (IncludePos.isMacroID())
1559    IncludePos = SourceMgr.getExpansionRange(IncludePos).second;
1560  FileID FID = SourceMgr.createFileID(File, IncludePos, FileCharacter);
1561  assert(!FID.isInvalid() && "Expected valid file ID");
1562
1563  // Finally, if all is good, enter the new file!
1564  EnterSourceFile(FID, CurDir, FilenameTok.getLocation());
1565}
1566
1567/// HandleIncludeNextDirective - Implements \#include_next.
1568///
1569void Preprocessor::HandleIncludeNextDirective(SourceLocation HashLoc,
1570                                              Token &IncludeNextTok) {
1571  Diag(IncludeNextTok, diag::ext_pp_include_next_directive);
1572
1573  // #include_next is like #include, except that we start searching after
1574  // the current found directory.  If we can't do this, issue a
1575  // diagnostic.
1576  const DirectoryLookup *Lookup = CurDirLookup;
1577  if (isInPrimaryFile()) {
1578    Lookup = 0;
1579    Diag(IncludeNextTok, diag::pp_include_next_in_primary);
1580  } else if (Lookup == 0) {
1581    Diag(IncludeNextTok, diag::pp_include_next_absolute_path);
1582  } else {
1583    // Start looking up in the next directory.
1584    ++Lookup;
1585  }
1586
1587  return HandleIncludeDirective(HashLoc, IncludeNextTok, Lookup);
1588}
1589
1590/// HandleMicrosoftImportDirective - Implements \#import for Microsoft Mode
1591void Preprocessor::HandleMicrosoftImportDirective(Token &Tok) {
1592  // The Microsoft #import directive takes a type library and generates header
1593  // files from it, and includes those.  This is beyond the scope of what clang
1594  // does, so we ignore it and error out.  However, #import can optionally have
1595  // trailing attributes that span multiple lines.  We're going to eat those
1596  // so we can continue processing from there.
1597  Diag(Tok, diag::err_pp_import_directive_ms );
1598
1599  // Read tokens until we get to the end of the directive.  Note that the
1600  // directive can be split over multiple lines using the backslash character.
1601  DiscardUntilEndOfDirective();
1602}
1603
1604/// HandleImportDirective - Implements \#import.
1605///
1606void Preprocessor::HandleImportDirective(SourceLocation HashLoc,
1607                                         Token &ImportTok) {
1608  if (!LangOpts.ObjC1) {  // #import is standard for ObjC.
1609    if (LangOpts.MicrosoftMode)
1610      return HandleMicrosoftImportDirective(ImportTok);
1611    Diag(ImportTok, diag::ext_pp_import_directive);
1612  }
1613  return HandleIncludeDirective(HashLoc, ImportTok, 0, true);
1614}
1615
1616/// HandleIncludeMacrosDirective - The -imacros command line option turns into a
1617/// pseudo directive in the predefines buffer.  This handles it by sucking all
1618/// tokens through the preprocessor and discarding them (only keeping the side
1619/// effects on the preprocessor).
1620void Preprocessor::HandleIncludeMacrosDirective(SourceLocation HashLoc,
1621                                                Token &IncludeMacrosTok) {
1622  // This directive should only occur in the predefines buffer.  If not, emit an
1623  // error and reject it.
1624  SourceLocation Loc = IncludeMacrosTok.getLocation();
1625  if (strcmp(SourceMgr.getBufferName(Loc), "<built-in>") != 0) {
1626    Diag(IncludeMacrosTok.getLocation(),
1627         diag::pp_include_macros_out_of_predefines);
1628    DiscardUntilEndOfDirective();
1629    return;
1630  }
1631
1632  // Treat this as a normal #include for checking purposes.  If this is
1633  // successful, it will push a new lexer onto the include stack.
1634  HandleIncludeDirective(HashLoc, IncludeMacrosTok, 0, false);
1635
1636  Token TmpTok;
1637  do {
1638    Lex(TmpTok);
1639    assert(TmpTok.isNot(tok::eof) && "Didn't find end of -imacros!");
1640  } while (TmpTok.isNot(tok::hashhash));
1641}
1642
1643//===----------------------------------------------------------------------===//
1644// Preprocessor Macro Directive Handling.
1645//===----------------------------------------------------------------------===//
1646
1647/// ReadMacroDefinitionArgList - The ( starting an argument list of a macro
1648/// definition has just been read.  Lex the rest of the arguments and the
1649/// closing ), updating MI with what we learn.  Return true if an error occurs
1650/// parsing the arg list.
1651bool Preprocessor::ReadMacroDefinitionArgList(MacroInfo *MI, Token &Tok) {
1652  SmallVector<IdentifierInfo*, 32> Arguments;
1653
1654  while (1) {
1655    LexUnexpandedToken(Tok);
1656    switch (Tok.getKind()) {
1657    case tok::r_paren:
1658      // Found the end of the argument list.
1659      if (Arguments.empty())  // #define FOO()
1660        return false;
1661      // Otherwise we have #define FOO(A,)
1662      Diag(Tok, diag::err_pp_expected_ident_in_arg_list);
1663      return true;
1664    case tok::ellipsis:  // #define X(... -> C99 varargs
1665      if (!LangOpts.C99)
1666        Diag(Tok, LangOpts.CPlusPlus11 ?
1667             diag::warn_cxx98_compat_variadic_macro :
1668             diag::ext_variadic_macro);
1669
1670      // OpenCL v1.2 s6.9.e: variadic macros are not supported.
1671      if (LangOpts.OpenCL) {
1672        Diag(Tok, diag::err_pp_opencl_variadic_macros);
1673        return true;
1674      }
1675
1676      // Lex the token after the identifier.
1677      LexUnexpandedToken(Tok);
1678      if (Tok.isNot(tok::r_paren)) {
1679        Diag(Tok, diag::err_pp_missing_rparen_in_macro_def);
1680        return true;
1681      }
1682      // Add the __VA_ARGS__ identifier as an argument.
1683      Arguments.push_back(Ident__VA_ARGS__);
1684      MI->setIsC99Varargs();
1685      MI->setArgumentList(&Arguments[0], Arguments.size(), BP);
1686      return false;
1687    case tok::eod:  // #define X(
1688      Diag(Tok, diag::err_pp_missing_rparen_in_macro_def);
1689      return true;
1690    default:
1691      // Handle keywords and identifiers here to accept things like
1692      // #define Foo(for) for.
1693      IdentifierInfo *II = Tok.getIdentifierInfo();
1694      if (II == 0) {
1695        // #define X(1
1696        Diag(Tok, diag::err_pp_invalid_tok_in_arg_list);
1697        return true;
1698      }
1699
1700      // If this is already used as an argument, it is used multiple times (e.g.
1701      // #define X(A,A.
1702      if (std::find(Arguments.begin(), Arguments.end(), II) !=
1703          Arguments.end()) {  // C99 6.10.3p6
1704        Diag(Tok, diag::err_pp_duplicate_name_in_arg_list) << II;
1705        return true;
1706      }
1707
1708      // Add the argument to the macro info.
1709      Arguments.push_back(II);
1710
1711      // Lex the token after the identifier.
1712      LexUnexpandedToken(Tok);
1713
1714      switch (Tok.getKind()) {
1715      default:          // #define X(A B
1716        Diag(Tok, diag::err_pp_expected_comma_in_arg_list);
1717        return true;
1718      case tok::r_paren: // #define X(A)
1719        MI->setArgumentList(&Arguments[0], Arguments.size(), BP);
1720        return false;
1721      case tok::comma:  // #define X(A,
1722        break;
1723      case tok::ellipsis:  // #define X(A... -> GCC extension
1724        // Diagnose extension.
1725        Diag(Tok, diag::ext_named_variadic_macro);
1726
1727        // Lex the token after the identifier.
1728        LexUnexpandedToken(Tok);
1729        if (Tok.isNot(tok::r_paren)) {
1730          Diag(Tok, diag::err_pp_missing_rparen_in_macro_def);
1731          return true;
1732        }
1733
1734        MI->setIsGNUVarargs();
1735        MI->setArgumentList(&Arguments[0], Arguments.size(), BP);
1736        return false;
1737      }
1738    }
1739  }
1740}
1741
1742/// HandleDefineDirective - Implements \#define.  This consumes the entire macro
1743/// line then lets the caller lex the next real token.
1744void Preprocessor::HandleDefineDirective(Token &DefineTok) {
1745  ++NumDefined;
1746
1747  Token MacroNameTok;
1748  ReadMacroName(MacroNameTok, 1);
1749
1750  // Error reading macro name?  If so, diagnostic already issued.
1751  if (MacroNameTok.is(tok::eod))
1752    return;
1753
1754  Token LastTok = MacroNameTok;
1755
1756  // If we are supposed to keep comments in #defines, reenable comment saving
1757  // mode.
1758  if (CurLexer) CurLexer->SetCommentRetentionState(KeepMacroComments);
1759
1760  // Create the new macro.
1761  MacroInfo *MI = AllocateMacroInfo(MacroNameTok.getLocation());
1762
1763  Token Tok;
1764  LexUnexpandedToken(Tok);
1765
1766  // If this is a function-like macro definition, parse the argument list,
1767  // marking each of the identifiers as being used as macro arguments.  Also,
1768  // check other constraints on the first token of the macro body.
1769  if (Tok.is(tok::eod)) {
1770    // If there is no body to this macro, we have no special handling here.
1771  } else if (Tok.hasLeadingSpace()) {
1772    // This is a normal token with leading space.  Clear the leading space
1773    // marker on the first token to get proper expansion.
1774    Tok.clearFlag(Token::LeadingSpace);
1775  } else if (Tok.is(tok::l_paren)) {
1776    // This is a function-like macro definition.  Read the argument list.
1777    MI->setIsFunctionLike();
1778    if (ReadMacroDefinitionArgList(MI, LastTok)) {
1779      // Forget about MI.
1780      ReleaseMacroInfo(MI);
1781      // Throw away the rest of the line.
1782      if (CurPPLexer->ParsingPreprocessorDirective)
1783        DiscardUntilEndOfDirective();
1784      return;
1785    }
1786
1787    // If this is a definition of a variadic C99 function-like macro, not using
1788    // the GNU named varargs extension, enabled __VA_ARGS__.
1789
1790    // "Poison" __VA_ARGS__, which can only appear in the expansion of a macro.
1791    // This gets unpoisoned where it is allowed.
1792    assert(Ident__VA_ARGS__->isPoisoned() && "__VA_ARGS__ should be poisoned!");
1793    if (MI->isC99Varargs())
1794      Ident__VA_ARGS__->setIsPoisoned(false);
1795
1796    // Read the first token after the arg list for down below.
1797    LexUnexpandedToken(Tok);
1798  } else if (LangOpts.C99 || LangOpts.CPlusPlus11) {
1799    // C99 requires whitespace between the macro definition and the body.  Emit
1800    // a diagnostic for something like "#define X+".
1801    Diag(Tok, diag::ext_c99_whitespace_required_after_macro_name);
1802  } else {
1803    // C90 6.8 TC1 says: "In the definition of an object-like macro, if the
1804    // first character of a replacement list is not a character required by
1805    // subclause 5.2.1, then there shall be white-space separation between the
1806    // identifier and the replacement list.".  5.2.1 lists this set:
1807    //   "A-Za-z0-9!"#%&'()*+,_./:;<=>?[\]^_{|}~" as well as whitespace, which
1808    // is irrelevant here.
1809    bool isInvalid = false;
1810    if (Tok.is(tok::at)) // @ is not in the list above.
1811      isInvalid = true;
1812    else if (Tok.is(tok::unknown)) {
1813      // If we have an unknown token, it is something strange like "`".  Since
1814      // all of valid characters would have lexed into a single character
1815      // token of some sort, we know this is not a valid case.
1816      isInvalid = true;
1817    }
1818    if (isInvalid)
1819      Diag(Tok, diag::ext_missing_whitespace_after_macro_name);
1820    else
1821      Diag(Tok, diag::warn_missing_whitespace_after_macro_name);
1822  }
1823
1824  if (!Tok.is(tok::eod))
1825    LastTok = Tok;
1826
1827  // Read the rest of the macro body.
1828  if (MI->isObjectLike()) {
1829    // Object-like macros are very simple, just read their body.
1830    while (Tok.isNot(tok::eod)) {
1831      LastTok = Tok;
1832      MI->AddTokenToBody(Tok);
1833      // Get the next token of the macro.
1834      LexUnexpandedToken(Tok);
1835    }
1836
1837  } else {
1838    // Otherwise, read the body of a function-like macro.  While we are at it,
1839    // check C99 6.10.3.2p1: ensure that # operators are followed by macro
1840    // parameters in function-like macro expansions.
1841    while (Tok.isNot(tok::eod)) {
1842      LastTok = Tok;
1843
1844      if (Tok.isNot(tok::hash) && Tok.isNot(tok::hashhash)) {
1845        MI->AddTokenToBody(Tok);
1846
1847        // Get the next token of the macro.
1848        LexUnexpandedToken(Tok);
1849        continue;
1850      }
1851
1852      if (Tok.is(tok::hashhash)) {
1853
1854        // If we see token pasting, check if it looks like the gcc comma
1855        // pasting extension.  We'll use this information to suppress
1856        // diagnostics later on.
1857
1858        // Get the next token of the macro.
1859        LexUnexpandedToken(Tok);
1860
1861        if (Tok.is(tok::eod)) {
1862          MI->AddTokenToBody(LastTok);
1863          break;
1864        }
1865
1866        unsigned NumTokens = MI->getNumTokens();
1867        if (NumTokens && Tok.getIdentifierInfo() == Ident__VA_ARGS__ &&
1868            MI->getReplacementToken(NumTokens-1).is(tok::comma))
1869          MI->setHasCommaPasting();
1870
1871        // Things look ok, add the '##' and param name tokens to the macro.
1872        MI->AddTokenToBody(LastTok);
1873        MI->AddTokenToBody(Tok);
1874        LastTok = Tok;
1875
1876        // Get the next token of the macro.
1877        LexUnexpandedToken(Tok);
1878        continue;
1879      }
1880
1881      // Get the next token of the macro.
1882      LexUnexpandedToken(Tok);
1883
1884      // Check for a valid macro arg identifier.
1885      if (Tok.getIdentifierInfo() == 0 ||
1886          MI->getArgumentNum(Tok.getIdentifierInfo()) == -1) {
1887
1888        // If this is assembler-with-cpp mode, we accept random gibberish after
1889        // the '#' because '#' is often a comment character.  However, change
1890        // the kind of the token to tok::unknown so that the preprocessor isn't
1891        // confused.
1892        if (getLangOpts().AsmPreprocessor && Tok.isNot(tok::eod)) {
1893          LastTok.setKind(tok::unknown);
1894        } else {
1895          Diag(Tok, diag::err_pp_stringize_not_parameter);
1896          ReleaseMacroInfo(MI);
1897
1898          // Disable __VA_ARGS__ again.
1899          Ident__VA_ARGS__->setIsPoisoned(true);
1900          return;
1901        }
1902      }
1903
1904      // Things look ok, add the '#' and param name tokens to the macro.
1905      MI->AddTokenToBody(LastTok);
1906      MI->AddTokenToBody(Tok);
1907      LastTok = Tok;
1908
1909      // Get the next token of the macro.
1910      LexUnexpandedToken(Tok);
1911    }
1912  }
1913
1914
1915  // Disable __VA_ARGS__ again.
1916  Ident__VA_ARGS__->setIsPoisoned(true);
1917
1918  // Check that there is no paste (##) operator at the beginning or end of the
1919  // replacement list.
1920  unsigned NumTokens = MI->getNumTokens();
1921  if (NumTokens != 0) {
1922    if (MI->getReplacementToken(0).is(tok::hashhash)) {
1923      Diag(MI->getReplacementToken(0), diag::err_paste_at_start);
1924      ReleaseMacroInfo(MI);
1925      return;
1926    }
1927    if (MI->getReplacementToken(NumTokens-1).is(tok::hashhash)) {
1928      Diag(MI->getReplacementToken(NumTokens-1), diag::err_paste_at_end);
1929      ReleaseMacroInfo(MI);
1930      return;
1931    }
1932  }
1933
1934  MI->setDefinitionEndLoc(LastTok.getLocation());
1935
1936  // Finally, if this identifier already had a macro defined for it, verify that
1937  // the macro bodies are identical, and issue diagnostics if they are not.
1938  if (const MacroInfo *OtherMI=getMacroInfo(MacroNameTok.getIdentifierInfo())) {
1939    // It is very common for system headers to have tons of macro redefinitions
1940    // and for warnings to be disabled in system headers.  If this is the case,
1941    // then don't bother calling MacroInfo::isIdenticalTo.
1942    if (!getDiagnostics().getSuppressSystemWarnings() ||
1943        !SourceMgr.isInSystemHeader(DefineTok.getLocation())) {
1944      if (!OtherMI->isUsed() && OtherMI->isWarnIfUnused())
1945        Diag(OtherMI->getDefinitionLoc(), diag::pp_macro_not_used);
1946
1947      // Warn if defining "__LINE__" and other builtins, per C99 6.10.8/4 and
1948      // C++ [cpp.predefined]p4, but allow it as an extension.
1949      if (OtherMI->isBuiltinMacro())
1950        Diag(MacroNameTok, diag::ext_pp_redef_builtin_macro);
1951      // Macros must be identical.  This means all tokens and whitespace
1952      // separation must be the same.  C99 6.10.3.2.
1953      else if (!OtherMI->isAllowRedefinitionsWithoutWarning() &&
1954               !MI->isIdenticalTo(*OtherMI, *this)) {
1955        Diag(MI->getDefinitionLoc(), diag::ext_pp_macro_redef)
1956          << MacroNameTok.getIdentifierInfo();
1957        Diag(OtherMI->getDefinitionLoc(), diag::note_previous_definition);
1958      }
1959    }
1960    if (OtherMI->isWarnIfUnused())
1961      WarnUnusedMacroLocs.erase(OtherMI->getDefinitionLoc());
1962  }
1963
1964  DefMacroDirective *MD =
1965      appendDefMacroDirective(MacroNameTok.getIdentifierInfo(), MI);
1966
1967  assert(!MI->isUsed());
1968  // If we need warning for not using the macro, add its location in the
1969  // warn-because-unused-macro set. If it gets used it will be removed from set.
1970  if (isInPrimaryFile() && // don't warn for include'd macros.
1971      Diags->getDiagnosticLevel(diag::pp_macro_not_used,
1972          MI->getDefinitionLoc()) != DiagnosticsEngine::Ignored) {
1973    MI->setIsWarnIfUnused(true);
1974    WarnUnusedMacroLocs.insert(MI->getDefinitionLoc());
1975  }
1976
1977  // If the callbacks want to know, tell them about the macro definition.
1978  if (Callbacks)
1979    Callbacks->MacroDefined(MacroNameTok, MD);
1980}
1981
1982/// HandleUndefDirective - Implements \#undef.
1983///
1984void Preprocessor::HandleUndefDirective(Token &UndefTok) {
1985  ++NumUndefined;
1986
1987  Token MacroNameTok;
1988  ReadMacroName(MacroNameTok, 2);
1989
1990  // Error reading macro name?  If so, diagnostic already issued.
1991  if (MacroNameTok.is(tok::eod))
1992    return;
1993
1994  // Check to see if this is the last token on the #undef line.
1995  CheckEndOfDirective("undef");
1996
1997  // Okay, we finally have a valid identifier to undef.
1998  MacroDirective *MD = getMacroDirective(MacroNameTok.getIdentifierInfo());
1999  const MacroInfo *MI = MD ? MD->getMacroInfo() : 0;
2000
2001  // If the callbacks want to know, tell them about the macro #undef.
2002  // Note: no matter if the macro was defined or not.
2003  if (Callbacks)
2004    Callbacks->MacroUndefined(MacroNameTok, MD);
2005
2006  // If the macro is not defined, this is a noop undef, just return.
2007  if (MI == 0) return;
2008
2009  if (!MI->isUsed() && MI->isWarnIfUnused())
2010    Diag(MI->getDefinitionLoc(), diag::pp_macro_not_used);
2011
2012  if (MI->isWarnIfUnused())
2013    WarnUnusedMacroLocs.erase(MI->getDefinitionLoc());
2014
2015  appendMacroDirective(MacroNameTok.getIdentifierInfo(),
2016                       AllocateUndefMacroDirective(MacroNameTok.getLocation()));
2017}
2018
2019
2020//===----------------------------------------------------------------------===//
2021// Preprocessor Conditional Directive Handling.
2022//===----------------------------------------------------------------------===//
2023
2024/// HandleIfdefDirective - Implements the \#ifdef/\#ifndef directive.  isIfndef
2025/// is true when this is a \#ifndef directive.  ReadAnyTokensBeforeDirective is
2026/// true if any tokens have been returned or pp-directives activated before this
2027/// \#ifndef has been lexed.
2028///
2029void Preprocessor::HandleIfdefDirective(Token &Result, bool isIfndef,
2030                                        bool ReadAnyTokensBeforeDirective) {
2031  ++NumIf;
2032  Token DirectiveTok = Result;
2033
2034  Token MacroNameTok;
2035  ReadMacroName(MacroNameTok);
2036
2037  // Error reading macro name?  If so, diagnostic already issued.
2038  if (MacroNameTok.is(tok::eod)) {
2039    // Skip code until we get to #endif.  This helps with recovery by not
2040    // emitting an error when the #endif is reached.
2041    SkipExcludedConditionalBlock(DirectiveTok.getLocation(),
2042                                 /*Foundnonskip*/false, /*FoundElse*/false);
2043    return;
2044  }
2045
2046  // Check to see if this is the last token on the #if[n]def line.
2047  CheckEndOfDirective(isIfndef ? "ifndef" : "ifdef");
2048
2049  IdentifierInfo *MII = MacroNameTok.getIdentifierInfo();
2050  MacroDirective *MD = getMacroDirective(MII);
2051  MacroInfo *MI = MD ? MD->getMacroInfo() : 0;
2052
2053  if (CurPPLexer->getConditionalStackDepth() == 0) {
2054    // If the start of a top-level #ifdef and if the macro is not defined,
2055    // inform MIOpt that this might be the start of a proper include guard.
2056    // Otherwise it is some other form of unknown conditional which we can't
2057    // handle.
2058    if (!ReadAnyTokensBeforeDirective && MI == 0) {
2059      assert(isIfndef && "#ifdef shouldn't reach here");
2060      CurPPLexer->MIOpt.EnterTopLevelIFNDEF(MII);
2061    } else
2062      CurPPLexer->MIOpt.EnterTopLevelConditional();
2063  }
2064
2065  // If there is a macro, process it.
2066  if (MI)  // Mark it used.
2067    markMacroAsUsed(MI);
2068
2069  if (Callbacks) {
2070    if (isIfndef)
2071      Callbacks->Ifndef(DirectiveTok.getLocation(), MacroNameTok, MD);
2072    else
2073      Callbacks->Ifdef(DirectiveTok.getLocation(), MacroNameTok, MD);
2074  }
2075
2076  // Should we include the stuff contained by this directive?
2077  if (!MI == isIfndef) {
2078    // Yes, remember that we are inside a conditional, then lex the next token.
2079    CurPPLexer->pushConditionalLevel(DirectiveTok.getLocation(),
2080                                     /*wasskip*/false, /*foundnonskip*/true,
2081                                     /*foundelse*/false);
2082  } else {
2083    // No, skip the contents of this block.
2084    SkipExcludedConditionalBlock(DirectiveTok.getLocation(),
2085                                 /*Foundnonskip*/false,
2086                                 /*FoundElse*/false);
2087  }
2088}
2089
2090/// HandleIfDirective - Implements the \#if directive.
2091///
2092void Preprocessor::HandleIfDirective(Token &IfToken,
2093                                     bool ReadAnyTokensBeforeDirective) {
2094  ++NumIf;
2095
2096  // Parse and evaluate the conditional expression.
2097  IdentifierInfo *IfNDefMacro = 0;
2098  const SourceLocation ConditionalBegin = CurPPLexer->getSourceLocation();
2099  const bool ConditionalTrue = EvaluateDirectiveExpression(IfNDefMacro);
2100  const SourceLocation ConditionalEnd = CurPPLexer->getSourceLocation();
2101
2102  // If this condition is equivalent to #ifndef X, and if this is the first
2103  // directive seen, handle it for the multiple-include optimization.
2104  if (CurPPLexer->getConditionalStackDepth() == 0) {
2105    if (!ReadAnyTokensBeforeDirective && IfNDefMacro && ConditionalTrue)
2106      CurPPLexer->MIOpt.EnterTopLevelIFNDEF(IfNDefMacro);
2107    else
2108      CurPPLexer->MIOpt.EnterTopLevelConditional();
2109  }
2110
2111  if (Callbacks)
2112    Callbacks->If(IfToken.getLocation(),
2113                  SourceRange(ConditionalBegin, ConditionalEnd));
2114
2115  // Should we include the stuff contained by this directive?
2116  if (ConditionalTrue) {
2117    // Yes, remember that we are inside a conditional, then lex the next token.
2118    CurPPLexer->pushConditionalLevel(IfToken.getLocation(), /*wasskip*/false,
2119                                   /*foundnonskip*/true, /*foundelse*/false);
2120  } else {
2121    // No, skip the contents of this block.
2122    SkipExcludedConditionalBlock(IfToken.getLocation(), /*Foundnonskip*/false,
2123                                 /*FoundElse*/false);
2124  }
2125}
2126
2127/// HandleEndifDirective - Implements the \#endif directive.
2128///
2129void Preprocessor::HandleEndifDirective(Token &EndifToken) {
2130  ++NumEndif;
2131
2132  // Check that this is the whole directive.
2133  CheckEndOfDirective("endif");
2134
2135  PPConditionalInfo CondInfo;
2136  if (CurPPLexer->popConditionalLevel(CondInfo)) {
2137    // No conditionals on the stack: this is an #endif without an #if.
2138    Diag(EndifToken, diag::err_pp_endif_without_if);
2139    return;
2140  }
2141
2142  // If this the end of a top-level #endif, inform MIOpt.
2143  if (CurPPLexer->getConditionalStackDepth() == 0)
2144    CurPPLexer->MIOpt.ExitTopLevelConditional();
2145
2146  assert(!CondInfo.WasSkipping && !CurPPLexer->LexingRawMode &&
2147         "This code should only be reachable in the non-skipping case!");
2148
2149  if (Callbacks)
2150    Callbacks->Endif(EndifToken.getLocation(), CondInfo.IfLoc);
2151}
2152
2153/// HandleElseDirective - Implements the \#else directive.
2154///
2155void Preprocessor::HandleElseDirective(Token &Result) {
2156  ++NumElse;
2157
2158  // #else directive in a non-skipping conditional... start skipping.
2159  CheckEndOfDirective("else");
2160
2161  PPConditionalInfo CI;
2162  if (CurPPLexer->popConditionalLevel(CI)) {
2163    Diag(Result, diag::pp_err_else_without_if);
2164    return;
2165  }
2166
2167  // If this is a top-level #else, inform the MIOpt.
2168  if (CurPPLexer->getConditionalStackDepth() == 0)
2169    CurPPLexer->MIOpt.EnterTopLevelConditional();
2170
2171  // If this is a #else with a #else before it, report the error.
2172  if (CI.FoundElse) Diag(Result, diag::pp_err_else_after_else);
2173
2174  if (Callbacks)
2175    Callbacks->Else(Result.getLocation(), CI.IfLoc);
2176
2177  // Finally, skip the rest of the contents of this block.
2178  SkipExcludedConditionalBlock(CI.IfLoc, /*Foundnonskip*/true,
2179                               /*FoundElse*/true, Result.getLocation());
2180}
2181
2182/// HandleElifDirective - Implements the \#elif directive.
2183///
2184void Preprocessor::HandleElifDirective(Token &ElifToken) {
2185  ++NumElse;
2186
2187  // #elif directive in a non-skipping conditional... start skipping.
2188  // We don't care what the condition is, because we will always skip it (since
2189  // the block immediately before it was included).
2190  const SourceLocation ConditionalBegin = CurPPLexer->getSourceLocation();
2191  DiscardUntilEndOfDirective();
2192  const SourceLocation ConditionalEnd = CurPPLexer->getSourceLocation();
2193
2194  PPConditionalInfo CI;
2195  if (CurPPLexer->popConditionalLevel(CI)) {
2196    Diag(ElifToken, diag::pp_err_elif_without_if);
2197    return;
2198  }
2199
2200  // If this is a top-level #elif, inform the MIOpt.
2201  if (CurPPLexer->getConditionalStackDepth() == 0)
2202    CurPPLexer->MIOpt.EnterTopLevelConditional();
2203
2204  // If this is a #elif with a #else before it, report the error.
2205  if (CI.FoundElse) Diag(ElifToken, diag::pp_err_elif_after_else);
2206
2207  if (Callbacks)
2208    Callbacks->Elif(ElifToken.getLocation(),
2209                    SourceRange(ConditionalBegin, ConditionalEnd), CI.IfLoc);
2210
2211  // Finally, skip the rest of the contents of this block.
2212  SkipExcludedConditionalBlock(CI.IfLoc, /*Foundnonskip*/true,
2213                               /*FoundElse*/CI.FoundElse,
2214                               ElifToken.getLocation());
2215}
2216